Deep Learning-Enhanced Analysis of Consumer Decision-Making Behavior

Understanding consumer decision-making in digital environments necessitates analytical models that are both scalableand adaptive. Traditional econometric and rule-based models often fail to capture the non-linear, context-sensitive, andmulti-modal nature of real-world decision behavior. This study introduces a deep learning–enhanced framework that integratescognitive modeling with neural architectures to overcome these limitations. Central to the framework is the Attentive UtilityReconstruction Network (AURN), a novel model that learns utility-aware latent representations by fusing multi-modal consumerdata, including attribute embeddings, historical behaviors, and contextual signals. AURN features an attention-gated utilitydecoder and a recurrent preference updater, enabling it to capture both static preferences and temporal dynamics. Byemploying temperature-controlled softmax and probabilistic attention distributions, AURN reconstructs personalized utilityfunctions with enhanced interpretability and flexibility. To further embed behavioral rationality, the framework introduces theContext-Responsive Choice Optimization (CRCO) strategy. CRCO augments AURN with domain-informed regularization,including structural priors (e.g., monotonicity, dominance), decoy-aware counterfactual penalties, and context-sensitive decisiontemperatures. It also incorporates behavioral biases such as salience, default bias, and compromise effects into utilityestimation. Together, these elements ensure predictive robustness and psychological plausibility across diverse consumersegments. Empirical results across four benchmark datasets demonstrate consistent outperformance of the proposed modelover state-of-the-art baselines in terms of accuracy, AUC, and F1 score. This integrated framework not only improves predictionperformance but also advances interpretability, offering a comprehensive tool for understanding digital consumer cognition.