Support Vector Machine And Its Applications On Signal Processing

Support vector machine (SVM) is a new technology on data mining. It is a new tool toslove machine learning problem by using optimization method, which has much excellencebeyond compare. And it offers a noval idea to solve the nonlinear problem. Statistical learningtheory (SLT) as SVM's foundation is studied in the paper. SVM is used to model in nonlinearsystem of oil field. A novel adaptive filter based on SVM is proposed in the paper.The main wok includes:(1) VC dimension theory and structure risk minimization principle are discussed in detailwhich is the foundmental theory of SVM.(2) Some popular SVMs and their realization of process are also introduced briefly. Thesolution of optimization educed from SVM is discussed. Especially, sequential minimaloptimization (SMO) method is widely used in majority softwares of SVM.(3) A new way of data analysis of record well in oil field is proposed based on SVC.Firstly the recognition model of oil/gas/water zone is established. Then we can use it to predictthe law of oil and gas distribution in some oil district. The result of prediction coincides withactual testing results. The interpretation of oil/gas/water distribution provides a real timemethod for record well data, which advances the recognition rate of oil and gas.(4) A novel method on fault diagnoses of oil pump based on SVC is proposed in the paper.Compared with other methods, it cannot recognize the states of oil pump but has great ability toclassify and the best generalization.(5) A new approach based on support vector machine to predict the production in Oil Fieldis put forward in the paper, which has superior accuracy in small samples. It also has otherwisequalities. The model is simple, and has better generalization.(6) A new modeling method of well test pressure data based on SVR is used to fit thepressure curve automatically, rapidly, which offers a feasible method to reduce the time of thecomprehensive interpretation on well testing. It has economic value and utility.(7) A new adaptive filter based on SVR is proposed. The simulation experiment given hasproved the validity of the new algorithm. The time-variable noise is removed.The research above is the foundation for SVM used in oil field system which has sometheory and practical value.