An improved hybrid stacked classifier for multi label text categorization

Abstract

Nowadays, the applications of multi label classification are increasing very rapidly with the growth of information technology. One among it is, multi label text categorization which deals with the automatic categorization of documents or comments posted in a social site. Because of the exponential growth of digitization of unstructured categorical data, there is an emerging need for text categorization in particular with multiple labels. Conventionally, it has been solved by either transforming the problem into single class or extending the existing classifiers. An improved hybrid stacked classifier has been proposed to address the challenges in multiple label assignment for text document. The model has been built with three classifiers stacked together with Label Power set by taking class probabilities and a Meta classifier. The experimental results show that the proposed method outperforms well than the existing methods.