Application Of Support Vector Machines In High-dimensional Data

With the advent of the information age, access to information technology and networktechnology have been developing rapidly, high dimensional data are quickly produce in scientificresearch, engineering application, social life, and other fields. These data have high dimension,the nonlinear characteristics, it is primary problem that how to reduce dimension or extractfeature in data modeling and data mining. This question becomes research focus in the field ofartificial intelligence and machine learning.Manifold learning aims to explore the inherent regularity of high dimensional datadistribution, reveal low dimensional structure or inherent law hidden in the high dimension spaceand reconstructs to reduce dimension or realize visualization according to learning the limiteddiscrete sample data samples and finding embedded inner low dimensional smooth manifold inhigh dimensional data space. This dimension reduction method is more to be able to reflect thenature of things than the traditional linear dimension reduction methods (such as PCA, etc.), andmore conducive to understand and deal with the data.Statistical learning theory is a new statistics theory system set up based on the limited samples,provides a strong theoretical foundation for people to systematically investigate the machinelearning problems with the small sample data. And support vector machine (SVM) is a new andeffective machine learning method based on the statistical learning theory. It makes the nonlinearspace problem into linear space during using the structural risk minimization principle and thekernel function thought, and solves problems of small sample, nonlinear, learning, localminimum point and so on in the many learning method, has strong generalization ability and alsogreatly reduces the complexity of the algorithm. The support vector machine shows goodextension, so is becoming a new research hot spot in the artificial intelligence and machinelearning field.According to the high dimensional data dimension high and nonlinear characteristics, thispaper puts forward a kind of manifold learning dimension reduction method based on automaticselection model parameters, applies to high dimension data and processes original data usinglinear and nonlinear dimension reduction method. Then these data are classified by support vectormachine. Finally, this paper gives the simulation experimental analysis for the colon cancer dataand face data. The results show that the classification accuracy is improved using the proposedapproach in this paper compared with the traditional methods.