Application Research Of Support Vector Machine On Prediction Of The Protein Structure

Support Vector Machine is a kind of new data mining technology, based on the statistical theory. It can solve the complicated problems of machinery learning by the means of the most optimal methods with favorable theoretical support. The dissertation discuss the basic mathematical theory of the SVM. It starts from the problem of classification to analyse SVM algorithm theory of the linearly separable problem, approximation linearly separable problem and linearly inseparable problem, based on which, the algorithm of the SVM is deeply investigated.The main work is as follows:1.This paper researched the SMO Algorithm on SVM that used for large-scale training samples deeply, and analysised the advantages of the grid search method for the parameters choice problem that effects the performance of the SVM algorithm, and combined Wisconsin Breast Cancer Database, drawn a better method by which SVM can applies in large-scale samples and parameters choice.2.This paper puts forward the method of a slip window with evolutional information. Firstly, it chooses the data bank—RS126 and takes advantage of a method with slipping windows to vector the data, and at the same time, makes sure of SVM's input vectors, and makes them normalized. Secondly, it gets the optimal parameters on the problem with the help of the grid-search method, and also gets the classified accuracy 69.0397% of SVM on the prediction of the secondary protein structure. Finally, compared with the classified effects of other algorithms, it proves that SVM with a slipping window and the grid-search method can make good effects in predicting the secondary protein structure.