The model parameters and kernel parameter of SVM are random, which are not conducive to search the optimal parameters, thereby reducing the prediction accuracy and efficiency.
Block population initialized measure
PSO inertia factor is fixed, and the local and global search abilities are limited, which reduces obtaining the optimal solution ability.
Adaptive inertia factor
PSO algorithm is easy to fall into local minimum in the latter prediction part.
Nonevolution number of mutation strategies
When low-dimensional space transforms into high dimensional space and solves quadratic programming problems, if the number of input spaces is high, computational efficiency will be a new problem.