Efficient and Intelligent Density and Delta-Distance Clustering Algorithm

Density and delta-distance clustering (DDC) is an ideal clustering method that computes the density and delta distance of data. When data derived from the two indicators are large, these areas can be defined as cluster centers. DDC has good clustering performance compared with some other clustering algorithms. However, DDC has a high time complexity and requires manual identification of cluster centers. To fill these gaps, an efficient and intelligent DDC (EIDDC) algorithm is proposed in this study. EIDDC begins from using a sampling method based on locality-sensitive hashing (LSH) to obtain a small-scale dataset. The density and delta distance of each data point are calculated from this dataset to reduce time complexity. Cluster centers are intelligently recognized by utilizing density-based spatial clustering of applications with noise-based outlier detection technology. Experiment results show that LSH can obtain good representatives of the original dataset and that the proposed outlier detection method can recognize the cluster centers of a given dataset. The results also reveal the efficiency of EIDDC.