NeuroCharter: A Neural Networks Software to Visually Discover the Effects and Contributions between Interrelated Features

Abstract

NeuroCharter is an open-source software that helps in prediction problems in scientific research through artificial neural networks. The program is designed mainly for researchers who focus on details of the neural-network’s parameters, in addition to easy reuse of the trained network. The program outputs almost all the necessary graphs regarding the network and features contributions and relative outputs for both numeric and categorical features. The program was implemented in Python 2.7.11 and is open sourced for reuse and future development. The program consists of four main classes, one for the neural networks calculation, one for data manipulation, one for plotting the neural network, and the main class that manages and links the other classes. The source code and some experimental data are freely available at the GitHub code repository http://j.mp/NeuroCharter. Funding Statement: The project was financially supported by King Saud University, Vice Deanship of Research Chairs.

Keywords

• Artificial Neural Networks
• Python
• Backpropagation
• predictions