Consumer Psychological Modeling and Behavior Prediction System Based on Graph Neural Network

E-commerce, where billions of people engage daily, requires an understanding of consumer psychology. CNNs, RNNs, and matrix factorization are commonly used in recommender systems, although they often overlook psychological factors and user-product-context relationships. This research combines consumer psychology with advanced graph-based learning to provide a psychologically informed prediction framework. Latent cognitive and emotional aspects are modeled to increase customer behavior prediction accuracy, customization, and interpretability. The proposed NeuroGraph-CPM creates a heterogeneous user–product graph including behavioral data, environmental information, and psychological signals from reviews and interactions. The model captures structural links and hidden psychological states that influence consumer decision-making using graph neural networks with affect-aware message forwarding and psychologically regularized attention. Research on a real-world Amazon Electronics dataset shows that NeuroGraph-CPM improves behavior prediction accuracy by 19.6%, click-through rate estimation by 16.3%, and personalization relevance by 21.8% over strong baseline models. Further study demonstrates that the model delivers human-centered transparency with interpretable attention distributions linked with psychological priors. The findings show that psychology and graph learning improve recommender system predictive performance, user engagement, and customisation.