Research On Fusion Technology Based On SVM And MOEA And Its Application

In the field of scientific research and engineering design, there are a large number of multi-objective nonlinear problems. Modeling and optimization of such problems are difficult issues. Support vector machine (SVM) as a specialized case study on small samples of machine learning theory has the better promoting generalization capability than the traditional statistical learning theory and neural network, and the modeling of nonlinear problems can be resolved satisfactorily. Evolutionary algorithm is a stochastic optimization method of simulating the natural evolutionary process, and also is the overall situation is a probability of optimization method. And a large number pareto optimal solutions can be obtained by multi-objective evolutionary algorithm (MOEA) at one time, and the solutions are kept consistency well. So it is very suitable of multi-objective evolutionary algorithm for solving multi-objective optimization problems.Based on the analysis and research on the principle of SVM and previous works,improvement of SVM is proposed. That is, to avoid difficult choice of parameters of SVM, a method of using differential evolutionary algorithm for parameter choice is presented. Numerical experiments show that the parameter choice is solved by improved SVM better .And also on the basis of the analysis and research on principle of MOEA, through combination of the two advanced algorithms non-dominated sorting genetic algorithm (NSGAII) and differential evolutionary for multi-objective optimization (DEMO), added a polynomial mutation operator, a new improvement multi-objective evolutionary algorithm (INSDE) is put forward. It is proved that INSDE is superior to NSGAII by multi-objective function tests problems, and also INSDE can converge to the pareto front of test problems effectively and has a good distribution of uniform. By combining SVM and MOEA, a set of calculation framework on modeling and optimization of non-linear multi-objective problems is put forward and function tests show that has good performance.In response to the proposed calculation framework, it is applied to an example modeling and optimization of product quality of an iron and steel company, and the issue of its modeling and optimization is resolved remarkably. Finally, with the actual situation of enterprise production, a production proposal is presented. Thus, a new approach to improve product quality is provided.