This project was for the course CS286: Machine Learning on Graphs.

Graph Classification of Twitch User Networks

The main objective of this project was to develop a binary classification model to predict whether a Twitch user plays a single game or multiple games based on their ego-net.

Dataset

The dataset consists of the ego-nets of Twitch users who participated in the partnership program in April 2018, and can be downloaded from here. More information on this dataset is available here.

This dataset needed to be preprocessed into a .csv file which can be found in Processed Dataset/balanced_dataset_2000_graphs.csv file.

Machine Learning and Deep Learning approaches used

• Random Forest
• Logistic Regression
• Support Vector Machine
• Graph Convolutional Network
• Graph Attention Network

Report

A detailed report on this project can be found in the report.pdf file.

Code files

All code files can be found in the Code folder, named accordingly.