This project is a POC of fraud detection on Minecraft using anomaly detections. The project contains a client with a click recorder, an API, a streamlit application and a jupyter notebook
The client is coded in Java and can record your clicks and also send requests to the Flask API enabling your to predict anomalies on your click data.
The API is coded in Flask, and contains a route for IsolationTree as well as well as a route for heuristics. The API also integrates a SVM (support vector machine) which is WIP and just for tests.
The notebook goes in depth about why I did this and how I tested multiple algorithms and used ITree to detect autoclicking behaviours.
The streamlit app is a data visualization app where you can upload click data as a csv file. Once uploaded you can see the proportion of anomalies in your dataset, as well as plot it.
