Researches Advanced in Clustering Algorithms

Abstract

Clustering is a technique to find the intrinsic structure between data and is a fundamental problem in many data-driven application fields. Currently, clustering is generally modeled as an unsupervised learning task, aiming to mine similar features between different samples and cluster samples with similar features into clusters. Ideally, objects in the same cluster are expected to be similar in the clustering results, while objects in different clusters are quite different. This study summarizes the research status of clustering algorithms in recent years. Specifically, the relevant critical steps of clustering algorithms are first introduced. From two aspects of partition and hierarchical clustering, representative clustering algorithms such as K-means, K-medoids, CLARANS, BIRCH, DBSCAN, and CURE are further detailed. This study also analyzes and summarizes the above algorithms in terms of critical technologies, algorithm ideas, benefits, and shortcomings and compares the distance accuracy of different algorithms on standard data sets. The above work will provide a valuable reference for cluster analysis and data mining research.