The basic SOA has some drawbacks such as low solution accuracy, slow convergence rate, and easy to fall into local optima while solving complex function optimization problems. This paper proposed an improved SOA based on somersault foraging strategy(SFSOA). Firstly, the proposed SFSOA updated the positions of seagull individuals by using nonlinearly decrease strategy of control parameter A based on the inverted sigmoid function. This way improved individual solution quality and accelerates convergence speed of algorithm. Then, it introduced novel learning mechanism based on the somersault foraging strategy for the global best individual to increase the position diversity of seagull individual and avoid the algorithm falling into a local optimum at the lately stage. The proposed SFSOA was compared against the basic SOA, grey wolf optimizer(GWO), and improved SOA(ISOA) by using eight complex benchmark function optimization problems. Experimental results reveal that the proposed SFSOA has higher solution accuracy, faster convergence rate and stronger global search ability than its competitors and can effectively deal with complex function optimization problems. Finally, SFSOA is applied to solve the feature selection problems and the satisfactory results are obtained.

[2022-07-19] Accepted Paper /doc/2022.05.0224/v1

[2022-12-05] Printed Article /doc/2022.05.0224/v2