A Social Spider Optimization Algorithm with Chaotic Initialization for Robust Clustering
Sakshi Aggarwal, Parijeet Chatterjee, Raj Prakash Bhagat, Keshav Kr. Purbey, Satyasai Jagannath Nanda · Procedia Computer Science · 2018
Several real-life datasets at times contain few observations whose characteristics are completely different from the rest of the patterns. Such patterns occur due to faulty readings taken or due to presence of noise are termed as outliers. In this paper a recently developed algorithm Social Spider Optimization (SSO) is employed for robust clustering. The robust clustering takes care of the outliers with the use of Robust Distance (RD) as the cost function. The SSO is a swarm intelligence algorithm inspired by the cooperative behavior of spiders on a web. In this paper the initialization of spiders are carried out by Chaotic sequence which enhances its performance compared to the original SSO which uses random sequence. The simulation studies are carried out on two synthetic and three real life datasets. Comparative analysis reveals superior performance of proposed Chaotic SSO (C-SSO) over Original SSO, Chaotic PSO, PSO, Chaotic K-means and K-means algorithms. It is observed that the Chaotic versions of the algorithms have better accuracy ( here represented by lower mis-classification index) over the regular algorithms.