An improved kNN learning based korean text classifier with heuristic information

Automatic text categorization is a problem of assigning predefined categories to free text documents based on the likelihood suggested by a training set of labelled texts. kNN learning based text classifier is a well known statistical approach and its algorithm is quite simple. While the method has been applied to many systems and shown relatively good performance, a through evaluation of the method has rarely been done. There are some parameters which play important roles in the performance of the method: decision function, k value of kNN, and size of feature set. This paper focuses on an improving method for a kNN learning based Korean text classifier by using heuristic information found experimentally. Our results show that kNN method with carefully chosen parameters is very significant in improving the performance and decreasing the size of feature set.