Research on Intelligent Classification System for Artworks Based on Random Forest Feature Selection and Hierarchical Retrieval

With the rapid development of digital technology, the handling of high-dimensional data such as digital art collections in classification and retrieval has become an important research topic. However, traditional methods often face the challenges of feature redundancy and retrieval efficiency, especially in large datasets where performance is easily limited. To address these issues, this study proposes an intelligent classification and retrieval system that incorporates random forest feature selection and hierarchical retrieval strategies. The system improves classification accuracy by reducing feature dimensionality, while using structured retrieval paths to improve query efficiency. We conducted experimental validation on a high-dimensional art dataset. The results show that after feature selection, the classification accuracy improves from 85.0% to 92.5%, and the macro-averaged F1 score improves from 0.83 to 0.91, while the training time is reduced by about 35%. The hierarchical retrieval strategy also significantly improves response time from an average of 1.2 seconds to 0.45 seconds and mAP from 0.79 to 0.87. These results confirm the effectiveness and usefulness of our method in dealing with high-dimensional data. The innovation of this study is to combine random forest and hierarchical retrieval into a scalable system framework that is not only applicable to art data, but can also be generalized to other domains, providing theoretical support and practical references for subsequent research and practical applications.