The Myers–Briggs Type Indicator (MBTI) is commonly used to classify individuals in contexts such as career placement, recreation, and social matching. However, the potential of using MBTI-based clothing features for fashion categorization remains un...
The Myers–Briggs Type Indicator (MBTI) is commonly used to classify individuals in contexts such as career placement, recreation, and social matching. However, the potential of using MBTI-based clothing features for fashion categorization remains underrated, despite its significant promise. While related work exists, it primarily focuses on recommending outfits based on MBTI types. This study addresses a more fundamental research question: determining the optimal configuration method for identifying the relationship between MBTI and clothing during model training and testing. We employ an ablation study framework to evaluate the contribution of individual features and identify which are essential or redundant for accurate classification. Through rigorous benchmarking, we find that a regularized Ensemble Configuration model achieves peak accuracy (±93.2%) using concatenation, augmented image and augmented text datasets. The key insight is that in this domain, algorithmic diversity (ensembling) provides a far greater performance lift than either intensive single-model tuning or manual feature-weighting schemes.