A Novel Neighborhood Granular Meanshift Clustering Algorithm

Abstract

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

Combining Improved Meanshift and Adaptive Shi-Tomasi Algorithms for a Photovoltaic Panel Segmentation Strategy

“…Regarding the fundamental Meanshift formulation described in Equations ( 1) and ( 2), there exists an issue: within the region of S h , every point contributes equally to x [17]. However, in reality, this contribution is related to the distance from x to each point.…”

Dynamic analysis and optimal control of a stochastic investor sentiment contagion model considering sentiments isolation with random parametric perturbations

“…And then apply the DCFS model to predict a synthetic chaotic plus random time series and the Hang Seng Index of the Hong Kong stock market. Chen et al 26 found that the granular mean shift clustering algorithm has better clustering performance than traditional clustering algorithms, such as Kmeans, Gaussian mixture, etc. Sang et al 27 proposed a fuzzy rough feature selection method based on robust non-linear vague quantifier for ordinal classification.…”