Customer Loyalty Prediction and Customer Segmentation: A Random Forest Prediction and PCA-Integrated K-Means++ Approach

This research discusses the prediction of company loyalty. Every company typically has a loyalty program in place. Customer loyalty is a key driver of business success, especially in competitive retail environment. Understanding the condition of customer loyalty presents a unique challenge for companies. It is also crucial for companies to understand the characteristics of their loyal customers. This research proposes a predictive model to assess the loyalty of a company's customers and builds a clustering model from the predicted loyal customers. Using the company's transaction data from January 1, 2023, to September 31, 2024, a predictive Random Forest model is developed. The Random Forest model is evaluated using ROC-AUC, accuracy, F1 Score, and Recall, showing ROC - AUC 0.93, accuracy of 87%, a F1 Score of 91 %, and a recall of 92 %. performance. The Random Forest model will generate customers predicted to be loyal and those predicted to churn. K-Means++ clustering will be built using the data of customers predicted to be loyal by the Random Forest model. Clustering is implemented by PCA (Principal Component Analysis) to reduce data dimensionality. Loyal customers are divided into 3 clusters, each displaying its distinct characteristics. This research identifies customers who have the potential to become loyal, and the clustering results reveal the characteristics of each cluster. Companies can leverage the findings of this research to develop better strategies to enhance customer loyalty.