Fuzzy Interval Number K-Means Clustering for Region Division of Pork Market

Abstract

This article proposes a new clustering algorithm named FINK-means. First, this article converts original data into a fuzzy interval number (FIN). Second, it proves the F that denotes the collection of FINs is a lattice. Finally, it introduces a novel metric distance on the lattice F. The contrast experiments about FINK-means, k-means, and FCM algorithm are carried out on two simulated datasets and four public datasets. The results show that the FINK-means algorithm has better clustering performance on three evaluation indexes including the purity, loss cost, and silhouette coefficient. FINK-means is applied to the task of region division of pork market in China based on the daily data of pork price for different provinces of China from August 9, 2017 to August 9, 2018. The results show that regions of pork market in China was divided into five categories, namely very low, low, medium, high, and very high. Every category has been discussed as well. At last, an additional experiment about region division in Canada was carried out to prove the efficiency of FINK-means further.