การจัดกลุ่มข้อมูลด้วยขั้นตอนวิธี เคอินเวอร์สฮาร์โมนิกมีน

Abstract

Data clustering is an important process for data analysis. It is an unsupervised learning utilizing the similarity of data. An ordinary clustering algorithm is K-Means clustering (KM). Its computation is simple and it is showing satisfaction with clustering performance. The K-Harmonic means (KHM) is developed from the KM. The Min function of KM is replaced by the harmonic mean function. This yields the better clustering results. This paper presents an improvement of KHM called K-Inverse harmonic means clustering algorithm (KIHM). The harmonic means of the Euclidean distance is replaced by the harmonic means of reverse radial basis function of the distance from the cluster center. Experimental results show that the iterations of computation are reduced. Once the experiments are conducted on the Thai web snippets data, KIHM shows better clustering results than that of KHM.