The parameter adjusts the ratio between configuration risk and experience risk, avoiding the overlearning problems and improving the prediction model generalization ability.
Less software reliability sample set
The slack variable is introduced to reduce the error sensitivity of prediction model. The model transforms the original problem into a dual problem, making solving process into a convex quadratic programming problem, getting the optimal solution easily.
Complex prediction process
The kernel function is introduced, which makes multidimensional input space into high dimensional space in order to solve multidimensional space problems.
More software reliability characteristic parameters
The prediction process is operated in the high-dimensional space, which makes original non-linear prediction problem transform into a linear problem, and then the results are reduced to the nonlinear problem solution.
Nonlinear characteristics of software reliability prediction
PSO is used to search the optimal solution of parameters to improve the overall prediction accuracy.
It is difficult to get the optimal solution for the software reliability prediction model parameters.