Adult Income Prediction based on Neural Network

Shengli Zhou (12212232)

Installation

The required packages are included in requirements.txt, you can build the environment for running the code by executing the following command in the project folder:

pip install -r requirements.txt

TL;DR

python train.py --dataset train
python inference.py --dataset train

Then you can find the answers in data/testlabel.txt.

Dataset

This project is based on the Adult Census Income dataset, which can be downloaded from kaggle.

For simplicity, we've placed the downloaded data in data/full folder. Another version of dataset (the same data but different way of splitting data) provided by the project in CS311 is placed in data/train.

Quick Start

• For preprocessing the dataset:

cd data
python preprocess.py --dataset [train | full] --sep
cd ..

• For training the model:

python train.py --dataset [train | full]

• For evaluating the model (note that only full dataset is available here as we don't have the answers to the train dataset)

python evaluate.py --dataset full

• For making inference:

python inference.py --dataset [train | full]

Then, you can find the predicted labels in data/testlabel.txt (or data/testlabel_full if you use the full dataset), each line in the text file represents an answer predicted according to the given information.

Checkpoints

The official checkpoints (weights) can be found in the checkpoints folder.