Yin-Yang Firefly Algorithm and a Research on Its Application in Global Optimization Problems

Aiming at the shortcomings of Firefly Algorithm (FA) which is prone to premature and unstable convergence, a novel Yin-Yang firefly algorithm (YYFA) was proposed based on the connotation of China Yin Yang theory. Firstly, a new randomly attraction model was designed to reduce the time complexity of the algorithm. Secondly, a Yin-Yang firefly deep learning strategy which arises from the idea of Deep Learning (DL) was employed to tap into valid information from the current optimal firefly to guide the firefly swarm to move towards a better direction. The experiment results on 13 typical test functions for global optimization show that YYFA algorithm has a better performance in terms of the ability and stability in global optimization in comparisons with different combined strategy algorithms. And the Friedman test outputs reveal that the proposed Yin-Yang firefly deep learning strategy is an effective improvement method. Finally, YYFA algorithm was applied to deal with an annual runoff forecasting problem from Huangfuchuan Station. The YYFA-SVR prediction model was constructed via using YYFA to find preferred hyperparameters in support vector regression (SVR) model. The application results demonstrate that the forecast effect of YYFA-SVR model is better than BOA-SVR model, WOA-SVR model and ESDA-SVR model, and has a higher forecast accuracy, which can provide new ideas for related forecasting work. © 2022, The Editorial Board of Journal of Basic Science and Engineering. All right reserved.