A Novel Nonlinear Function Fitting Model Based on FOA and GRNN
Table 1
Parameters setting of algorithms.
Model Type
Training and Testing samples
Algorithm parameters
FOA-GRNN
Training samples: 8000 dataset Testing samples: 2000 dataset
Individual: the SPREAD parameter; Individual fitness function: Sum of absolute errors; Evolution generations:20; Population size:20;
GA-BP
Same
Individual: the weights and thresholds of BP network; Individual fitness function: Same; Evolution generations: Same; Population size: Same; Crossover probability: 0.2; Mutation probability: 0.1;
PSO-BP
Same
Individual: the weights and thresholds of BP network; Individual fitness function: Same; Evolution generations: Same; Population size: Same; Particle maximum:0.55; Particle minimum:0.05; Velocity maximum:1; Velocity minimum: -1; Acceleration constants c1& c2: 1.49445;
