Methods Of Optimal Cost And Kernal Parameters Selection In Support Vector Machine

With the development of computer and informational technology, more and moreprice need to be paid for collecting, storing and processing vast data. The data miningtechnology comes from the large amount of data which are produced by database tech-nology. We use the data management system to store the data, use the machine studymethods to analysis and extract the useful knowledge of the data.Consequently, how toextract useful information from sets of data becomes a problem need to be solved immi-nently. Data mining technology comes into being in this background. Data mining is anon-trivial process searching for useful, potential and understandable form from sets ofdata. It involves a lot of intercross subjects and technologies such as machine learning,mathematical programming, statistics, pattern recognition and so on.Support vector machine is a new technology of data mining, and one of the impor-tant results of applying mathematical programming to data mining. It is also a machinelearning method that contains some technologys of the machine learning domain. It con-tains the largest marging hyperplane,Mercer kernal,convex programming,slackvaribles and such techologies. Untill now it has got the best properties in some chal-lengable applyings. Support vector machine and the kernal methods are treated as "themost popular and sucessful example of the machine learning domain, and a developmentdirection which attracts so many people's concerns " in the American magzines.At first a method for selecting optimal cost and kernalparameters C andγin sup-port vector machine (SVM) is presented, The cost and kernal parameters are obtainedby solving a mathematical program with equilibrium constraints (MPEC) via combiningthe genetic algorithm and deterministic algorithm, but C andγare treated as variablesof optimization problem in this paper. The genetic algorithm is used to solve the opti-mization problem with respect to C andγ, and deterministic algorithm is used to obtainthe constraints of problem.And then the results show that the generalization performance of the support vectormachine can be improved clearly by the proposed method in this paper compared withthe grid search method through the numerical experiments.