ABSTRACT
This study is to analyze Garena Free Fire game users, and what significant variables influence a person to purchase diamonds through a telecom games portal, by using machine learning predictive analytics. Four models build in python with optimized hyper-parameter to get optimum recall value; these are logistic regression, random forest, extra gradient boost and neural network. Results of those models are evaluated by confusion matrix and one model chosen as the selected model to predict potential buyers compared to customers who actually do not buy. Analysis of feature importance will give marketing insight into what variables must be followed up to increase customer awareness of diamond purchases through a games portal. Analytics were based on customer behaviors with 117 unique variables per customer within 390,908 customers. This research method can be used by cellular operators to know their customers so that they can determine the right promotion and marketing strategy based on the customer’s behavior.
