Fig. 4
The initial value is preset to initialize the ABC algorithm, and the dimension of variables and boundary conditions are set. The fitness function is applied as F(xi) = 1/MSEi, where MSEi represents the root mean square error of the MLS-SVM of i th solution, MSEi of the initial MLS-SVM training sample is calculated as the initial value of the fitness function, and the minimum value of the fitness function in the global range is the optimal MLS-SVM soft-sensing model, in this case, the corresponding (γ, σ) is the optimal solution in the set search range