A Novel Neighborhood Granular Meanshift Clustering Algorithm

The most popular algorithms used in unsupervised learning are clustering algorithms. Clustering to group samples into a number of classes or clusters based on the distances given sample features. Therefore, how define distance between is important for algorithm. Traditional generally Mahalanobis and Minkowski distance, which have difficulty dealing with set-based data uncertain nonlinear data. To solve this problem, we propose granular vectors relative absolute neighborhood granule operation. Further, meanshift algorithm also proposed. Finally, effectiveness proved from two aspects internal metrics (Accuracy Fowlkes–Mallows Index) external metric (Silhouette Coeffificient) multiple datasets UC Irvine Machine Learning Repository (UCI). We find that has better effect than traditional algorithms, such as Kmeans, Gaussian Mixture so on.