Machine Learning Based Potential Customer Recommendation System for E-Commerce Using Feature Selection and XAI

In this era of digital commerce, systems that provide personalized recommendations are paramount in enhancing user experience and facilitating sales on online commerce platforms. For this reason, merchants need to know their potential customers. This work is devoted to developing a potential customer recommendation system using the well-known customer behavior and shopping habits Kaggle dataset. A customer is considered as potential customer having purchase amount greater than 70 USD and a review rating of at least 4. One of the contributions of this work is the utilization of the Select K-best with Mutual Information (MI) feature selection method in the e-commerce recommendation system. This study proposes a smart customer recommendation system using Machine Learning (ML) and Explainable AI to help merchants make data-driven decisions on which customers are the most potential ones. The proposed work makes the best use of several machine learning algorithms including Random Forest, Logistic Regression, and K-Nearest Neighbor. Among these models, Random Forest reveals the best performance of 100% accuracy. In addition, the Local Interpretable Model-agnostic Explanations (LIME) from Explainable AI are used to detect the responsible transaction records, product features and purchasing behavior of the consumers.