Player Pattern Prediction Using Action Logs of Players

The hike in mobile games has changed the game industry's outlook. Plenty of information about the players are now available to Game developers and thus can predict the players pattern using reliable models. Predicting the player's exit moment in a game generates several opportunities to understand and improve players' lifetime and revenue earnings. Churn prediction, a common challenge faced by variety of sectors, is also one of the most important problem for gaming industry, as player retention is critical for the monetization of a game. Users inclination towards a game and churn prediction in advance can help us to increase profit through effective services. The paper proposes dynamic difficulty algorithm which provides predictions on accumulated Playtime and Number of Sessions until that moment. It is well suited for real time analyses, even with million users for games. The method is evaluated by experimenting some of the classifiers. The result shows that the approach is well defined and successfully applicable to various datasets and response variables.