PREDICTIVE ANALYTICS FOR CUSTOMER SEGMENTATION AND TARGETED MARKETING

Customer segmentation is an important component of modern marketing, as customers differ in their demographic characteristics, income levels, and spending behaviour. Traditional marketing approaches often rely on broad customer categories, which may fail to capture meaningful differences among customers. Machine learning and data analytics provide effective techniques for identifying customer groups and developing targeted marketing strategies. This study presents a data-driven framework for customer segmentation and targeted marketing using K-means clustering and statistical analysis. The analysis uses the publicly available Mall Customer Segmentation dataset, comprising 200 customer records with information on age, gender, annual income, and spending score. After examining the dataset, annual income and spending score were standardized and used as input features for clustering. K-means models with different numbers of clusters were evaluated using inertia, the silhouette coefficient, the Davies–Bouldin index, and the Calinski–Harabasz score. A five-cluster solution was selected based on its overall clustering performance and interpretability, achieving a silhouette coefficient of 0.5547 and a Davies–Bouldin index of 0.5722. The identified segments represent customers with moderate income and spending, high income and high spending, lower income and high spending, high income and low spending, and lower income and low spending. Further statistical analysis indicated significant differences in age across the identified clusters, whereas the association between gender and cluster membership was not statistically significant. Based on these customer profiles, segment-specific marketing recommendations were developed to support personalized promotions, customer engagement, and marketing resource allocation. The findings demonstrate the potential of unsupervised machine learning to identify interpretable customer segments. However, the effectiveness of the recommended marketing strategies requires further validation using actual transaction records and campaign-response data..