A Novel Optimum Clustering Method Using Variant of NOA

Abstract

Cluster analysis, one of the most prominent data analysis concepts in data mining field, is constantly applied in data analysis. Escaping local optima and obtaining optimal clusters are still challenging tasks in the knowledge extraction process. This work mainly focuses on the design and development of a novel optimum clustering method. The NOA (northern bald ibis optimization algorithm), a recently developed optimization algorithm, was designed by mimicking the energy-saving flying pattern of Northern Ibis birds. Initially, the optimization performance of standard NOA is enhanced by including chaotic space transformation search (CSTS). The enhanced algorithm is termed as CNOA. The CNOA improves the capability of two prominent phases of optimization algorithm, that is, exploration and exploitation, in the search space and successfully avoids trapping into local optima. Later, CNOA is adopted in the design of a novel optimum clustering method. The optimization performance of the improved NOA is tested on seven unimodal and six multimodal benchmark functions. Numerical complexity of proposed optimum clustering is tested on ten UCI clustering problems. Descriptive statistics are considered in ranking improved NOA algorithm and proposed advanced clustering method. Based on the experimental simulations and analysis and comparisons with other popular methods, the superiority of proposed clustering method is confirmed.