Intelligent Group Optimization Algorithm PSO And Its Application In Several Types Models Optimization

The optimization technology is a application technology based on mathematics, which can be used to solve the optimal solution of various problems. Among them, There is an important branch - intelligent optimization algorithm, The intelligent optimization algorithm developed through the simulation or interpretation of some natural phenomena or processes, like the general search algorithm is also an iterative algorithm. The intelligent optimization algorithm have those advantages: the overall situation, the efficient parallel optimal performance, robustness, high universality, etc.The uncertainty problem which can be usually found in the actual situation, The research of uncertainty problem has been a difficult problems. The people makes a lot of ways to deal with the traditional uncertainty problems, which constituted an important branch of science—the research and forecast of uncertainties problem. The number of samples of these issues are usually very limited, and even were very rare, and most of the data sequence does not contain a clear characteristics relationship between those data. That makes these problems more difficult to deal with, there are many methods which have achieved good results, but those methods can also improve.The optimization algorithm will be applied, use optimization algorithms to optimize the model to improve model accuracy.The main contributions of this paper are listed as follows:(1) The intelligent optimization algorithm PSO will be applied to the grey prediction model GM(1, 1) and the improved new grey prediction model FGM(1, 1). First of all, because the grey model's prediction ability is effected by the mean sequence. The change of the mean sequence parameters a will affected the prediction ability of the model, In the past the choice of the parametersαgenerally used of the default parameters or based on actual data and the models's forecast effect, with a lot of randomness, and there is no a definite rule; Secondly, the grey prediction model GM(1, 1) at least need four data while the new and improved grey prediction model FGM(1, 1) at least need three data to build the model, The less of data is needed in building the model, the less data information can be learned, so this will definitely affect the model prediction. In light of the above,The intelligent optimization algorithm PSO will be applied in the grey model GM(1, 1) and FGM(1, 1) to optimize the parametersαof the mean sequence of the model. Through the establishment of the appropriate fitness function, then use the intelligent optimization algorithm PSO to optimize and search the appropriate parametersα, finally, to apply the searched parametersαin the the model to improve model prediction ability.(2) The intelligent optimization algorithm PSO will be applied to support vector machine model, because the correct classification rate and the effect of regression of the support vector machine model affected by support vector machine penalty factor, the kernel function and the kernel function parameters. The different penalty factor and the kernel function parameters have an great effect to the model prediction ability. There is no a definite rule in the choice of the penalty factor and the kernel function parameters, it is often based on experience. The selected parameters are often not very suitable to different types of data. In view of this, The intelligent optimization algorithm PSO will be applied to support vector machine model. According to different data to select appropriate penalty factor and the kernel function parameters, through the establishment of an appropriate fitness function, and apply the intelligent optimization algorithm PSO to optimize the support vector machine penalty factor and the kernel function parameters. Finally, to apply the searched penalty factor and the kernel function parameters in the the support vector machine model to improve model prediction ability.(3) Test of optimization results of the intelligent optimization algorithm PSO: Empirical test, The intelligent optimization algorithm PSO optimized grey prediction model GM (1,1) and intelligent optimization algorithm PSO optimized the new and improved grey prediction model FGM (1,1) have markedly improved prediction ability. The intelligent optimization algorithm PSO optimized SVM multi-classification model also have an markedly improved in the rates of classification accuracy. Through empirical testing we known that the intelligent optimization algorithm PSO have an good optimization results.