| The traditional discriminant analysis is one statistical method for discriminating the sample respective type,widespread application to be on a par with regression anal-ysis.Commonly used method have distance-distinguish method,fisher discriminant ,bayes discriminant ,stepwise discrimination for it.The rapid advancement in scientific and technological development,about artificial intelligence classified question (pattern recognition, discriminant analysis) research,it had the massive sorting algorithm to be published. For example,genetic algorithm,text classification algorithm,bayesian classification algo-rithm,SVM(support vector machine)for classification,fingerprint classification algorithm and so on. Because the classified question is the complex extension problem of function to explain theoretically,it does not have one kind of most superior sorting algorithm to be suitable for each kind of different situation.We have the necessity to its unceasing research.In order to solve the problem about choice of kernel functions in support vector machines for classification, in this paper we present a new approach—the combination kernel functions method for classification based on particle swarm optimization.This method overcame the limitations of the single nuclear function which describes inaccuracy discriminant result.In this paper,important aim at discriminant method for the thing having many indexs which have the harassment function to judge the result,we proposed based on the principal components analysis and the support vector machines combination discriminant analysis method for this kind of thing,and this article gives its sequence of operation.Through this method,both can simplify the operation,and overcome the flaw of linear methods,thus enhancement judgment rate of accuracy.Finally the application in the sand storm early warning's research,the experiment proved that the combination discriminant analysis method is effective.
|
PDF Full Text Request