A Novel Batch Framework-Based Performance Improvement of Evolutionary Algorithm

Evolutionary algorithms are widely used methodologies for problem-solving and are covered under the umbrella term of evolutionary computation (EC). They are stochastic techniques that take inspiration from natural evolutionary processes to solve complex optimization problems. Numerous studies have been conducted in a quest to improve the performance of the evolutionary computation framework. In a similar row, we have proposed to incorporate batch approach along with one of the most widely used evolutionary algorithms, i.e., genetic algorithm (GA), to enhance its performance. We named it Batch GA. The proposed study is about improving the algorithm’s search strategy, which incurs numerous associated benefits when put to use. It initiates by dividing the population into several batches, exposing each to different crossover and mutation probabilities, which are further made to run over specified iterations and finally get merged to form the population. Due to several batches tuned to different parameters, this methodology enhances search capability by doing better exploration and exploitation of search space. The credibility of the approach is tested over 20 benchmark functions and five classical engineering problems. The results obtained confirm the better performance of the batch approach over its non-batch counterparts. This approach incurs less overhead as compared to other state-of-art techniques. Thus, it is easy to understand and implement. Also, it is a general concept and can be incorporated with other evolutionary algorithms as well.