2016/03/02 by Azad Naik, Huzefa Rangwala, Naik, Azad +1
Computer Science · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Face and Expression Recognition #Machine Learning and Data Classification #Text and Document Classification Technologies
paper · pdf · doi:10.48550/arxiv.1603.00772
openalex publication_date 2016/03/02 · openalex created_date 2022/10/03 · openalex updated_date 2026/07/28
Hierarchical Classification (HC) is a supervised learning problem where\nunlabeled instances are classified into a taxonomy of classes. Several methods\nthat utilize the hierarchical structure have been developed to improve the HC\nperformance. However, in most cases apriori defined hierarchical structure by\ndomain experts is inconsistent; as a consequence performance improvement is not\nnoticeable in comparison to flat classification methods. We propose a scalable\ndata-driven filter based rewiring approach to modify an expert-defined\nhierarchy. Experimental comparisons of top-down HC with our modified hierarchy,\non a wide range of datasets shows classification performance improvement over\nthe baseline hierarchy (i:e:, defined by expert), clustered hierarchy and\nflattening based hierarchy modification approaches. In comparison to existing\nrewiring approaches, our developed method (rewHier) is computationally\nefficient, enabling it to scale to datasets with large numbers of classes,\ninstances and features. We also show that our modified hierarchy leads to\nimproved classification performance for classes with few training samples in\ncomparison to flat and state-of-the-art HC approaches.\n