Parameter Optimization of a Stochastic Configuration Network Based on the Chaotic Feedback Sooty Tern Optimization Algorithm

To solve the problem that the selection and assignment of hidden layer parameters of stochastic configuration network (SCN) affects its prediction accuracy, a SCN parameter optimization method was proposed based on the chaotic feedback sooty tern optimization algorithm (CFSTOA). First, the Tent mapping, linear factor adjustment strategy and disadvantaged population feedback principle were used to improve the sooty tern optimization algorithm (STOA) to enhance the local search capability and to obtain a CFSTOA with faster convergence speed and higher convergence accuracy. Then, the CFSTOA was used to optimize the regularization parameter and the scale factor of weight biases of the SCN to obtain the optimal hidden layer parameters. Finally, the performance of the CFSTOA was tested through 10 benchmark functions and 4 standard regression datasets. Results show that the CFSTOA has a faster convergence speed and is less likely to fall into a local optimum, which can improve the prediction accuracy and training speed of the SCN algorithm.