An improved centroid classifier for text categorization

In the context of text categorization, Centroid Classifier has proved to be a simple and yet efficient method. However, it often suffers from the inductive bias or model misfit incurred by its assumption. In order to address this issue, we propose a novel batch-updated approach to enhance the performance of Centroid Classifier. The main idea behind this method is to take advantage of training errors to successively update the classification model by batch. The technique is simple to implement and flexible to text data. The experimental results indicate that the technique can significantly improve the performance of Centroid Classifier.