Multi-Level Clustering Approach for Customer Behavior Analysis in Data-Driven Marketing

The current study presents a comprehensive multi-level clustering model to investigate customer behavior in the context of data-driven marketing. Utilize the abundantly available Online Retail dataset, the study initiates by using RFM (Recency, Frequency, Monetary) analysis to extract prevalent behavioral characteristics. Subsequently, two prominent clustering models, namely K-Means and Gaussian Mixture Model (GMM), are used to segment customers. To further enhance the clustering outcome, secondary clustering methods like DBSCAN and Agglomerative Clustering are applied to the preliminary output of K-Means and GMM. Silhouette Score and Davies-Bouldin Index are used to measure the performance of every clustering configuration and through it, it is proved that the K-Means and DBSCAN combination provides the best segmentation performance. The results offer valuable insights for marketers to better understand and classify customer behavior patterns, as well as future research opportunities with more heterogeneous data sources and some new hybrid methods.