Applications of matrices in big data analysis: Modeling user behavior in digital systems

Abstract

This applied study aims to implement matrix-based techniques on real-world data to model user behavior in digital systems. The study was conducted on large-scale data extracted from an e-commerce platform comprising more than 100,000 users and 50,000 products. Non-negative Matrix Factorization (NMF) and similarity matrices were employed to analyze browsing and purchasing patterns. A predictive model based on principal component analysis of matrices was developed to forecast users’ future behavior. The applied results demonstrated a predictive accuracy of 87.3% in anticipating user preferences and 82.1% in forecasting purchasing behavior. Furthermore, a recommendation system based on matrix analysis was developed, achieving a 34% improvement in click-through rate compared to traditional systems. This study provides a practical and applicable framework for leveraging matrix techniques in big data analysis for organizations and enterprises.