Abstract
Text clustering is one of common techniques in mining large scales of document data. The paper presents an improved K-Means text clustering algorithm in which a local search mechanism is introduced. By the iteration process of K-Means algorithm, our approach can quickly get a local extreme point, and then use the search strategy of local search mechanism to have K-Means jump out of that point and get a better solution. The experimental results show that our approach achieves better performance in the terms of entropy than the traditional algorithm while not slowing down the clustering speed.