This repository focuses training of a neural network for regression prediction using "Keras". Please check this medium post for all of the theoretical and practical details! Please contact if you need professional projects are based non-linear regression with the super high accuracy.
In this study,"the yacht hydrodynamics data set" was used as a case study and it was reached 0.99 R-square value which is awesome!
To implement a neural network for regression, it must to be defined the architecture itself. It is used a simple Multilayer Perceptron (MLP) as shown at the figure below to define the architecture.
And, here is the implementation using Keras:
The approaches and codes that shared in this tutorial can be adopted for any other regression tasks such as "computer hardware", "energy efficiency" and more!
Here are the results from left to right: plot of training history, plot of actual vs prediction for training set, plot of actual vs prediction for validation set.
Here is the main python notebook that explains all of the stuff step by step!
If you use this code for your publications, please cite it as:
@ONLINE{hse,
author = "Ahmet Özlü",
title = "Non-linear regression using Keras",
year = "2020",
url = "https://github.com/ahmetozlu/nonlinear_regression_keras"
}
Ahmet Özlü
This system is available under the MIT license. See the LICENSE file for more info.