Churn Prediction and Analysis Based on User Behaviour Telemetry Data of a Freemium Online Game
The thesis deals with using data from a BlueByte game and attempt to predict specific user actions, in particular possible churn. The general approach is exploratory, the exact target is not fixed. If new data characteristics or variable interactions are detected, this is already very interesting. Machine Learning and Optimization methods can be used as necessary, also the available statistical methods (possibly including, e.g., PCA, SOM, clustering methods).