Re-ranking by optimizing NDCG measure.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "toc": true
   },
   "source": [
    "<h1>Table of Contents<span class=\"tocSkip\"></span></h1>\n",
    "<div class=\"toc\"><ul class=\"toc-item\"><li><span><a href=\"#Task\" data-toc-modified-id=\"Task-1\"><span class=\"toc-item-num\">1&nbsp;&nbsp;</span>Task</a></span></li><li><span><a href=\"#Approach\" data-toc-modified-id=\"Approach-2\"><span class=\"toc-item-num\">2&nbsp;&nbsp;</span>Approach</a></span></li><li><span><a href=\"#Research\" data-toc-modified-id=\"Research-3\"><span class=\"toc-item-num\">3&nbsp;&nbsp;</span>Research</a></span><ul class=\"toc-item\"><li><span><a href=\"#Definitions\" data-toc-modified-id=\"Definitions-3.1\"><span class=\"toc-item-num\">3.1&nbsp;&nbsp;</span>Definitions</a></span><ul class=\"toc-item\"><li><span><a href=\"#Learning-to-Rank-methods-overview\" data-toc-modified-id=\"Learning-to-Rank-methods-overview-3.1.1\"><span class=\"toc-item-num\">3.1.1&nbsp;&nbsp;</span>Learning-to-Rank methods overview</a></span></li><li><span><a href=\"#NDCG-measure\" data-toc-modified-id=\"NDCG-measure-3.1.2\"><span class=\"toc-item-num\">3.1.2&nbsp;&nbsp;</span>NDCG measure</a></span></li></ul></li><li><span><a href=\"#Initial-findings\" data-toc-modified-id=\"Initial-findings-3.2\"><span class=\"toc-item-num\">3.2&nbsp;&nbsp;</span>Initial findings</a></span><ul class=\"toc-item\"><li><span><a href=\"#Overall-Research-Papers-and-other-Sources\" data-toc-modified-id=\"Overall-Research-Papers-and-other-Sources-3.2.1\"><span class=\"toc-item-num\">3.2.1&nbsp;&nbsp;</span>Overall Research Papers and other Sources</a></span></li><li><span><a href=\"#Ranking-Metric-Optimization\" data-toc-modified-id=\"Ranking-Metric-Optimization-3.2.2\"><span class=\"toc-item-num\">3.2.2&nbsp;&nbsp;</span>Ranking Metric Optimization</a></span></li><li><span><a href=\"#Unbiased-Learning-to-Rank\" data-toc-modified-id=\"Unbiased-Learning-to-Rank-3.2.3\"><span class=\"toc-item-num\">3.2.3&nbsp;&nbsp;</span>Unbiased Learning-to-Rank</a></span></li><li><span><a href=\"#Experiments-that-could-follow-after-initial-POC\" data-toc-modified-id=\"Experiments-that-could-follow-after-initial-POC-3.2.4\"><span class=\"toc-item-num\">3.2.4&nbsp;&nbsp;</span>Experiments that could follow after initial POC</a></span></li></ul></li></ul></li><li><span><a href=\"#Model-implementation\" data-toc-modified-id=\"Model-implementation-4\"><span class=\"toc-item-num\">4&nbsp;&nbsp;</span>Model implementation</a></span><ul class=\"toc-item\"><li><span><a href=\"#Library-imports\" data-toc-modified-id=\"Library-imports-4.1\"><span class=\"toc-item-num\">4.1&nbsp;&nbsp;</span>Library imports</a></span></li><li><span><a href=\"#Preparing-and-looking-at-dataset\" data-toc-modified-id=\"Preparing-and-looking-at-dataset-4.2\"><span class=\"toc-item-num\">4.2&nbsp;&nbsp;</span>Preparing and looking at dataset</a></span></li><li><span><a href=\"#Data-modeling\" data-toc-modified-id=\"Data-modeling-4.3\"><span class=\"toc-item-num\">4.3&nbsp;&nbsp;</span>Data modeling</a></span><ul class=\"toc-item\"><li><span><a href=\"#Defining-train/test-subsets\" data-toc-modified-id=\"Defining-train/test-subsets-4.3.1\"><span class=\"toc-item-num\">4.3.1&nbsp;&nbsp;</span>Defining train/test subsets</a></span></li><li><span><a href=\"#Defining-'y'-label\" data-toc-modified-id=\"Defining-'y'-label-4.3.2\"><span class=\"toc-item-num\">4.3.2&nbsp;&nbsp;</span>Defining 'y' label</a></span></li><li><span><a href=\"#Scaling-data-numerics\" data-toc-modified-id=\"Scaling-data-numerics-4.3.3\"><span class=\"toc-item-num\">4.3.3&nbsp;&nbsp;</span>Scaling data numerics</a></span></li><li><span><a href=\"#Defining-final-train/test-sets\" data-toc-modified-id=\"Defining-final-train/test-sets-4.3.4\"><span class=\"toc-item-num\">4.3.4&nbsp;&nbsp;</span>Defining final train/test sets</a></span></li><li><span><a href=\"#Building-ranker\" data-toc-modified-id=\"Building-ranker-4.3.5\"><span class=\"toc-item-num\">4.3.5&nbsp;&nbsp;</span>Building ranker</a></span></li><li><span><a href=\"#Predicting-and-re-ranking\" data-toc-modified-id=\"Predicting-and-re-ranking-4.3.6\"><span class=\"toc-item-num\">4.3.6&nbsp;&nbsp;</span>Predicting and re-ranking</a></span></li></ul></li></ul></li></ul></div>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br><br><br>\n",
    "# Task"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Using provided Learning to Rank dataset, the **task is to engineer a machine learning model\n",
    "that re-ranks documents optimising NDCG measure** (relevancy being approximated by ``clicked`` column).\n",
    "\n",
    "The dataset is provided in the form of a CSV file, which consists of the following columns:\n",
    "* listing identifier – ``listing_id``\n",
    "* rank of a document in the listing – ``doc_rank``\n",
    "* whether a document was relevant – ``clicked``\n",
    "* numerical features – ``n0`` through ``n44``\n",
    "* categorical features – ``c0`` through ``c4``"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br><br><br>\n",
    "# Approach"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As a Proof-of-Concept model a **XGBRanker** from XGBoost library v.0.81 was used. With objective defined as ``rank:ndcg`` parameter it uses *LambdaMART* algorithm to perform list-wise ranking where NDCG is maximized.<br>\n",
    "\n",
    "It is an implementation of the Scikit-Learn API for XGBoost Ranking and is based on XGBModel.\n",
    "\n",
    "**Note**: Confined to smaller scales, models like GBDT are arguably at par in performance and easier to handle than Deep NN aproach, and it is relevant to use them for focused medium sized problems."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br><br><br><br><br><br>\n",
    "# Research\n",
    "/ method, definitions, research papers, initial thoughts, concepts /"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Definitions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Learning-to-Rank methods overview"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Methods**\n",
    "\n",
    "* **Pointwise**<br>\n",
    "Pointwise approaches look at a *single document* at a time in the loss function. They essentially take a single document and train a classifier / regressor on it to predict how relevant it is for the current query. The final ranking is achieved by simply sorting the result list by these document scores. For pointwise approaches, the score for each document is independent of the other documents that are in the result list for the query.\n",
    "*All the standard regression and classification algorithms can be directly used for pointwise learning to rank.*<br><br>\n",
    "* **Pairwise**<br>\n",
    "Pairwise approaches look at a *pair of documents* at a time in the loss function. Given a pair of documents, they try and come up with the optimal ordering for that pair and compare it to the ground truth. The goal for the ranker is to minimize the number of inversions in ranking i.e. cases where the pair of results are in the wrong order relative to the ground truth.<br>\n",
    "Pairwise approaches work better in practice than pointwise approaches because predicting relative order is closer to the nature of ranking than predicting class label or relevance score. *Some of the most popular Learning to Rank algorithms like RankNet, LambdaRank and LambdaMART are pairwise approaches.*\n",
    "<br><br>\n",
    "* **Listwise**<br>\n",
    "Listwise approaches directly look at the entire *list of documents* and try to come up with the optimal ordering for it. There are 2 main sub-techniques for doing listwise Learning to Rank:\n",
    "    - Direct optimization of Information Retrieval measures such as NDCG. E.g. *SoftRank, AdaRank*\n",
    "    - Minimize a loss function that is defined based on understanding the unique properties of the kind of ranking you are trying to achieve. E.g. *ListNet, ListMLE*\n",
    "<br>\n",
    "\n",
    "source: https://medium.com/@nikhilbd/pointwise-vs-pairwise-vs-listwise-learning-to-rank-80a8fe8fadfd"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Sample algorithms**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* **RankNet**<br>\n",
    "RankNet was originally developed using neural nets, but the underlying model can be different and is not constrained to just neural nets. The cost function for RankNet aims to minimize the number of inversions in ranking. Here an inversion means an incorrect order among a pair of results, i.e. when we rank a lower rated result above a higher rated result in a ranked list. RankNet optimizes the cost function using Stochastic Gradient Descent.<br><br>\n",
    "\n",
    "* **LambdaRank**<br>\n",
    "It was found that during RankNet training procedure, you don't need the costs, only need the gradients (λ) of the cost with respect to the model score. One can think of these gradients as little arrows attached to each document in the ranked list, indicating the direction we’d like those documents to move.<br>\n",
    "Further it was found that scaling the gradients by the change in NDCG found by swapping each pair of documents gave good results. The core idea of LambdaRank is to use this new cost function for training a RankNet. On experimental datasets, this shows both speed and accuracy improvements over the original RankNet.<br><br>\n",
    "\n",
    "* **LambdaMART**<br>\n",
    "LambdaMART combines LambdaRank and MART (Multiple Additive Regression Trees). While MART uses gradient boosted decision trees for prediction tasks, LambdaMART uses gradient boosted decision trees using a cost function derived from LambdaRank for solving a ranking task. On experimental datasets, LambdaMART has shown better results than LambdaRank and the original RankNet.\n",
    "\n",
    "source: https://www.quora.com/What-is-the-intuitive-explanation-of-Learning-to-Rank-and-algorithms-like-RankNet-LambdaRank-and-LambdaMART-In-what-types-of-data-variables-can-these-techniques-be-used-What-are-their-strengths-and-limitations/answer/Nikhil-Dandekar"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Papers**\n",
    "\n",
    "* **From RankNet to LambdaRank to LambdaMART: An Overview** (Microsoft Research Technical Report MSR-TR-2010-82) -https://www.microsoft.com/en-us/research/publication/from-ranknet-to-lambdarank-to-lambdamart-an-overview/ / https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/MSR-TR-2010-82.pdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### NDCG measure\n",
    "(definition: https://en.wikipedia.org/wiki/Discounted_cumulative_gain)<br><br>\n",
    "NDCG (normalised discounted cumulative gain) is a single-number measure of effectiveness of a ranking algorithm that allows non-binary relevance judgments. It has a strong relationship to the average precision (AP), but has a discounting function with a very heavy tail that must be truncated.<br>\n",
    "\n",
    "NDCG has two advantages compared to many other measures: \n",
    "- allows each retrieved document has graded relevance while most traditional ranking measures only allow binary relevance. That is, each document is viewed as either relevant or not relevant by previous ranking measures, while there can be degrees of relevancy for documents in NDCG.\n",
    "- involves a discount function over the rank while many other measures uniformly weight all positions. This feature is particularly important for search engines as users care top ranked documents much more than others.\n",
    "\n",
    "NDCG is a normalization of the Discounted Cumulative Gain (DCG) measure. (For formal definition of both DCG and NDCG, please see Section 2.) DCG is a weighted sum of the degree of relevancy of the ranked items. The weight is a decreasing function of the rank (position) of the object, and therefore called discount.\n",
    "<br>(...)<br>\n",
    "Search engine systems also use a cut-off top-k version of NDCG. That is, the discount is set to be zero for ranks larger than k. Such NDCG measure is usually referred to as NDCG@k.\n",
    "<br><br>\n",
    "source: *A Theoretical Analysis of NDCG Ranking Measures* (in papers section below)\n",
    "\n",
    "<br><br>\n",
    "**Other measures**: MAP (Mean Average Precision), MRR (Mean Reciprocal Rank), AUC (Area Under the ROC Curve), ERR (Expected reciprocal rank), Yandex's pfound"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Papers**\n",
    "\n",
    "* **A Theoretical Analysis of NDCG Ranking Measures** (PMLR Volume 30: Conference on Learning Theory, 12-14 June 2013) - http://proceedings.mlr.press/v30/Wang13.pdf\n",
    "* **Metric Learning to Rank** (The 27th International Conference on Machine Learning (ICML 2010)) - https://icml.cc/Conferences/2010/papers/504.pdf / https://bmcfee.github.io/papers/mlr.pdf\n",
    "* **Learning to Rank by Optimizing NDCG Measure** (NIPS 2009) - https://papers.nips.cc/paper/3758-learning-to-rank-by-optimizing-ndcg-measure / https://papers.nips.cc/paper/3758-learning-to-rank-by-optimizing-ndcg-measure.pdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br>\n",
    "## Initial findings"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Overall Research Papers and other Sources"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* **Hashing as Tie-Aware Learning to Rank** (Optimizing NDCG and AP; 2018)  - https://arxiv.org/abs/1705.08562v4\n",
    "\n",
    "* **TF-Ranking: A Scalable TensorFlow Library for Learning-to-Rank** (2018) - https://ai.googleblog.com/2018/12/tf-ranking-scalable-tensorflow-library.html / https://arxiv.org/abs/1812.00073\n",
    "\n",
    "* **Learning Groupwise Scoring Functions Using Deep Neural Networks** - https://arxiv.org/abs/1811.04415\n",
    "\n",
    "* **Direct Learning to Rank and Rerank** (Proceedings of Machine Learning Research, Volume 84: International Conference on Artificial Intelligence and Statistics, 9-11 April 2018) http://proceedings.mlr.press/v84/rudin18a.html / http://proceedings.mlr.press/v84/rudin18a/rudin18a.pdf\n",
    "\n",
    "* **Learning to rank @ allegro.pl** - https://www.youtube.com/watch?v=O7v9u_gSy7Y"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Ranking Metric Optimization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "An important research challenge in learning-to-rank is direct optimization of ranking metrics (such as the previously mentioned NDCG and MRR). These metrics, while being able to measure the performance of ranking systems better than the standard classification metrics like Area Under the Curve (AUC), have the unfortunate property of being either discontinuous or flat. Therefore standard stochastic gradient descent optimization of these metrics is problematic.\n",
    "<br><br>\n",
    "In recent work, Google proposed a novel method, *LambdaLoss*, which provides a principled probabilistic framework for ranking metric optimization. In this framework, metric-driven loss functions can be designed and optimized by an expectation-maximization procedure. The TF-Ranking library integrates the recent advances in direct metric optimization and provides an implementation of LambdaLoss.\n",
    "\n",
    "source: https://ai.googleblog.com/2018/12/tf-ranking-scalable-tensorflow-library.html\n",
    "\n",
    "**Papers**\n",
    "* The LambdaLoss Framework for Ranking Metric Optimization - https://ai.google/research/pubs/pub47258 / https://ai.google/research/pubs/pub47258.pdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Unbiased Learning-to-Rank"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prior research has shown that given a ranked list of items, users are much more likely to interact with the first few results, regardless of their relevance. This observation has inspired research interest in unbiased learning-to-rank, and led to the development of unbiased evaluation and several unbiased learning algorithms, based on training instances re-weighting. In the Google's TF-Ranking library, metrics are implemented to support unbiased evaluation and losses are implemented for unbiased learning by natively supporting re-weighting to overcome the inherent biases in user interactions datasets. \n",
    "\n",
    "source: https://ai.googleblog.com/2018/12/tf-ranking-scalable-tensorflow-library.html\n",
    "\n",
    "**Papers**\n",
    "* Addressing Trust Bias for Unbiased Learning-to-Rank (2019) - https://ai.google/research/pubs/pub47859.pdf\n",
    "* Position Bias Estimation for Unbiased Learning to Rank in Personal Search (2018) - https://ai.google/research/pubs/pub46485.pdf\n",
    "* Counterfactual Learning-to-Rank for Additive Metrics and Deep (2018) - http://www.cs.cornell.edu/people/tj/publications/agarwal_etal_18b.pdf\n",
    "* Unbiased Learning-to-Rank with Biased Feedback (2017) - http://www.cs.cornell.edu/people/tj/publications/joachims_etal_17a.pdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Experiments that could follow after initial POC"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Further hyperparameters tuning with XGBRanker\n",
    "<br><br>\n",
    "* Checking for position/presentation bias\n",
    "    - **Paper:** Unbiased Learning-to-Rank with Biased Feedback (2017) - http://www.cs.cornell.edu/people/tj/publications/joachims_etal_17a.pdf\n",
    "<br><br>\n",
    "\n",
    "* Combine search phrases with image (transfer learning with CNNs trained on ImageNet) and text (Word2vec/GloVe embeddings)\n",
    "* Combine NN with LambdaMART algorithm\n",
    "* Using listing ids as features as index into an embedding, which allowes to learn a vector representation per listing, encoding their unique properties.\n",
    "    - **Paper:** Applying Deep Learning To Airbnb Search (2018) - https://arxiv.org/pdf/1810.09591.pdf\n",
    "<br><br>\n",
    "\n",
    "* Trying out the recent Google's TF-Ranking library\n",
    "    - **GitHub:** https://github.com/tensorflow/ranking\n",
    "<br><br>\n",
    "\n",
    "* Working on Counterfactual Learning-to-Rank framework\n",
    "    - **Paper:** http://www.cs.cornell.edu/people/tj/publications/agarwal_etal_18b.pdf <br>\n",
    "    Focusing on DCG, this conceptual generalization gives rise to two new learning methods that both directly optimize an unbiased estimate of DCG despite the bias in the implicit feedback data. The first, SVM PropDCG, generalizes the Propensity Ranking SVM (SVM PropRank), and we show how the resulting optimization problem can be addressed via the Convex Concave Procedure (CCP). The second, Deep PropDCG, further generalizes the counterfactual learning-to-rank approach to deep networks as non-linear ranking functions. In addition to the theoretical support, we empirically find that SVM PropDCG significantly outperforms SVM PropRank in terms of DCG, and that it is robust to varying severity of presentation bias, noise, and propensity-model misspecification. Moreover, the ability to train non-linear ranking functions via Deep PropDCG further improves DCG.\n",
    "<br><br>\n",
    "\n",
    "* Using Learning-to-Rank with Elasticsearch\n",
    "    - **Source:** https://opensourceconnections.com/blog/2017/02/14/elasticsearch-learning-to-rank/\n",
    "    - **GitHub:** https://github.com/o19s/elasticsearch-learning-to-rank"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br><br><br><br><br><br>\n",
    "# Model implementation "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Library imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Standard libs\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Model libs\n",
    "from sklearn.preprocessing import MinMaxScaler, StandardScaler\n",
    "# from sklearn.model_selection import cross_val_score, GroupKFold\n",
    "# from sklearn.metrics import accuracy_score, make_scorer\n",
    "\n",
    "from xgboost import XGBRanker\n",
    "from xgboost.callback import early_stop, print_evaluation\n",
    "from xgboost import plot_importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "XGBRanker v.0.81\n"
     ]
    }
   ],
   "source": [
    "# XGBRanker is implemented in XGBoost version >= 0.81\n",
    "import xgboost as xgb\n",
    "print('XGBRanker v.' + xgb.__version__)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br>\n",
    "## Preparing and looking at dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Loading dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('data/dataset_v2.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Shape of dataset. There are 100k listings with overall ~4.9m presented documents."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data shape: (4885214, 53)\n",
      "Number of listings: 100000\n"
     ]
    }
   ],
   "source": [
    "print('Data shape:', data.shape)\n",
    "print('Number of listings:', data['listing_id'].nunique())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Simple statistics for listing_id. The listings range from 2 to 60 presented documents in each with mean of ~49 per listing."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    100000.000000\n",
       "mean         48.852140\n",
       "std          19.479057\n",
       "min           2.000000\n",
       "25%          43.000000\n",
       "50%          60.000000\n",
       "75%          60.000000\n",
       "max          60.000000\n",
       "Name: listing_id, dtype: float64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['listing_id'].value_counts().describe()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Listings differ in presentation volume and a 2/3 of them contain 60 documents."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEXCAYAAABWNASkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvIxREBQAAIABJREFUeJzt3XuYXXV97/H3Z88lt0kmxJlwSUIu\nJKLhagkgB6to8Qi2hT6tFpSeilWpVaqt2gpqKaVarfZ4OS09Fi1asYiAF1JOlFYEBY8gQcIlhEBI\nIpkEkkzu17l++8daM9nZ2TOzZmbPzJ6Vz+t59jN7rfXbv/Xdl/nstX9r7bUVEZiZWb4UxroAMzOr\nPIe7mVkOOdzNzHLI4W5mlkMOdzOzHHK4m5nlkMO9Ckj6sqS/qlBfJ0raK6kmnb5f0rsr0Xfa3w8k\nvaNS/Q1ivZ+U1CrppYztQ9LCka7raCfp65I+mV7/dUmrh9jPmLyu8qx2rAvIO0nrgWOBTqALeBr4\nBnBTRHQDRMR7B9HXuyPiR321iYgXgIbhVd27vuuBhRHxB0X9X1yJvgdZxxzgw8DciNgy2usfa+We\nh2oUEQ8AJw/UrlpeV3nnLffR8dsRMRWYC3wG+Cjwr5VeiaS8vlnPBbYdjcFeTXL8+sqniPBlBC/A\neuDCknnnAN3Aqen014FPptebgLuBncB24AGSN+Fb0tscAPYCfwnMAwJ4F/AC8NOiebVpf/cDnwZ+\nAewC7gJmpMsuAFrK1QtcBLQDHen6Hi/q793p9QLwCeBXwBaSTySN6bKeOt6R1tYKfLyfx6kxvf3W\ntL9PpP1fmN7n7rSOr/dx+78AXgQ2AX+Urnthf30X3fY9wCpgD8knq19L5/f2UeZ5ugBoSZ+HLem6\nfwd4M/Bs+tx9rOi2BeAa4HlgG3B70fPQ52PVz/NwJbA2rXkdcEUfj8v1wJ3At9O2vwTOKFp+AvCd\n9LFZB3ygzG2/Cezued5L+j/iMSla9lFgY7re1cBv9HN/7ufQ6+pK4EHgH4AdaV0XF/U7n+S1vgf4\nEXAj8M2x/l+vtsuYF5D3C2XCPZ3/AvAn6fXif5BPA18G6tLLrwMq11dRKHwDmAJMony4bwROTdt8\np+cfofSfsXQd6T/3N0uWF/8T/hGwBlhAMhT0XeCWktq+ktZ1BtAGvLKPx+kbJG88U9PbPgu8q686\nS257EbC56D7eyuHh3l/fb00fn7MBAQtJhn9g4HDvBK5Ln6f3kATkrel6TgEOAgvS9n8GPATMBiYA\n/wJ8K8tjVfo8pPdxN3ByOn08cEofj831JEH6lrTOj5CEZR3JG86j6X2oT5/HtcCbSm77O2nbSWX6\nL31MWtLrJwMbgBOK7uNJGV9XV6brfQ9QA/wJyZt2z//Bz0mCvx54TfpYONxLLh6WGTubgBll5neQ\n/LPOjYiOiHgg0ld0P66PiH0RcaCP5bdExFMRsQ/4K+D3e3a4DtMVwOcjYm1E7AWuBS4v+fj+NxFx\nICIeBx4nCa7DpLVcBlwbEXsiYj3wv4H/lbGO3we+VnQfrx9E3+8GPhsRj0RiTUT8KuN6O4BPRUQH\ncBvJp64vpetZCawETk/b/jHJ1nhLRLSlNb5lsI9VkW7gVEmTIuLFdH19eTQi7kzr/DwwEXg1yRta\nc0TcEBHtEbGW5A3m8qLb/jwivh8R3f28vsrpInkTWyypLiLWR8Tzg7j9ryLiKxHRBfwbyf/EsZJO\nTOu+Lq35QWDpIPo9ajjcx84sko/upT5HsjX8n5LWSromQ18bBrH8VyRbbU2ZquzfCWl/xX3XkuxA\n7lF8dMt+yu/sbSLZCivta9Yg6ii9j1n7nkMyVDIU29LwgWToCJJPEBTN67m/c4HvSdopaSfJMFAX\ng3+sSN/ALgPeC7wo6f9JekU/dfY+NpHsxG8heczmAif01JTW9bGSmgZ6bZUVEWtIPq1cD2yRdJuk\nEwbRRe9jERH706sNad3bi+YNuca8c7iPAUlnk4TLg6XL0q2+D0fEAuC3gQ9J+o2exX10OdCW/Zyi\n6yeSbHG2AvuAyUV11QDNg+h3E0lAFPfdyeEBl0VrWlNpXxsz3v5FjryPWfveAJzUR7/7KXp8gOMy\n1lPOBpJx4+lFl4kRkeU+HvE8RMQ9EfFGki3aZ0i2uPvS+9hIKpAMDW1Ka1pXUtPUiHhzf+vOKiJu\njYjXkDz2Afz9cPskea5nSCp+Xub01fho5nAfRZKmSfotko/w34yIJ8u0+S1JCyWJZCyxK71AEpoL\nhrDqP5C0OP2HuAG4M93ifBaYKOk3JdWR7GicUHS7zcC8NBDK+Rbw55LmS2oA/g74dkR0Dqa4tJbb\ngU9JmippLvAhkh15WdwOXFl0H/96EH1/FfiIpLOUWJi2AVgBvF1SjaSLgNcN5n6V+HJaw1wASc2S\nLs1428OeB0nHSrpE0hSSsfm9HHqNlHOWpN9Nh4D+LL3NQyQ72XdL+qikSen9PDXd+BgWSSdLeoOk\nCST7Hg5w+Ou4v9dVn9Ihs+XA9ZLqJZ1HshFkJRzuo+M/JO0h2VL6OMm45zv7aLuI5AiAvSQ7jv45\nIu5Pl30a+ET6Efojg1j/LSQ7vl4iGW/9AEBE7ALeRxJwG0m25FuKbndH+nebpF+W6ffmtO+fkuyk\nOwj86SDqKvan6frXknyiuTXtf0AR8QPgi8CPSYa0fpy174i4A/hUOm8P8H0O7Qv5IElw7CTZv/D9\nId2zxJdIxob/M30tPAScm/G2pc9DgeS4/00kQ3uvI3ke+3IXyTDODpJ9Db+b7s/pIrl/Z5I8f60k\nr4XGQdyvvkwgOey3leR1N5NkyKfc/RmsK4DzSI46+iTJkUBtw6o2h3r2PptZDo2XL0ANh6RvA89E\nxF8P2Pgo4i13MxtXJJ0t6SRJhXS47FKG96kql/yNMzMbb44j+U7Fy0iGEf8kIh4b25Kqj4dlzMxy\nyMMyZmY5NGbDMk1NTTFv3ryxWr2Z2bj06KOPtkZE80Dtxizc582bx/Lly8dq9WZm45KkTKfH8LCM\nmVkOOdzNzHLI4W5mlkMOdzOzHHK4m5nlkMPdzCyHHO5mZjnkcDczyyGHu5lZDmX6hmp6Ws0vkfwS\n+Vcj4jNl2vw+ye8lBvB4RLy9gnWamY2aWx9+4Yh5bz/3xDItq9eA4Z7+ruaNwBtJTq/5iKSlEfF0\nUZtFwLXA+RGxQ9LMkSrYzMwGlmVY5hxgTUSsjYh2kt//LP3tx/cAN0bEDoCI2FLZMs3MbDCyhPss\nkt/+7NGSziv2cuDlkn4m6aF0GMfMzMZIljF3lZlX+gsftSQ/7HwBMBt4QNKpEbHzsI6kq4CrAE48\ncXyNX5mZjSdZttxbgDlF07NJfnW9tM1d6S+qrwNWk4T9YSLipohYEhFLmpsHPB2xmZkNUZZwfwRY\nJGm+pHrgcmBpSZvvA68HkNREMkyztpKFmplZdgOGe0R0AlcD9wCrgNsjYqWkGyRdkja7B9gm6Wng\nPuAvImLbSBVtZmb9y3Sce0QsA5aVzLuu6HoAH0ovZmY2xvwNVTOzHHK4m5nlkMPdzCyHHO5mZjnk\ncDczyyGHu5lZDjnczcxyyOFuZpZDDnczsxxyuJuZ5ZDD3cwshxzuZmY55HA3M8shh7uZWQ453M3M\ncsjhbmaWQw53M7MccribmeWQw93MLIcc7mZmOeRwNzPLIYe7mVkOOdzNzHLI4W5mlkMOdzOzHMoU\n7pIukrRa0hpJ15RZfqWkrZJWpJd3V75UMzPLqnagBpJqgBuBNwItwCOSlkbE0yVNvx0RV49AjWZm\nNkhZttzPAdZExNqIaAduAy4d2bLMzGw4soT7LGBD0XRLOq/U70l6QtKdkuZUpDozMxuSLOGuMvOi\nZPo/gHkRcTrwI+DfynYkXSVpuaTlW7duHVylZmaWWZZwbwGKt8RnA5uKG0TEtohoSye/ApxVrqOI\nuCkilkTEkubm5qHUa2ZmGWQJ90eARZLmS6oHLgeWFjeQdHzR5CXAqsqVaGZmgzXg0TIR0SnpauAe\noAa4OSJWSroBWB4RS4EPSLoE6AS2A1eOYM1mZjaAAcMdICKWActK5l1XdP1a4NrKlmZmZkPlb6ia\nmeWQw93MLIcc7mZmOeRwNzPLIYe7mVkOOdzNzHLI4W5mlkMOdzOzHHK4m5nlkMPdzCyHHO5mZjnk\ncDczyyGHu5lZDjnczcxyyOFuZpZDDnczsxxyuJuZ5ZDD3cwshxzuZmY55HA3M8shh7uZWQ453M3M\ncsjhbmaWQw53M7MccribmeVQpnCXdJGk1ZLWSLqmn3ZvkRSSllSuRDMzG6wBw11SDXAjcDGwGHib\npMVl2k0FPgA8XOkizcxscLJsuZ8DrImItRHRDtwGXFqm3d8CnwUOVrA+MzMbgizhPgvYUDTdks7r\nJelVwJyIuLu/jiRdJWm5pOVbt24ddLFmZpZNlnBXmXnRu1AqAF8APjxQRxFxU0QsiYglzc3N2as0\nM7NByRLuLcCcounZwKai6anAqcD9ktYDrwaWeqeqmdnYyRLujwCLJM2XVA9cDiztWRgRuyKiKSLm\nRcQ84CHgkohYPiIVm5nZgAYM94joBK4G7gFWAbdHxEpJN0i6ZKQLNDOzwavN0igilgHLSuZd10fb\nC4ZflpmZDYe/oWpmlkMOdzOzHHK4m5nlkMPdzCyHHO5mZjnkcDczyyGHu5lZDjnczcxyyOFuZpZD\nDnczsxxyuJuZ5ZDD3cwshxzuZmY55HA3M8shh7uZWQ453M3McsjhbmaWQw53M7MccribmeWQw93M\nLIcc7mZmOeRwNzPLIYe7mVkOOdzNzHLI4W5mlkOZwl3SRZJWS1oj6Zoyy98r6UlJKyQ9KGlx5Us1\nM7OsBgx3STXAjcDFwGLgbWXC+9aIOC0izgQ+C3y+4pWamVlmWbbczwHWRMTaiGgHbgMuLW4QEbuL\nJqcAUbkSzcxssGoztJkFbCiabgHOLW0k6f3Ah4B64A3lOpJ0FXAVwIknnjjYWs3MLKMsW+4qM++I\nLfOIuDEiTgI+CnyiXEcRcVNELImIJc3NzYOr1MzMMssS7i3AnKLp2cCmftrfBvzOcIoyM7PhyRLu\njwCLJM2XVA9cDiwtbiBpUdHkbwLPVa5EMzMbrAHH3COiU9LVwD1ADXBzRKyUdAOwPCKWAldLuhDo\nAHYA7xjJos3MrH9ZdqgSEcuAZSXzriu6/sEK12VmZsPgb6iameWQw93MLIcc7mZmOeRwNzPLIYe7\nmVkOOdzNzHLI4W5mlkMOdzOzHHK4m5nlkMPdzCyHHO5mZjnkcDczyyGHu5lZDjnczcxyyOFuZpZD\nDnczsxxyuJuZ5ZDD3cwshxzuZmY55HA3M8shh7uZWQ453M3McsjhbmaWQw53M7MccribmfWjs6ub\nh9dtY/fBjrEuZVAyhbukiyStlrRG0jVlln9I0tOSnpB0r6S5lS/VzGz0PbZhJ3et2MQVX3mYHfva\nx7qczAYMd0k1wI3AxcBi4G2SFpc0ewxYEhGnA3cCn610oWZmY+HpTbuZXF/D6s17uOymn7Nl98Gx\nLimTLFvu5wBrImJtRLQDtwGXFjeIiPsiYn86+RAwu7JlmpmNvraOLtZs3cur5kzn6+88mw3bD/CZ\nHz4z1mVlkiXcZwEbiqZb0nl9eRfwg3ILJF0labmk5Vu3bs1epZnZGHh2y166uoPFJzTyP05q4nUv\nb+bRX+0Y67IyyRLuKjMvyjaU/gBYAnyu3PKIuCkilkTEkubm5uxVmpmNgac37WJyfQ0nzpgMwBlz\npvOrbfvHxdh7lnBvAeYUTc8GNpU2knQh8HHgkohoq0x5ZmZjo7O7m9Wb9/DK46ZRU0i2cc+cMx2A\nx1t2jmVpmWQJ90eARZLmS6oHLgeWFjeQ9CrgX0iCfUvlyzQzG13rWvdxsKObxSdM65132uxGJFix\nIQfhHhGdwNXAPcAq4PaIWCnpBkmXpM0+BzQAd0haIWlpH92ZmY0Lq17cTV2NWDizoXdew4RaFs1s\n4PFxEO61WRpFxDJgWcm864quX1jhuszMxtTGHQeYfcxk6moO3wY+c850/uvpzUQEUrldktXB31A1\nMysREWzZ08ax0yYcseyMOdPZsb+DDdsPjEFl2TnczcxK7D7YSVtnNzOnTjxiWc9O1RVVvlPV4W5m\nVqLnW6gzy2y5v/zYqUysK7DiBYe7mdm4snlPcjR3uS33upoCp57QWPWHQzrczcxKbNl9kMn1NTRM\nKH/MyZlzpvPUxl10dnWPcmXZOdzNzEokO1OP3GrvcfJxU2nr7GbDjurdqepwNzMrkhwpc5CZU48c\nb++xoDk59n3t1r2jVdagOdzNzIps2dPGwY7ufsP9pOYpAKzdum+0yho0h7uZWZFnN+8BYGY/wzLT\nJ9dzzOQ61rZ6y93MbFx4bnMS2P1tuUMyNPO8t9zNzMaH57bsZVJd30fK9FjQNMXDMmZm48Vzm/cw\nc9qEAc8bs6C5gda9bVX7w9kOdzOzVETw3Ja9HFvmy0ulFqQ7VddV6da7w93MLNW6t51dBzpoHmC8\nHYqOmKnSnaoOdzOz1LrWZCs8S7ifOGMKNQVV7bi7w93MLLUu3Qpvahg43OtrC8w5ZpLD3cys2q1t\n3UddjZg+uS5T++RwSA/LmJlVtXVb9zH3ZVMoZPyFpQVNU1i/bR/d3THClQ2ew93MLLV+2z7mN03J\n3H5BcwMHO7rZtKv6TiDmcDczA7q6g/Xb9rNgUOFeveeYcbibmQGbdh6gvbN7cFvuTT3hXn3j7g53\nMzMOHQY5mHBvnjqBhgm1rG31lruZWVUaSrhLYkHzlN7bVhOHu5kZSbhPqa/J9AWmYtV6AjGHu5kZ\nyTHu85unDHjCsFILmhvYuPMAB9q7RqiyockU7pIukrRa0hpJ15RZ/lpJv5TUKektlS/TzGxkrWvd\ny/ymhkHfrvcEYlU2NDNguEuqAW4ELgYWA2+TtLik2QvAlcCtlS7QzGyktXV2sXHHgUGNt/dYkL4h\nVNsJxPo/G33iHGBNRKwFkHQbcCnwdE+DiFifLusegRrNzEbUhu376Q4GdYx7j/lN1Xmse5ZhmVnA\nhqLplnTeoEm6StJyScu3bt06lC7MzCquJ5iHsuU+qb6GWdMnVd2x7lnCvdzehSGdSCEiboqIJRGx\npLm5eShdmJlVXM9x6vOGEO6QjLtX27HuWcK9BZhTND0b2DQy5ZiZjb7VL+3huGkTaZyU7WyQpXoO\nh4yonhOIZQn3R4BFkuZLqgcuB5aObFlmZqPnmZf28Irjpw759guaG9jb1snWPW0VrGp4Bgz3iOgE\nrgbuAVYBt0fESkk3SLoEQNLZklqAtwL/ImnlSBZtZlYpHV3drNmyh5OPG3q494zVP19FO1WzHC1D\nRCwDlpXMu67o+iMkwzVmZuPKutZ9dHQFrxhGuC8o+j3V8056WaVKGxZ/Q9XMjmrPvLQHgJOPnTbk\nPk5onMTEugLrqmjL3eFuZke11S/tpqYgTpo5tCNlAAoFMe9l1XXEjMPdzI5qq1/aw4KmKUyorRlW\nPwtnNrA6/RRQDRzuZnZUW/XiHl5x/NCHZHqcNquRjTsPsH1fewWqGj6Hu5kdtfYc7GDjzgPD2pna\n47TZjQA8uXHXsPuqBIe7mR21nt3cszN1+OF+6qwk3J9yuJuZja3eI2UqsOU+bWId85um8ETLzmH3\nVQkOdzM7aq1+aQ8NE2qZfcykivR32qxGnmzxlruZ2Zh65sU9vPzYhkH/+lJfTp/dyKZdB6viNAQO\ndzM7KrV3dvPkxl2clo6VV0I1jbs73M3sqLRiw04OdHRx3klNFevzlBOmIVXHETMOdzM7Kj24ppWC\n4LwFlTsXzNSJdSxomsITVTDu7nA3s6PS/1/TyqmzGmmcPLRzuPfl9NnTeXLj2B8x43A3s6POvrZO\nVmzYyfkLKzck0+PUWY1s3t3G5t0HK973YDjczeyo84t12+nsDs6v4Hh7jyVzjwHgZ2taK973YDjc\nzeyo87M1rdTXFlgy75iK933arEaOb5zIsidfqnjfg+FwN7Ojzs+e38ZZJx7DxLrhnQmynEJBvOmU\n4/jpc1vZ29ZZ8f4z1zFmazYzGwPb9rax6sXdnL9w5H4x6c2nHU97Zzf3PbNlxNYxEIe7mR1VvvfY\nRgAuOHnmiK3jrLnH0NQwgR8+NXZDM5l+QzXPbn34hSPmvf3cE8egEjMbae2d3Xz1gXWcO39G77dJ\nR0JNQbzplGP57i83cqC9i0n1lR/+GYi33M3sqPH9xzby0u6DvO/1C0d8XW8+7XgOdHTxk2e3jvi6\nynG4m9lRoas7+PJPnueUE6bx2kWVPwSy1LnzZzBjSj23PLSeiBjx9ZVyuJcREbR1drFrfwf72zvp\n7h79J8bMKuuHT73E2tZ9vO+ChRU7C2R/amsK/PmFi/jZmm3ctWLTiK/viPWP+hqrzIH2Lp7fupfn\nt+5l8+42duxv57q7nqKzJNDrawtMqqthUl0NjZPqaJxUx7T07/TJdb3zGifV0Vg6PamOupry76Me\n8zcbeU+27OJj33uShTMbuOjU40ZtvW8/dy7f+eVG/vbup7ng5GamT64ftXUfdeHe0dXNig07eeC5\nVh54bisrXthJkIT38dMmsqBpCtMm1TGhtkBdTYGu7qCjq5uOrm7au4KOzm4OdHSxdW8bL2zfT0Gw\n80AH+9u7+l3vlPoapkyoZVJ98gYxMX2j2LG/nbqaQnoRdTUFWnbs752uLVlWWxD1tQVqC4fm1dUU\nqO29fuhvbaFAXW2BuoIOa1NbEDUFIclvLpZ7j72wgz+8+Rc0Tqrja1eeTU1h5Lfae9QUxKd/9zR+\n+x8f5G/vXsU/vPX0UfnUABnDXdJFwJeAGuCrEfGZkuUTgG8AZwHbgMsiYn1lSx2aiGBd6z4eXNPK\nT59t5aG129jb1klBcMac6Vxw8kwWzWxgzozJw3rSO7u7OdjRzYH2Lg60d3Kgoyu5tHexv6OLg+1d\ntHV2p28Uwb62Tnbsb0+mOyN980iWP7R22xGfHEZCbUFIUJAoKAn8guCf719DXU2BmoKoLYjanjeK\nmqRN8bLkb8l0TR/zC6KmkLzJJOtMTrsqCQESSOKc+TMo9NaVzOu5XtDhNR9ans4rHN6+Rjq0vDBw\nf8l1gEPze2orpP+UKrqdSPqh6HrP/SkkDQbuh0P3YbT+8fNuz8EOHnthJ3c82sI9K1/iuGkT+dZV\nr2bW9Mr84tJgvPL4afzx6xZw433Ps31fG3//ltOZOXXiiK93wHCXVAPcCLwRaAEekbQ0Ip4uavYu\nYEdELJR0OfD3wGUjUfCOfe3sOdhJ46Q6pk6sRUp2lOxr72Ln/nZa97azYft+1rXuY+WmXazYsJPW\nve0AzJkxiUvOPIHXLmrivAVNNE6uK7vlOhS1hQINEwo0TKgFJgy7v4igK4Lu7uT+dUUkf3suEXSX\nTPe5vGRZdwRd3dCdtumOnjaH5nX1LoOOzqAtOg/rM6B3eVdRP91R3G96ved+ZNyp9B+Pj/74ZLXp\neSOAQ28WyZtvz5tK0ZsHyRtX8ZvEoTen5E2n7zenw/vpWdeuAx2HtQeYMaW+9w1NRe0Bep/Z9DmO\noslIp3qe/t6/vW0OvS56ru7Y337Y4xHAtIm1Jes6sp8AtqcZAdA4qY63nT2H971+IcdOG/lA7cuH\n33gyTQ0T+MwPnuFNX/gpn7/sTF4/gsfZQ7Yt93OANRGxFkDSbcClQHG4Xwpcn16/E/gnSYoR2EV8\nx6Mb+LtlzwzYToL5TVN47cubOWvuMbxmYRNzXzal0uWMGEnUSrnb5d0T/EH6jx+HX+/myHkDXifS\neYeud0c6r8/+Dl/eXXSd3n4BDrWjqP/iNofNK9dH0fqPaFNUH701lKy3bJtkSXeZeWX7KNsmivo/\nvNbmqROOmNfVHYf6LQlsOPQm0DuN+ll2aEHPG0rPAilZf9m2h3dTtt/jGycyfVI9zVMnsGhmA7U1\nBe5dNXbfFIXkzfed58/n1xc18eE7nmBK/ciPiGdZwyxgQ9F0C3BuX20iolPSLuBlwGGnRZN0FXBV\nOrlX0uqhFJ3Vemi6r6SGcaIJ1z2aXPfoGpd1X1HBupcO7+ZzszTKEu7lBgFLt8iztCEibgJuyrDO\nipC0PCKWjNb6KsV1jy7XPbpc9+jI8qG/BZhTND0bKB0U7W0jqRZoBLZXokAzMxu8LOH+CLBI0nxJ\n9cDlHPmpYinwjvT6W4Afj8R4u5mZZTPgsEw6hn41cA/JoZA3R8RKSTcAyyNiKfCvwC2S1pBssV8+\nkkUPwqgNAVWY6x5drnt0ue5RIG9gm5nlT84OtDMzM3C4m5nlUi7DXdJFklZLWiPpmrGupy+Sbpa0\nRdJTRfNmSPovSc+lfyv/C77DJGmOpPskrZK0UtIH0/lVXbukiZJ+IenxtO6/SefPl/RwWve30wMH\nqo6kGkmPSbo7nR4vda+X9KSkFZKWp/Oq+rUCIGm6pDslPZO+1s8bD3X3yF24F50u4WJgMfA2SYvH\ntqo+fR24qGTeNcC9EbEIuDedrjadwIcj4pXAq4H3p49xtdfeBrwhIs4AzgQukvRqktNlfCGtewfJ\n6TSq0QeBVUXT46VugNdHxJlFx4lX+2sFkvNp/TAiXgGcQfLYj4e6E8nXsPNzAc4D7imavha4dqzr\n6qfeecBTRdOrgePT68cDq8e6xgz34S6Scw+Nm9qBycAvSb5t3QrUlnv9VMuF5Psl9wJvAO4m+eJg\n1ded1rYeaCqZV9WvFWAasI70oJPxUnfxJXdb7pQ/XcKsMaplKI6NiBcB0r8je3ahYZI0D3gV8DDj\noPZ0aGMFsAX4L+B5YGdEdKZNqvX18kXgL0lPJ0Nyeo/xUDck31b/T0mPpqcggep/rSwAtgJfS4fC\nvippCtVfd688hnumUyHY8ElqAL4D/FlE7B7rerKIiK6IOJNkS/gc4JXlmo1uVf2T9FvAloh4tHh2\nmaZVVXeR8yPi10iGSt8v6bVjXVAGtcCvAf83Il4F7KOah2DKyGO4ZzldQjXbLOl4gPTv2J7Org+S\n6kiC/d8j4rvp7HFRO0BE7ATuJ9lnMD09bQZU5+vlfOASSeuB20iGZr5I9dcNQERsSv9uAb5H8qZa\n7a+VFqAlIh5Op+8kCftqr7uqxvcUAAAC/ElEQVRXHsM9y+kSqlnxqRzeQTKeXVWUnMT7X4FVEfH5\nokVVXbukZknT0+uTgAtJdpLdR3LaDKjCuiPi2oiYHRHzSF7PP46IK6jyugEkTZE0tec68D+Bp6jy\n10pEvARskHRyOus3SE5zXtV1H2asB/1HaGfIm4FnScZTPz7W9fRT57eAF4EOki2Fd5GMpd4LPJf+\nnTHWdZap+zUkQwBPACvSy5urvXbgdOCxtO6ngOvS+QuAXwBrgDuACWNdaz/34QLg7vFSd1rj4+ll\nZc//Y7W/VtIazwSWp6+X7wPHjIe6ey4+/YCZWQ7lcVjGzOyo53A3M8shh7uZWQ453M3McsjhbmaW\nQw53M7MccrjbuCZpb/r3BEl39tNuuqT3FU33236ItbxX0h+WmT+v+LTOZqPBx7nbuCZpb0Q0ZGg3\nj+TLP6eOeFFVtG47ennL3XKheOtY0inpj3KskPSEpEXAZ4CT0nmfK2l/paTvSvph+iMMny3q912S\nnpV0v6SvSPqnfmq4XtJH0utnpT8K8nPg/SN6583KqB24idm4817gSxHx7+n5hWpIzuh3aiRnhOzZ\nmi52Jsmpi9uA1ZL+EegC/orkhFF7gB+TfI0+i68BfxoRP5H0ueHdHbPB85a75dHPgY9J+igwNyIO\nZLjNvRGxKyIOkpwgai7J2Qt/EhHbI6KD5PwtA5LUCEyPiJ+ks24Z/F0wGx6Hu+VORNwKXAIcAO6R\n9IYMN2srut5F8qm23DnTsxDVe251O0o43C13JC0A1kbE/yE5RevpJMMqUwfZ1S+A10k6Jj1v+u9l\nuVEk54rfJek16awrBrles2FzuFseXQY8lf6c3iuAb0TENuBnkp7KOgYeERuBvyP5CcEfkQzX7MpY\nwzuBG9MdqlmGhcwqyodCmvVDUkNE7E233L8H3BwR3xvruswG4i13s/5dn34CeApYR/KjDWZVz1vu\nZoMk6ePAW0tm3xERnxqLeszKcbibmeWQh2XMzHLI4W5mlkMOdzOzHHK4m5nl0H8D3CTXlLHPRoQA\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('Distribution of documents per listing')\n",
    "sns.distplot(data['listing_id'].value_counts());"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAELCAYAAADURYGZAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.countplot(data[data['listing_id'] <= 20]['listing_id'], palette='Set2');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sample observations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>listing_id</th>\n",
       "      <th>doc_rank</th>\n",
       "      <th>clicked</th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.142328</td>\n",
       "      <td>0.007693</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>99999.0</td>\n",
       "      <td>0.048142</td>\n",
       "      <td>-0.131414</td>\n",
       "      <td>-0.125540</td>\n",
       "      <td>19.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.023810</td>\n",
       "      <td>0.758929</td>\n",
       "      <td>0.091036</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.060197</td>\n",
       "      <td>-0.131414</td>\n",
       "      <td>-0.125540</td>\n",
       "      <td>17.90</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.133929</td>\n",
       "      <td>0.001270</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>5.0</td>\n",
       "      <td>-0.119726</td>\n",
       "      <td>-0.276582</td>\n",
       "      <td>-0.213422</td>\n",
       "      <td>9.75</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.062500</td>\n",
       "      <td>2.175595</td>\n",
       "      <td>0.276685</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>600.0</td>\n",
       "      <td>0.191341</td>\n",
       "      <td>-0.131414</td>\n",
       "      <td>-0.125540</td>\n",
       "      <td>19.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.026786</td>\n",
       "      <td>1.547619</td>\n",
       "      <td>0.044791</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>99999.0</td>\n",
       "      <td>-0.137913</td>\n",
       "      <td>-0.131414</td>\n",
       "      <td>-0.125540</td>\n",
       "      <td>7.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 53 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   listing_id  doc_rank  clicked    n0    n1    n2        n3        n4  \\\n",
       "0           0         0        0  87.0  86.0  87.0  0.000000  0.142328   \n",
       "1           0         1        0  87.0  86.0  87.0  0.023810  0.758929   \n",
       "2           0         2        0  87.0  86.0  87.0  0.000000  0.133929   \n",
       "3           0         3        0  87.0  86.0  87.0  0.062500  2.175595   \n",
       "4           0         4        0  87.0  86.0  87.0  0.026786  1.547619   \n",
       "\n",
       "         n5   n6  ...      n40       n41       n42       n43    n44  c0  c1  \\\n",
       "0  0.007693  0.0  ...  99999.0  0.048142 -0.131414 -0.125540  19.99   0   0   \n",
       "1  0.091036  0.0  ...   1000.0  0.060197 -0.131414 -0.125540  17.90   0   0   \n",
       "2  0.001270  0.0  ...      5.0 -0.119726 -0.276582 -0.213422   9.75   0   0   \n",
       "3  0.276685  0.0  ...    600.0  0.191341 -0.131414 -0.125540  19.99   0   0   \n",
       "4  0.044791  0.0  ...  99999.0 -0.137913 -0.131414 -0.125540   7.99   0   0   \n",
       "\n",
       "   c2  c3  c4  \n",
       "0   4   0   0  \n",
       "1   4   0   0  \n",
       "2   3   0   0  \n",
       "3   4   0   0  \n",
       "4   4   0   0  \n",
       "\n",
       "[5 rows x 53 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>listing_id</th>\n",
       "      <th>doc_rank</th>\n",
       "      <th>clicked</th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4885209</th>\n",
       "      <td>99999</td>\n",
       "      <td>55</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002976</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>50.0</td>\n",
       "      <td>0.520656</td>\n",
       "      <td>0.143999</td>\n",
       "      <td>0.323712</td>\n",
       "      <td>25.90</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4885210</th>\n",
       "      <td>99999</td>\n",
       "      <td>56</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6.0</td>\n",
       "      <td>0.716215</td>\n",
       "      <td>0.143999</td>\n",
       "      <td>0.323712</td>\n",
       "      <td>109.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4885211</th>\n",
       "      <td>99999</td>\n",
       "      <td>57</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000148</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>10.0</td>\n",
       "      <td>0.420662</td>\n",
       "      <td>0.143999</td>\n",
       "      <td>0.323712</td>\n",
       "      <td>39.00</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4885212</th>\n",
       "      <td>99999</td>\n",
       "      <td>58</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>7.0</td>\n",
       "      <td>0.772205</td>\n",
       "      <td>0.143999</td>\n",
       "      <td>0.323712</td>\n",
       "      <td>124.49</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4885213</th>\n",
       "      <td>99999</td>\n",
       "      <td>59</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000179</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.896056</td>\n",
       "      <td>0.143999</td>\n",
       "      <td>0.323712</td>\n",
       "      <td>109.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 53 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         listing_id  doc_rank  clicked   n0   n1   n2   n3        n4  \\\n",
       "4885209       99999        55        0  0.0  0.0  0.0  0.0  0.002976   \n",
       "4885210       99999        56        0  0.0  0.0  0.0  0.0  0.000000   \n",
       "4885211       99999        57        0  0.0  0.0  0.0  0.0  0.000000   \n",
       "4885212       99999        58        0  0.0  0.0  0.0  0.0  0.000000   \n",
       "4885213       99999        59        0  0.0  0.0  0.0  0.0  0.000000   \n",
       "\n",
       "               n5   n6  ...   n40       n41       n42       n43     n44  c0  \\\n",
       "4885209  0.000000  0.0  ...  50.0  0.520656  0.143999  0.323712   25.90   0   \n",
       "4885210  0.000000  0.0  ...   6.0  0.716215  0.143999  0.323712  109.99   0   \n",
       "4885211  0.000148  0.0  ...  10.0  0.420662  0.143999  0.323712   39.00   0   \n",
       "4885212  0.000000  0.0  ...   7.0  0.772205  0.143999  0.323712  124.49   0   \n",
       "4885213  0.000179  0.0  ...   3.0  0.896056  0.143999  0.323712  109.99   0   \n",
       "\n",
       "         c1  c2  c3  c4  \n",
       "4885209   0   4   1   3  \n",
       "4885210   0   4   1   3  \n",
       "4885211   0   4   1   3  \n",
       "4885212   0   4   1   3  \n",
       "4885213   0   4   1   3  \n",
       "\n",
       "[5 rows x 53 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.tail()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "There are slightly more documents clicked (~108k) than listings (100k)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    4776361\n",
       "1     108853\n",
       "Name: clicked, dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['clicked'].value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br><br><br><br><br><br>\n",
    "## Data modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Defining train/test subsets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Splitting original dataset for train/test subsets \n",
    "\n",
    "threshold = 10000\n",
    "threshold_upper = 12000\n",
    "\n",
    "data_train = data[data['listing_id'] <= threshold]\n",
    "data_test = data[(data['listing_id'] > threshold) & (data['listing_id'] <= threshold_upper)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br>\n",
    "### Defining 'y' label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Defining relevance label\n",
    "y_train = data_train.pop('clicked')\n",
    "y_test = data_test.pop('clicked')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((494834, 52), (97408, 52), (494834,), (97408,))"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_train.shape, data_test.shape, y_train.shape, y_test.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "<br><br>\n",
    "### Scaling data numerics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_train_numerics = data_train.loc[:,'n0':]\n",
    "data_test_numerics = data_test.loc[:,'n0':]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>n7</th>\n",
       "      <th>n8</th>\n",
       "      <th>n9</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.00000</td>\n",
       "      <td>0.142328</td>\n",
       "      <td>0.007693</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>99999.0</td>\n",
       "      <td>0.048142</td>\n",
       "      <td>-0.131414</td>\n",
       "      <td>-0.12554</td>\n",
       "      <td>19.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>87.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>87.0</td>\n",
       "      <td>0.02381</td>\n",
       "      <td>0.758929</td>\n",
       "      <td>0.091036</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>1000.0</td>\n",
       "      <td>0.060197</td>\n",
       "      <td>-0.131414</td>\n",
       "      <td>-0.12554</td>\n",
       "      <td>17.90</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 50 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     n0    n1    n2       n3        n4        n5   n6   n7   n8    n9  ...  \\\n",
       "0  87.0  86.0  87.0  0.00000  0.142328  0.007693  0.0  0.0  0.0  0.01  ...   \n",
       "1  87.0  86.0  87.0  0.02381  0.758929  0.091036  0.0  0.0  0.0  0.01  ...   \n",
       "\n",
       "       n40       n41       n42      n43    n44  c0  c1  c2  c3  c4  \n",
       "0  99999.0  0.048142 -0.131414 -0.12554  19.99   0   0   4   0   0  \n",
       "1   1000.0  0.060197 -0.131414 -0.12554  17.90   0   0   4   0   0  \n",
       "\n",
       "[2 rows x 50 columns]"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_train_numerics.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>n7</th>\n",
       "      <th>n8</th>\n",
       "      <th>n9</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>494834</th>\n",
       "      <td>80.0</td>\n",
       "      <td>98.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>0.032738</td>\n",
       "      <td>0.791667</td>\n",
       "      <td>0.136692</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.01</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.03</td>\n",
       "      <td>...</td>\n",
       "      <td>1200.0</td>\n",
       "      <td>0.867136</td>\n",
       "      <td>0.426012</td>\n",
       "      <td>0.198007</td>\n",
       "      <td>19.99</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>494835</th>\n",
       "      <td>80.0</td>\n",
       "      <td>98.0</td>\n",
       "      <td>86.0</td>\n",
       "      <td>0.023810</td>\n",
       "      <td>0.684524</td>\n",
       "      <td>0.113259</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.01</td>\n",
       "      <td>...</td>\n",
       "      <td>10000.0</td>\n",
       "      <td>0.856878</td>\n",
       "      <td>0.426012</td>\n",
       "      <td>0.198007</td>\n",
       "      <td>29.50</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 50 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "          n0    n1    n2        n3        n4        n5   n6    n7   n8    n9  \\\n",
       "494834  80.0  98.0  86.0  0.032738  0.791667  0.136692  0.0  0.01  0.0  0.03   \n",
       "494835  80.0  98.0  86.0  0.023810  0.684524  0.113259  0.0  0.00  0.0  0.01   \n",
       "\n",
       "        ...      n40       n41       n42       n43    n44  c0  c1  c2  c3  c4  \n",
       "494834  ...   1200.0  0.867136  0.426012  0.198007  19.99   0   0   4   1   1  \n",
       "494835  ...  10000.0  0.856878  0.426012  0.198007  29.50   0   0   4   1   1  \n",
       "\n",
       "[2 rows x 50 columns]"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_test_numerics.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Defining scaler\n",
    "numeric_scaler = StandardScaler()\n",
    "\n",
    "# Fit & transform train set\n",
    "train_numerics_scaled = numeric_scaler.fit_transform(data_train_numerics)\n",
    "train_numerics_scaled = pd.DataFrame(train_numerics_scaled, columns=data_train_numerics.columns, index=data_train_numerics.index)\n",
    "\n",
    "# Transform test set\n",
    "test_numerics_scaled = numeric_scaler.transform(data_test_numerics)\n",
    "test_numerics_scaled = pd.DataFrame(test_numerics_scaled, columns=data_test_numerics.columns, index=data_test_numerics.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>n7</th>\n",
       "      <th>n8</th>\n",
       "      <th>n9</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.211761</td>\n",
       "      <td>1.108089</td>\n",
       "      <td>1.1926</td>\n",
       "      <td>-0.343223</td>\n",
       "      <td>-0.184692</td>\n",
       "      <td>-0.229624</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.322053</td>\n",
       "      <td>-0.346921</td>\n",
       "      <td>-0.514283</td>\n",
       "      <td>...</td>\n",
       "      <td>5.007544</td>\n",
       "      <td>-3.156401</td>\n",
       "      <td>-2.511417</td>\n",
       "      <td>-1.626134</td>\n",
       "      <td>-0.043119</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>-1.137446</td>\n",
       "      <td>-0.957742</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1.211761</td>\n",
       "      <td>1.108089</td>\n",
       "      <td>1.1926</td>\n",
       "      <td>1.494542</td>\n",
       "      <td>0.995122</td>\n",
       "      <td>1.524552</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.322053</td>\n",
       "      <td>-0.346921</td>\n",
       "      <td>-0.514283</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.181954</td>\n",
       "      <td>-3.100811</td>\n",
       "      <td>-2.511417</td>\n",
       "      <td>-1.626134</td>\n",
       "      <td>-0.043738</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>-1.137446</td>\n",
       "      <td>-0.957742</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 50 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         n0        n1      n2        n3        n4        n5        n6  \\\n",
       "0  1.211761  1.108089  1.1926 -0.343223 -0.184692 -0.229624 -0.276655   \n",
       "1  1.211761  1.108089  1.1926  1.494542  0.995122  1.524552 -0.276655   \n",
       "\n",
       "         n7        n8        n9  ...       n40       n41       n42       n43  \\\n",
       "0 -0.322053 -0.346921 -0.514283  ...  5.007544 -3.156401 -2.511417 -1.626134   \n",
       "1 -0.322053 -0.346921 -0.514283  ... -0.181954 -3.100811 -2.511417 -1.626134   \n",
       "\n",
       "        n44        c0        c1        c2        c3        c4  \n",
       "0 -0.043119 -0.336061 -0.295113 -0.029593 -1.137446 -0.957742  \n",
       "1 -0.043738 -0.336061 -0.295113 -0.029593 -1.137446 -0.957742  \n",
       "\n",
       "[2 rows x 50 columns]"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_numerics_scaled.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>n7</th>\n",
       "      <th>n8</th>\n",
       "      <th>n9</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>494834</th>\n",
       "      <td>1.034377</td>\n",
       "      <td>1.401779</td>\n",
       "      <td>1.167539</td>\n",
       "      <td>2.183704</td>\n",
       "      <td>1.057764</td>\n",
       "      <td>2.485525</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.178179</td>\n",
       "      <td>-0.346921</td>\n",
       "      <td>-0.414861</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.171471</td>\n",
       "      <td>0.620480</td>\n",
       "      <td>-0.457059</td>\n",
       "      <td>-0.712891</td>\n",
       "      <td>-0.043119</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>0.566174</td>\n",
       "      <td>0.294427</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>494835</th>\n",
       "      <td>1.034377</td>\n",
       "      <td>1.401779</td>\n",
       "      <td>1.167539</td>\n",
       "      <td>1.494542</td>\n",
       "      <td>0.852755</td>\n",
       "      <td>1.992307</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.322053</td>\n",
       "      <td>-0.346921</td>\n",
       "      <td>-0.514283</td>\n",
       "      <td>...</td>\n",
       "      <td>0.289823</td>\n",
       "      <td>0.573175</td>\n",
       "      <td>-0.457059</td>\n",
       "      <td>-0.712891</td>\n",
       "      <td>-0.040301</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>0.566174</td>\n",
       "      <td>0.294427</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 50 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              n0        n1        n2        n3        n4        n5        n6  \\\n",
       "494834  1.034377  1.401779  1.167539  2.183704  1.057764  2.485525 -0.276655   \n",
       "494835  1.034377  1.401779  1.167539  1.494542  0.852755  1.992307 -0.276655   \n",
       "\n",
       "              n7        n8        n9  ...       n40       n41       n42  \\\n",
       "494834 -0.178179 -0.346921 -0.414861  ... -0.171471  0.620480 -0.457059   \n",
       "494835 -0.322053 -0.346921 -0.514283  ...  0.289823  0.573175 -0.457059   \n",
       "\n",
       "             n43       n44        c0        c1        c2        c3        c4  \n",
       "494834 -0.712891 -0.043119 -0.336061 -0.295113 -0.029593  0.566174  0.294427  \n",
       "494835 -0.712891 -0.040301 -0.336061 -0.295113 -0.029593  0.566174  0.294427  \n",
       "\n",
       "[2 rows x 50 columns]"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_numerics_scaled.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br>\n",
    "### Defining final train/test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Merging listing and docs columns with numerics dataframes\n",
    "train_adult = pd.concat([data_train.loc[:,:'doc_rank'], train_numerics_scaled], axis=1)\n",
    "test_adult = pd.concat([data_test.loc[:,:'doc_rank'], test_numerics_scaled], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((494834, 52), (97408, 52))"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Shape of final dataframes\n",
    "train_adult.shape, test_adult.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>listing_id</th>\n",
       "      <th>doc_rank</th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>n7</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1.211761</td>\n",
       "      <td>1.108089</td>\n",
       "      <td>1.1926</td>\n",
       "      <td>-0.343223</td>\n",
       "      <td>-0.184692</td>\n",
       "      <td>-0.229624</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.322053</td>\n",
       "      <td>...</td>\n",
       "      <td>5.007544</td>\n",
       "      <td>-3.156401</td>\n",
       "      <td>-2.511417</td>\n",
       "      <td>-1.626134</td>\n",
       "      <td>-0.043119</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>-1.137446</td>\n",
       "      <td>-0.957742</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1.211761</td>\n",
       "      <td>1.108089</td>\n",
       "      <td>1.1926</td>\n",
       "      <td>1.494542</td>\n",
       "      <td>0.995122</td>\n",
       "      <td>1.524552</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.322053</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.181954</td>\n",
       "      <td>-3.100811</td>\n",
       "      <td>-2.511417</td>\n",
       "      <td>-1.626134</td>\n",
       "      <td>-0.043738</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>-1.137446</td>\n",
       "      <td>-0.957742</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 52 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   listing_id  doc_rank        n0        n1      n2        n3        n4  \\\n",
       "0           0         0  1.211761  1.108089  1.1926 -0.343223 -0.184692   \n",
       "1           0         1  1.211761  1.108089  1.1926  1.494542  0.995122   \n",
       "\n",
       "         n5        n6        n7  ...       n40       n41       n42       n43  \\\n",
       "0 -0.229624 -0.276655 -0.322053  ...  5.007544 -3.156401 -2.511417 -1.626134   \n",
       "1  1.524552 -0.276655 -0.322053  ... -0.181954 -3.100811 -2.511417 -1.626134   \n",
       "\n",
       "        n44        c0        c1        c2        c3        c4  \n",
       "0 -0.043119 -0.336061 -0.295113 -0.029593 -1.137446 -0.957742  \n",
       "1 -0.043738 -0.336061 -0.295113 -0.029593 -1.137446 -0.957742  \n",
       "\n",
       "[2 rows x 52 columns]"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_adult.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>listing_id</th>\n",
       "      <th>doc_rank</th>\n",
       "      <th>n0</th>\n",
       "      <th>n1</th>\n",
       "      <th>n2</th>\n",
       "      <th>n3</th>\n",
       "      <th>n4</th>\n",
       "      <th>n5</th>\n",
       "      <th>n6</th>\n",
       "      <th>n7</th>\n",
       "      <th>...</th>\n",
       "      <th>n40</th>\n",
       "      <th>n41</th>\n",
       "      <th>n42</th>\n",
       "      <th>n43</th>\n",
       "      <th>n44</th>\n",
       "      <th>c0</th>\n",
       "      <th>c1</th>\n",
       "      <th>c2</th>\n",
       "      <th>c3</th>\n",
       "      <th>c4</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>494834</th>\n",
       "      <td>10001</td>\n",
       "      <td>0</td>\n",
       "      <td>1.034377</td>\n",
       "      <td>1.401779</td>\n",
       "      <td>1.167539</td>\n",
       "      <td>2.183704</td>\n",
       "      <td>1.057764</td>\n",
       "      <td>2.485525</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.178179</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.171471</td>\n",
       "      <td>0.620480</td>\n",
       "      <td>-0.457059</td>\n",
       "      <td>-0.712891</td>\n",
       "      <td>-0.043119</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>0.566174</td>\n",
       "      <td>0.294427</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>494835</th>\n",
       "      <td>10001</td>\n",
       "      <td>1</td>\n",
       "      <td>1.034377</td>\n",
       "      <td>1.401779</td>\n",
       "      <td>1.167539</td>\n",
       "      <td>1.494542</td>\n",
       "      <td>0.852755</td>\n",
       "      <td>1.992307</td>\n",
       "      <td>-0.276655</td>\n",
       "      <td>-0.322053</td>\n",
       "      <td>...</td>\n",
       "      <td>0.289823</td>\n",
       "      <td>0.573175</td>\n",
       "      <td>-0.457059</td>\n",
       "      <td>-0.712891</td>\n",
       "      <td>-0.040301</td>\n",
       "      <td>-0.336061</td>\n",
       "      <td>-0.295113</td>\n",
       "      <td>-0.029593</td>\n",
       "      <td>0.566174</td>\n",
       "      <td>0.294427</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 52 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        listing_id  doc_rank        n0        n1        n2        n3  \\\n",
       "494834       10001         0  1.034377  1.401779  1.167539  2.183704   \n",
       "494835       10001         1  1.034377  1.401779  1.167539  1.494542   \n",
       "\n",
       "              n4        n5        n6        n7  ...       n40       n41  \\\n",
       "494834  1.057764  2.485525 -0.276655 -0.178179  ... -0.171471  0.620480   \n",
       "494835  0.852755  1.992307 -0.276655 -0.322053  ...  0.289823  0.573175   \n",
       "\n",
       "             n42       n43       n44        c0        c1        c2        c3  \\\n",
       "494834 -0.457059 -0.712891 -0.043119 -0.336061 -0.295113 -0.029593  0.566174   \n",
       "494835 -0.457059 -0.712891 -0.040301 -0.336061 -0.295113 -0.029593  0.566174   \n",
       "\n",
       "              c4  \n",
       "494834  0.294427  \n",
       "494835  0.294427  \n",
       "\n",
       "[2 rows x 52 columns]"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_adult.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br>\n",
    "\n",
    "### Building ranker\n",
    "\n",
    "**XGBRanker** model is build as a PoC. With ``rank:ndcg`` as objective it uses *LambdaMART* to perform list-wise ranking where **NDCG is maximized**.<br>\n",
    "\n",
    "<u>Group information is required for ranking tasks.</u><br>\n",
    "Before fitting the model, the data need to be sorted by group. When fitting the model, you need to provide an additional array that contains the size of each group."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Defining grouping array\n",
    "groups = train_adult['listing_id'].value_counts().sort_index()\n",
    "\n",
    "# Defining evaluation sets and groups\n",
    "eval_set = [(train_adult.drop('doc_rank', axis=1), y_train),\n",
    "            (test_adult.drop('doc_rank', axis=1), y_test)]\n",
    "eval_group = [train_adult['listing_id'].value_counts().sort_index(),\n",
    "              test_adult['listing_id'].value_counts().sort_index()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Building ranker. Using LambdaMART with NDCG as objective\n",
    "param_dist = {'objective':'rank:ndcg', \n",
    "              'n_estimators':150,\n",
    "              'random_state':42,\n",
    "              'n_jobs':-1\n",
    "             }\n",
    "\n",
    "ranker = XGBRanker(**param_dist)\n",
    "\n",
    "ranker.fit(X=train_adult.drop('doc_rank', axis=1), \n",
    "           y=y_train,\n",
    "           group=groups,\n",
    "           eval_set=eval_set,\n",
    "           eval_group=eval_group,\n",
    "           eval_metric='ndcg',\n",
    "           verbose=False)\n",
    "\n",
    "evals_result = ranker.evals_result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Train NDCG score: 0.41954768000000003 \n",
      " Test NDCG score:  0.41412291333333334\n"
     ]
    }
   ],
   "source": [
    "print(' Train NDCG score: {} \\n Test NDCG score:  {}'.format(np.mean(evals_result['eval_0']['ndcg']),\n",
    "                np.mean(evals_result['eval_1']['ndcg'])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<br><br>\n",
    "### Predicting and re-ranking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds = ranker.predict(test_adult.drop('doc_rank', axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAc8AAAEWCAYAAAAASRzMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_importance(ranker, max_num_features=10, importance_type='gain');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>listing_id</th>\n",
       "      <th>doc_rank</th>\n",
       "      <th>predictions</th>\n",
       "      <th>clicked</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>495079</th>\n",
       "      <td>10005</td>\n",
       "      <td>5</td>\n",
       "      <td>1.762337</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495088</th>\n",
       "      <td>10005</td>\n",
       "      <td>14</td>\n",
       "      <td>1.127221</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495090</th>\n",
       "      <td>10005</td>\n",
       "      <td>16</td>\n",
       "      <td>1.090742</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495082</th>\n",
       "      <td>10005</td>\n",
       "      <td>8</td>\n",
       "      <td>0.937371</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495083</th>\n",
       "      <td>10005</td>\n",
       "      <td>9</td>\n",
       "      <td>0.904550</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        listing_id  doc_rank  predictions  clicked\n",
       "495079       10005         5     1.762337        0\n",
       "495088       10005        14     1.127221        1\n",
       "495090       10005        16     1.090742        0\n",
       "495082       10005         8     0.937371        0\n",
       "495083       10005         9     0.904550        0"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "listing_df = test_adult.loc[:,:'doc_rank']\n",
    "preds_df = pd.DataFrame(preds, index=test_adult.index, columns=['predictions'])\n",
    "clicked_df = pd.DataFrame(y_test, index=test_adult.index, columns=['clicked'])\n",
    "\n",
    "results = pd.concat([\n",
    "                        listing_df, \n",
    "                        preds_df,\n",
    "                        clicked_df\n",
    "                    ],\n",
    "                        axis=1).sort_values(by=['listing_id','predictions'], ascending=(True,False))\n",
    "\n",
    "# Sample re-ranked listing\n",
    "results[results['listing_id'] == 10005].head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>listing_id</th>\n",
       "      <th>doc_rank</th>\n",
       "      <th>clicked</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>495074</th>\n",
       "      <td>10005</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495075</th>\n",
       "      <td>10005</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495076</th>\n",
       "      <td>10005</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495077</th>\n",
       "      <td>10005</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495078</th>\n",
       "      <td>10005</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495079</th>\n",
       "      <td>10005</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495080</th>\n",
       "      <td>10005</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495081</th>\n",
       "      <td>10005</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495082</th>\n",
       "      <td>10005</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495083</th>\n",
       "      <td>10005</td>\n",
       "      <td>9</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495084</th>\n",
       "      <td>10005</td>\n",
       "      <td>10</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495085</th>\n",
       "      <td>10005</td>\n",
       "      <td>11</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495086</th>\n",
       "      <td>10005</td>\n",
       "      <td>12</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495087</th>\n",
       "      <td>10005</td>\n",
       "      <td>13</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>495088</th>\n",
       "      <td>10005</td>\n",
       "      <td>14</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        listing_id  doc_rank  clicked\n",
       "495074       10005         0        0\n",
       "495075       10005         1        0\n",
       "495076       10005         2        0\n",
       "495077       10005         3        0\n",
       "495078       10005         4        0\n",
       "495079       10005         5        0\n",
       "495080       10005         6        0\n",
       "495081       10005         7        0\n",
       "495082       10005         8        0\n",
       "495083       10005         9        0\n",
       "495084       10005        10        0\n",
       "495085       10005        11        0\n",
       "495086       10005        12        0\n",
       "495087       10005        13        0\n",
       "495088       10005        14        1"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Original ranking - clicked on 14th place\n",
    "data[data['listing_id'] == 10005].loc[:,:'clicked'].head(15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD8CAYAAAB+UHOxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting predictions from sample listing 10005\n",
    "plt.plot(range(len(results[results['listing_id'] == 10005]['predictions'])), \n",
    "         results[results['listing_id'] == 10005]['predictions']);"
   ]
  }
 ],
 "metadata": {
  "hide_input": false,
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  },
  "notify_time": "10",
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": true,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "273.594px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}