Adaptive elitist-ant system for medical clustering problem

Abstract

In general, population based algorithms are superior to local search based algorithms in term of exploration the search space. In any case, the primary downside in different population based algorithms is in exploiting the search space. Recently, a Hybrid Elitist ant system approach is considered as a decent population based algorithm over various optimization problems. Along these lines, the objective of this work is to evaluate the hybrid elitist ant system approach performance to exploit the search space in clustering problem. Six benchmark datasets for medical clustering problem are utilized as a test domain (UCI Machine Learning Repository). Keeping in mind the end goal is to explore the performance of hybrid elitist ant system approach, a comparison performed between the hybrid elitist-ant system and different methodologies in the literature. The result of elitist ant system approach contrasted with different methodologies outcomes exhibits its viability. In any case, the impediment of hybrid elitist ant system approach is in tuning the importance of constraints parameter, where in each dataset needs to tune its own importance for every problem. Subsequently, its props motivate to enhance the hybrid elitist ant system approach by adaptively tuning the importance parameter. Computational outcomes demonstrate that the proposed adaptive elitist ant system approach is competent in delivering higher quality solutions (outcomes) than hybrid elitist ant system approach and different methodologies (outcomes) in the literature in all datasets.