A Data Clustering Approach Using Bees Algorithm with a Memory Scheme

Abstract

The Bees Algorithm (BA) is one of the most recent swarm-based meta-heuristic algorithms that mimic the natural foraging behavior of honey bees in order to solve optimization problems and find the optimal solution. Clustering analysis, used in various science fields and applications, is an important tool and a descriptive process attempting to identify similar classes of objects based on the values of their attributes. To solve clustering problems there are diverse ways, including machine learning techniques, statistics, and metaheuristic methods. In this work, an improved Bees Algorithm with memory scheme (BAMS), which is a modified version of the BA algorithm, is used for data clustering. In the BAMS algorithm, a simple memory scheme is introduced to prevent visiting sites which are close to previously visited sites and to avoid visiting sites with the same fitness or worse. Four real-life data sets are applied to validate the proposed algorithm, and results of this study are compared to BA and others state-of-the-art methods. The experimental results show that the proposed algorithm outperforms other methods.