Customer Purchase Intent Prediction using Feature Aggregation on E-Commerce Clickstream Data

This paper presents a machine learning model for predicting customer purchase intent using e-commerce clickstream data. The model is built using the LightGBM framework, chosen for its efficiency in handling large-scale datasets and complex feature interactions. Key challenges addressed include the high dimensionality of clickstream data, the inherent class imbalance between purchase and non-purchase sessions, and the temporal variability of user behavior. The feature engineering process involved creating and selecting features that capture relevant user behaviors, such as session duration, event counts, and interaction diversity. The model was evaluated using ROC-AUC, F1-score, precision, and recall metrics, demonstrating strong performance in identifying sessions likely to result in a purchase. This study contributes to the field of e-commerce analytics by providing a robust framework for conversion prediction, enabling more effective customer engagement strategies. Our findings underscore the potential of machine learning to enhance e-commerce conversion rates, thereby optimizing customer engagement.