The rapid growth of e-commerce platforms has generated massive volumes of customer transaction data, making intelligent customer segmentation essential for personalized marketing and customer relationship management. Traditional segmentation techniques often fail to capture complex and sequential purchasing behaviours, resulting in less effective marketing decisions. This paper presents a deep learning framework for consumer segmentation using the Online Retail dataset from the UCI Machine Learning Repository. The approach is built on Long Short-Term Memory (LSTM). The proposed methodology includes comprehensive data preprocessing, label encoding, RFM (Recency, Frequency, Monetary) feature engineering, Min-Max feature scaling, and supervised LSTM model training. The framework classifies customers into meaningful behavioural segments, including High-Value, Loyal, Potential, Occasional, and At-Risk customers, enabling businesses to implement targeted marketing strategies. The proposed model demonstrates excellent predictive performance, achieving 99.77% accuracy (ACC), 99.39% precision (PRE), 99.93% recall (REC), 99.66% F1-score (F1), and 99.94% ROC-AUC. Comparative analysis with existing customer segmentation approaches confirms the superiority of the proposed LSTM framework in learning customer purchasing patterns. The findings indicate that the proposed approach provides an effective and reliable solution for intelligent customer segmentation, personalized recommendations, and enhanced marketing decision-making in modern e-commerce platforms.