An Improved Method Based on the Density and K-means Nearest Neighbor Text Clustering Algorithm

For k-means algorithm to the initial cluster centers sensitive to outliers shortcomings, we propose a density-based method to improve the k-means algorithm. Density-based methods are used, by setting the neighborhood and the neighborhood of the object that contains at least to exclude isolated point, and will not repeat the core point as the initial cluster centers We use the ratio of the distance between the distance and class within the class as a criterion evaluation function, the number of clusters to obtain the minimum value of the criterion function as the best number of clusters. These improvements effectively overcome the shortcomings of K-means algorithm. Finally, a few examples of the improved algorithm introduces specific application examples show that the improved algorithm has a higher accuracy than the original clustering algorithm, can help achieve tight class within the class room away from the clustering effect.