A k-nearest neighbor text classification algorithm based on fuzzy integral

This paper presents a k -nearest neighbor text classification algorithm based on fuzzy integral. It regards the k nearest training samples as k evidences, and fuses it using fuzzy integral, which avoids independence demand of D-S theory and improves performance of text classification. Experiment compares the new method with improved kNN algorithms and other text classification algorithms, which result shows that performance of the new method is priori to other methods and the combination of it with SVM can provide a practical resolution for cosmic text classification.