Modeling Rating Uncertainty with Interval Type-2 Fuzzy Decision Trees in Multi-Criteria Recommendation

Managing uncertainty in user-generated data is a persistent challenge for multi-criteria recommender systems. This study introduces a redundancy-based interval type-2 (IT2) fuzzy approach that models rating ambiguity across five movie evaluation criteria—overall, story, acting, direction, and visuals. Using Histogram-Based Interval Type-2 Fuzzy Membership Functions (HBIT2FMFs), the method captures both membership values and their uncertainty bounds. User rating vectors are first clustered with hierarchical density-based spatial clustering of applications with noise (HDBSCAN) to identify dense regions, then refined via k-means into subclusters that serve as categorical targets for interval type-2 IT2 fuzzy ID3 and C4.5 decision trees. Evaluations on the Yahoo! Movies dataset show consistent improvements in MAE, precision, recall, and F1-score over crisp and type-1 models. Personalized recommendations are generated by computing cosine similarity within each subcluster. Overall, the framework provides an interpretable, robust, and uncertainty-aware solution for multi-criteria collaborative recommendation.