An empirical study to predict churn of online multiplayer games and its impact on revenue of the game developing company

Abstract

Online multiplayer games are becoming massively popular nowadays. However, the churn of the players is becoming a significant concern as it is challenging to predict whether a player will churn or not, impacting business revenue. In this research, authors tried to solve this problem by predicting the player churn in advance using predictive analytics, thereby enabling the business owners to undertake steps to prevent player churn resulting in revenue stability. To achieve this, the authors collected the data from online and gaming platforms and then applied various pre-processing steps such as data conversion to make data suitable to use and then tested and applied a machine learning-based model for prediction by selecting churn period as the threshold value. Finally, various classifiers, such as logistic regression, were applied to predict whether a player will churn. The results were very satisfactory, as predicting churn with perfect accuracy was possible. The decision tree provides the best results, which were proximately 99.1 %, and other algorithms like logistic regression, random forest, and Adaboost gave predictive results of 96.86 %, 95.47 %, and 98.8 %, respectively. The accuracy of all the models has also been summarised. Hence, by making predictions in advance, the online platforms will take preventive measures to minimize the churn of players and increase revenue accordingly.