The Role of Big Data Analytics in Predicting Customer Purchase Behavior in Iranian Online Clothing Stores with Emphasis on Transactional Data and Social Media Interactions

Abstract

ABSTRACT

The present study aimed to develop and empirically test a big data analytics model for predicting customer purchase behavior in Iranian online clothing stores by integrating transactional purchase data with social media interaction indicators. This applied quantitative study employed a predictive analytics design based on big data methodology. The research population consisted of active customers of major online clothing retailers in Tehran, Iran, from which behavioral data of 1,248 verified users were extracted. Data were collected through integrated digital sources including e-commerce transactional databases, customer relationship management systems, and social media analytics platforms. Transactional variables included purchase frequency, order value, browsing behavior, cart abandonment rate, and discount usage, while social media indicators captured engagement intensity, sentiment polarity, influencer exposure, and interaction responsiveness. Data preprocessing procedures involved cleaning, normalization, feature engineering, and behavioral profile integration. Predictive modeling was conducted using machine learning algorithms including Random Forest, Gradient Boosting, Support Vector Machine, and Artificial Neural Networks. Cluster analysis was also applied to identify customer behavioral segments, and model performance was evaluated using accuracy, precision, recall, F1-score, and AUC indicators. Results indicated significant positive relationships between social media engagement, customer loyalty, sentiment toward brands, and repurchase intention. Machine learning models demonstrated high predictive capability, with Artificial Neural Networks achieving the strongest performance in forecasting purchase probability. Behavioral segmentation identified four statistically distinct customer groups characterized by loyalty orientation, promotion sensitivity, social influence dependency, and occasional purchasing patterns. Engagement-based variables exhibited stronger predictive power than price-related indicators, suggesting that emotional interaction and digital experience play a more influential role than discounts alone. Integrated models combining transactional and social interaction data significantly improved prediction accuracy compared with single-source behavioral models. The findings confirm that big data analytics provides an effective framework for predicting online clothing purchase behavior by capturing multidimensional consumer interactions across digital environments. Integrating transactional records with social media behavioral signals enables retailers to understand customer decision processes more accurately, optimize marketing strategies, enhance customer loyalty, and support data-driven retail management in competitive e-commerce ecosystems.