Support Vector Machine Based On Semi-definite Programming And Its Applications

Support Vector Machine (SVM) is a good classification&recognition method incomputational intelligence methods, it has a good ability to handle non-linear, highclassification&recognition accuracy and generalization capability, and is widely usedin many fields. However, there are some problems in the traditional SVM, such as theselection on kernel function and responding parameters, in order to improve theclassification&recognition accuracy, the novel SVM based on Semi-DefiniteProgramming (SDP) and its applications in oil well logging are mainly studied.The main work and innovations are as follows:(1) Study on Support Vector Machine based on Quantum Culture particle swarmoptimization (QCPSO-SVM). To optimize the relevant parameters in SVM andovercome the problem of PSO algorithm easily falls into local minima in SVMtraining, an improved method of the support vector machine based on QuantumCulture particle swarm optimization is studied, which is used to optimize parameters,where QCPSO uses quantum algorithm to simplify the iterative process, and usesculture algorithm to expand the search space, so as it can avoid falling into localminima. Typical simulation shows that the recognition accuracy of QCPSO-SVM isgreatly improved.(2) Study on Support Vector Machine based on Semi-Definite Programming(SDP-SVM). To overcome the problems of selecting the kernel function and itsparameters, as well as the evolutionary operation of evolutionary computation is morecomplicated, from the combination of multi-kernel functions, the SVM based on SDPis studied, which uses SDP to solve the combination of operating parameters inmulti-kernel functions, the simulation results show the SDP-SVM is better thancommon SVM at recognition accuracy.(3) Application study on practical logging with SDP-SVM. To solve theproblems in current recognition methods, such as the low recognition accuracy andthe poor generalization capability in complex oil and gas layer recognition, theSDP-SVM method is the first time to use in the actual oil layer recognition, theapplication results show that its recognition effect is very significant, the SDP-SVM issuperior to PSO-SVM and QCPSO-SVM at recognition accuracy, which has a broad application prospects.