Abstract

Clustering is a method of grouping items based on shared qualities. Clustering can be used for market segmentation, crime analysis, urban landscape quality assessment, and pattern discovery in large datasets. Although a few clustering methods have linear time complexity but lower accuracy, clustering strategies are computationally costly. The K_means algorithm is a clustering algorithm that randomly select K centroids at the initial stages, the results are affected by these initial selections. To overcome this problem, an IDK_means algorithm was introduced to determine these initial centroids. IDK_means has shown better results than the K_means algorithm. This paper extends the IDK_means algorithm by improving centroid selection using Particle Swarm Optimization (PSO). PSO replaces the averaging of centroids used in the IDK_means with finding the best fit for each centroid within its cluster. IDK_means++ is tested on many datasets. The experiments were also repeated using different clusters. We applied Euclidean distance and cosine similarity and assessed the results using the Silhouette Coefficient, Davies–Bouldin Index, and Calinski–Harabasz Index, where IDK_means algorithm never outperformed IDK_means++. The paper indicates that incorporating PSO into IDK_mean gave better results for all types of datasets.

Statistics