An integrated interpretation and clustering model based on attribute grouping

Clustering is a technique in unsupervised learning used to group unlabeled data. However, traditional clustering algorithms cannot provide explanations for the clustering process and its results, which limits their applicability in certain fields. Existing methods to address the lack of interpretability in clustering algorithms typically focus on explaining the results after the clustering process is complete. Few studies explore embedding interpretability directly into the clustering process, and most of these methods rely on data prototypes to express interpretability, which often leads to explanations that are not intuitive and user-friendly. To address this, a feature-based method is proposed to embed interpretability into the clustering process. This approach provides users with intuitive and easy-to-understand explanations and introduces a new direction for research on embedding interpretability into clustering. The method operates in two stages: in the first stage, all attributes are grouped; in the second stage, an optimization formula is used to complete both the clustering and the weighting of each attribute group. The proposed method was evaluated on multiple synthetic and real-world datasets and compared with other methods. The experimental results show that the method improves clustering accuracy by approximately 5 percent and interpretability by around 40 percent compared to existing approaches.