KNN text categorization algorithm based on semantic centre

As a classical statistical pattern recognition algorithm characterized with high accuracy and stability, KNN has been used widely in text categorization. But since KNN's time complexity is directly proportional to the sample size, its classification speed is very slow. In this paper, we propose a new KNN text categorization algorithm based on semantic centre, which we call SKNN, to speed up the classification. The basic thread is to replace the large number of original sample documents with a small amount of sample semantic centers. Experiments have proved that the SKNN's clarification is over 10 times as fast as that of the traditional KNN and its F1 value is approximately equal to SVM and traditional KNN algorithm.