Linear and Non-Linear Models for Purchase Prediction

In this paper, we present our approach for the task of product purchase prediction. In the task, there are a collection of sequences of click events: click sessions. For some of the sessions, there are also buying events. The target of this task is to predict whether a user is going to buy something or not in a session, and if the user is buying, which products (items) the user is going to buy. In our approach, we treat the task as a classification problem and use linear and non-linear models to make the predictions, and then build an ensemble system based on the output of the individual systems. The evaluation results show that our final system is effective on the test data.