Hierarchical rough decision theoretic framework for text classification

Hierarchical classification problems have been wide investigated in the past years. The available hierarchical classification methods, which use the top-down level-based scheme, often suffer from the burden of inter-level error transmission. In this paper, an instance-centric hierarchical classification framework based on decision-theoretic rough set model is proposed. The procedure of classification will be divided into two phases. Firstly, a hierarchical rough decision model is constructed to acquire all possible paths as well as reduce error transmission. A general loss function for supervised leaning is also defined by which the cost and benefit of assigning an instance to a specific subcategory can be evaluated. Subsequently, a novel classification routing method special for support vector machine is put forward in order to select an optimal classification path. Comparative experimental results with Chinese text classification benchmark TanCorp illustrate the effectiveness of proposed notions.