A STUDY ON THE MODELING OF SIGNIFICANT CROSS EFFECTS IN PURCHASE INCIDENCE: ANALYZING ARTIFICIAL NEURAL NETWORK METHODS AND MULTIVARIATE PROBIT MODELING

Abstract

This study focuses on modeling enormous cross-effects in purchase incidence in hopes of better understanding consumer behavior across related product categories. It tests how well many modeling approaches, including Artificial Neural Networks (ANNs) and Multivariate Probit (MVP), capture complex cross-category interactions. While MVP models provide a clear statistical framework grounded on economic theory, ANNs are known for their adaptability when dealing with large datasets and nonlinear interactions. Studies evaluate these methods by comparing their accuracy in predicting results, clarity of presentation, and ability to capture the nuances of buying behavior across various product types. The findings demonstrate that MVP models provide much higher interpretability and theoretical consistency compared to ANNs, despite the fact that ANNs perform pretty well in terms of prediction accuracy. The findings add to broadening methodologies in consumer choice modeling and give practical insights into marketing strategies and decision-making. To examine large-scale cross effects on sales incidence, this study compares and contrasts the effectiveness of ANN approaches with MVP modeling. The goal of this research is to provide light on the relative strengths of two cutting-edge methods for understanding complicated relationships and making predictions about consumer behavior in large datasets by comparing and contrasting them. By using ANN and MVP algorithms to purchase incidence data, this study intends to assess the methodology's capacity to provide more precise and actionable insights into consumer behavior. This comparative study aims to assist researchers doing marketing or consumer research in selecting the most appropriate modeling approaches by focusing on model accuracy, computing efficiency, and interpretability.