HYBRID DBSCAN-GMM CLUSTERING MODEL FOR EFFECTIVE CUSTOMER SEGMENTATION: A TECHNICAL APPROACH FOR PERSONALIZED MARKETING STRATEGY

Abstract

Customer segmentation plays a vital role in developing personalized marketing
strategies that enable organizations to target consumers more effectively.
Conventional clustering methods such as K-means and Recency, Frequency,
and Monetary analysis often encounter limitations when applied to complex
retail data, particularly in managing noise, sparse purchasing patterns, and nonspherical clusters. This study introduces a hybrid approach that integrates
Density-Based Spatial Clustering Applications with Noise and Gaussian
Mixture Model to address these challenges. Density-Based Spatial Clustering
Applications with Noise first identifies dense regions of customer data and
separates noise, after which Gaussian Mixture Model probabilistically refines
the clusters for greater accuracy and flexibility. Data were obtained from Kaggle
and the UCI Machine Learning Repository and analyzed using the proposed
hybrid model. The hybrid Density-Based Spatial Clustering Applications with
Noise and Gaussian Mixture Model approach demonstrated a clustering
accuracy, reduced misclassification, and provided better adaptability for largescale and datasets. Findings indicate that this model supports more precise
customer profiling, thereby enabling personalized marketing strategies that
enhance customer engagement, loyalty, and organizational profitability.