Cluster Boost: Stacked Machine Learning Model for Customer Behaviour Analysis in E-commerce

This study focuses on the development of a stacked model, named Cluster Boost, which integrates K-means clustering and Gradient Boosting to analyse customer behaviour in e-commerce. Cluster Boost harnesses the strengths of both unsupervised and supervised learning methods, combining them to create a more robust and accurate predictive model. The proposed Cluster Boost model achieves an optimal testing accuracy of $\mathbf{9 7. 9 2 \%}$, significantly outperforming individual models such as Random Forest and Decision Tree on a public ecommerce customer churn analysis dataset. It also demonstrates an impressive precision of $\mathbf{9 5. 3 7 \%}$, a recall of $\mathbf{9 1. 9 6 \%}$, and an F 1 -score of $\mathbf{9 3. 6 4 \%}$. These results underscore the model’s potential for accurately predicting and understanding customer behaviour in e-commerce, highlighting the advantages of stacking K-means clustering with Gradient Boosting to capture intricate data relationships and enhance predictive accuracy.