Github-Recommender-System

Introduction

In this work, we used implicit ratings and an auto-encoder with a modified cost function to make a GitHub Recommender System.

Here we collected the data and constructed the confidence and prediction matrices based on implicit rating schemes. This data was then used to train an auto-encoder with a modified cost function and test the trained model using recall metric.

Refer the report for a detailed description.

Contents

• Report.pdf - a detailed description of the methodology and results of the project.
• scripts/autoencoder.py - the autoencoder code.
• scripts/data_curation - contains all the scripts used for dataset creation.
• report_tex - contains the .tex files of the Report.pdf.