A parameter adaptive sine cosine algorithm for global optimization problems

Abstract Sine cosine algorithm (SCA) is a random search technique for global numerical optimization. However, SCA still exhibits low efficiency for some complex problems. In this paper, we develop a novel SCA variant (PASCA) by introducing a parameter adaptive mechanism. Specifically, the control parameter is produced based on a Cauchy distribution rather than the original linear decreasing scheme. After that, at the end of each generation, the successful historical information is utilized to dynamically update the control parameter, which helps to strengthen the search ability and reduce dependence on the problem to be solved. To verify the effectiveness of PASCA algorithm, comparison experiments are conducted on 43 benchmark functions including 13 classic problems and CEC 2017 test suite as well as a time series prediction problem. Simulation results demonstrate that PASCA is an efficient and promising optimization method compared with the classic SCA and its four recent strong variants.