Machine Learning-Based Analysis of Emotional Responses to Food Labels: A Case Study of Thai Young Adults

Understanding the emotional drivers of consumer choice is critical for effective food packaging design. This study proposes a novel 'Emotion-AI Framework' to decode consumer responses to ten processed fish product labels using the circumplex model of emotion. Explicit emotional responses and purchase intentions were collected from 100 participants, and unsupervised machine learning (K-Means clustering) successfully classified consumers into three distinct segments (Enthusiasts, Passives, and Rejectors) strictly based on their multidimensional emotional profiles. Furthermore, a supervised Random Forest regression model, coupled with permutation feature importance, revealed that aggregated emotional states (specifically the low-arousal/pleasant and high-arousal/unpleasant quadrants) are the dominant drivers of purchase intention. Crucially, these emotional states significantly outperformed the direct impact of physical label attributes. The findings demonstrate that integrating theoretical emotional models with predictive machine learning provides robust, data-driven insights for the food industry, enabling the optimization of product labels to evoke targeted affective states and maximize consumer acceptance.