A new algorithm based on centroid for text categorization

Text categorization is a hot topic and a key technology in data mining and information retrieval, so that it received wide attention recently. Centroid-based algorithm is an effective and robust approach. However it often suffers from the inductive bias or model misfit. In order to solve this problem, many researchers have put forward a number of improvement strategies which makes the centroid-based algorithm have a better performance. The paper proposed a novel approach to adjust the centroids which is called Weighted Margin adjusted Centroid based Algorithm (WMCA). Then it presented a lot of experimental comparison with some other algorithms by using 5 different public corpuses. The results showed that the WMCA algorithm has the best performance.