TOWARDS INTELLIGENT E-COMMERCE: HYBRID UNSUPERVISED LEARNING FOR DEEP SEGMENTATION OF CUSTOMER BEHAVIOR AND PURCHASE TRENDS

E-commerce has rapidly grown in India, with revenues crossing $75 billion in 2023. The Indian e-commerce market is expected to reach $188 billion by 2025, driven by the rise in internet users, digital payments, and convenience. The objective is to develop a robust hybrid unsupervised learning system to identify and segment high- and low-revenue customers in e-commerce, enabling more effective marketing, personalized campaigns, and improved customer retention. Before machine learning, businesses used manual segmentation methods like RFM (Recency, Frequency, Monetary) analysis, simple demographic segmentation, or revenue threshold analysis based on transactional history to classify customers. Traditional customer segmentation techniques are limited by static rules and lack of real-time adaptability. These methods cannot effectively capture dynamic customer behaviors, leading to inaccurate segmentation, missed revenue opportunities, and poor customer engagement. With increasing customer data complexity and the need for accurate real-time segmentation, manual methods fall short. A hybrid unsupervised learning approach address this gap by automatically segmenting customers based on their behavior patterns, enabling personalized marketing and improving business revenue strategies. Machine learning, particularly hybrid unsupervised learning models, automatically group customers by analyzing complex behavioral patterns and transactional data. This system identifies hidden segments within customer data, allowing e-commerce companies to focus on high-value customers for personalized marketing and improve retention for low-value customers.