A novel text classification based on Mahalanobis distance

In text mining field, The KNN (K Nearest Neighbors) is one of the oldest and simplest methods of text classification. But it is known to be sensitive to the distance (or similarity) function used in classifying a test instance, this disadvantage can cause low classification accuracy and limit the KNN classifier´s utilization in text classification in text mining. In this paper, we introduce Mahalanobis distance in text classification area, and proposed an algorithm (MDKNN) base on this theory. Experiment show that our method has comparable or better performance than KNN Classifier and Naïve Bayes classifier in text classification.