Testing the fraud detection algorithms of online chess platform and exploring ways to improve them using data mining techniques

Abstract

This thesis is to analyze the efficiency implemented chess fraud detection algorithms and research the possibilities of their improvement with data mining methods and techniques. The testing was performed by creating a bot for playing chess online guessing about possible parameters analyzed by the system, which should be detected during its work. As it remained unnoticed, the research for improvement was performed getting additional parameters to include into detection analysis and discovering methods for distinguish chess engine game behavior pattern. The methods were implemented in Python (Keras, Selenium) with the help of other tools as Google Colaboratory, RapidMiner. Leela chess engine was used as core for bot. Dataset for analysis was provided as open data by lichess.org.