A Review of Clustering Algorithms: Evolution, Comparison, and Challenges from Traditional Clustering to Deep Clustering

Against the background of big data, clustering algorithms, as a core task of unsupervised learning, have been widely applied, and the related theories and technologies have continued to iterate and develop. Existing reviews mostly focus on the classification and comparison of traditional clustering algorithms, while insufficient attention has been paid to the evolutionary relationship between traditional clustering and deep clustering. Therefore, this paper reviews clustering algorithms and their evolutionary logic, systematically sorts out the evolutionary route of clustering algorithms from traditional methods to deep models, describes the relatively classic and main clustering algorithms, and compares and analyzes their advantages and disadvantages. It also shows the key transition from classic clustering to deep clustering. In addition, this paper describes the core challenges that clustering algorithms may face in the future and analyzes further research directions according to current research progress.