The Research And Application Of Least Square Support Vector Machine

Artificial Intelligence is the goal for people to chase. We hope that machine canhelp for hunman beings. Pattern recognition is an important part of ArtificialIntelligence. Pattern recognition can be seen everywhere in our daily life. This papergives an account of the concept and significance of Pattern recognition and tellssomething about the procedure of Pattern recognition. There are five steps of anexperiment of Pattern recognition. Such as the acquisition of data, pretreatment of data,select of characteristics, classification decision and systematic review. Like datanormalization is used in the process of data preprocessing, principal component analysisis used in the select of characteristics, t test is used in the classification decision andclassification accuracy is used in systematic review.Support vector machine is an important algorithm of classification decision in patternrecognition. This algorithm is well known for the public since it has been put forwardby Vanpik. SVM is extremely usefully in classification and regression. This papermainly tells about the content of SVM algorithm, uses the KKT condition in thisalgorithm step by step and proposes specific steps in the solution of linear and nonlinearclassifier. Although SVM has a lot of advantages, it still has some disadvantages. Sosome researchers improve this algorithm in some aspects.This paper gives a new algorithm about SVM, it is called least square supportvector machine. It has been put forward in the late20th. And it attracts a lot ofresearchers to study. LSSVM can make up the shortage of SVM in the accuracy ofclassification and fitting. In this paper, LSSVM use RBF as kernel function. Soparameter is an important variable for the algorithm. And there are four methods to findthe optimal parameters. This paper gives a new algorithm called three steps of search onPSO. Finally, it analyzes and compares the property of each algorithm in the numericalexperiment.