Customer Segmentation In Financial Services Using Gaussian Mixture Models (GMM) And K-Means

Abstract

In the financial services industry, customer segmentation has emerged as a vital analytical process for delivering personalized banking, customer retention, and intelligent decision-making. In banking datasets, which are typically huge, traditional segmentation methods fall short in capturing dynamic and overlapping behaviors of customers. This research introduces an AI-based customer segmentation approach, based on K-Means clustering and Gaussian Mixture Models (GMM) on publicly available Indian banking transaction data, which consists of more than 1 million banking transaction records and more than 800,000 customers. It includes data preprocessing, feature engineering based on Recency-Frequency-Monetary (RFM), Principal Component Analysis (PCA), and probabilistic clustering for meaningful customer group identification. The experimental analysis showed that PCA was able to retain 91.4% of the variance of the data set while at the same time reducing the dimensionality of the feature space. The Silhouette Score of K-Means is 0.742, while the Davies-Bouldin Index is 0.481; GMM improved the results in comparison to K-Means, with a Silhouette Score of 0.861 and Davies-Bouldin Index of 0.294. The proposed framework was able to successfully identify the five major segments of customers, such as premium customers, digitally active customers, dormant customers, and high-risk transaction groups. Results validate the modeling of the customer behavior and the quality of the customer segmentation in a financial environment by using probabilistic clustering. The proposed framework offers a lot of support to the provision of personalized banking services, customer intelligence generation, fraud monitoring, and strategic financial decision-making.