A new selection operator for genetic algorithms that balances between premature convergence and population diversity

Abstract

The research objective is to find a balance between premature convergence and population diversity with respect to genetic algorithms (GAs). We propose a new selection scheme, namely, split-based selection (SBS) for GAs that ensures a fine balance between two extremes, i.e. exploration and exploitation. The proposed selection operator is further compared with five commonly used existing selection operators. A rigorous simulation-based investigation is conducted to explore the statistical characteristics of the proposed procedure. Furthermore, performance evaluation of the proposed scheme with respect to competing methodologies is carried out by considering 14 diverse benchmarks from the library of the traveling salesman problem (TSPLIB). Based on t-test statistic and performance index (PI), this study demonstrates a superior performance of the proposed scheme while maintaining the desirable statistical characteristics.