The Research On Support Vector Machine Based On The Geometry Estimate Value Of Data Sample

With the development of various industries, Data Mining is applied in the industries such as bank, insurance which a lot of data in. But,This is just data, not information. It takes too much time to complete the task by artificially. This needs more effectively means and methods to gain useful information quickly from a lot of data. This is the main task of the Data Mining. The research and application background of the Data Mining are introduced in the paper at first. Because the main research of the paper is around Support Vector Machine, the basic concepts of Support Vector Machine and Classification and prediction methods are introduced. Those classic algorithms about Support Vector Machine and advantages and disadvantages of them are also analyzed in detail. In order to solve the problem of the Support Vector Machine that the problem with the constraints, the basic theories of the problem without constraints and classic algorithms of them are introduced. After that, the recent research situation of the Support Vector Machine of domestic and foreign countries is introduced and analyzed and also compared in Classification performance and the training time of data.After analyzing the Classification methods of the Support Vector Machine with the geometry character of data sample, a new Support Vector Machine Classification method based on geometry estimate value of data sample is prospected in the paper. Estimated value of the data sample is defined by the geometry character of the data samples. It reflects the geometry character of data sample much better and effectively. It is calculated in a simple way. Then the new objective function and constrained function of the new Support Vector Machine are formed. The new Support Vector Machine based on the geometry estimate value of data sample is formed in linear and nonlinear conditions. It's solved by multiplier method. The multiplier method about the new Support Vector Machine also derived in mathematics on theory. Then the experiment is executed to compare the new Classification method of the Support Vector Machine based on geometry estimate value of data sample to the conventional Classification methods of the Support Vector Machine. The experiment shows that the new Classification method of the Support Vector Machine based on geometry estimate value of data sample has better Classification performance and shorter training time.Specially, if there are a lot of data samples, the training time of the Support Vector Machine is long. In order to shorten the training time, Assistant Classification Strategy of the Support Vector Machine based on the geometry estimate value of data sample is proposed. The unitary bias degree is defined with the geometry character of data sample. The purpose is that utilizing the geometry estimate value of data sample to assistant the Classification of the New Support Vector Machine, not propose the assistant Classification method lonely. It means that the geometry character of data sample is properly used. Analogously, the experiment is carried out on the new Assistant Classification Strategy method in classification performance, training time, classification precision. It shows that the new Classification method of Support Vector Machine based on the data estimated value with the Assistant Classification Strategy has better Classification performance. These methods not only improve the performance of the Classification on theory, but also promote the practical application.