NI-DBSCAN: DBSCAN under Non-IID

Abstract

Abstract DBSCAN (Density Based Spatial Clustering of Application with Noise) is an example of density-based clustering algorithm. Aiming at problem that DBSCAN algorithm assumes that the data are independent and identically distributed and the traditional distance formula is difficult to accurately calculate the similarity degree between categorical data. Density Based Spatial clustering algorithm of Application with Noise under Non-IID (NI-DBSCAN) is proposed. The unsupervised clustering problem of categorical data is dealt with by means of the Non-IID (non-independent and identical distribution) thought. Using coupling similarity to measure similarity can better reflect the “real relationship” between categorical data. The experimental results on the UCI dataset show that the algorithm can obtain satisfactory clustering results and improve the applicability and accuracy of the algorithm.