Comparison of multilabel problem transformation methods for text mining

Primarily, the need for automatic text categorization and medical diagnosis was the start of Multi-label classification. Multi-label classification received a great attention and used in several real world applications The demand of its applications increased to cover additional fields like functional genomics, music, biology, scene, video etc. For example, a text document may belong to many subjects or topics like Scientific, Cultural, or Politics. There exist a variety of multi-label classification algorithms developed based on two basic approaches: algorithm adaptation and problem transformation. Our contribution consists to present an analysis and experimental comparison of 4 problem transformation algorithms applied to two text benchmark datasets using 4 evaluation measures. In the experimental study, each problem transformation method is applied against all 54 classifiers found in the MEKA software in order to find the classifier that gives the best performance for each dataset and classification method.