Research Of Support Vector Machines In Quality Management

Support Vector Machines(SVM) is a new data mining method. As it is strictly based on the foundation of statistical learning theory and introduces the kernel technology, it considered to be one with the the most superior performance among machine learning methods at present.Traditional machine learning methods depent on the principle of minimizing the risk, so often fall into local optimum, but support vector machines use the structural risk minimization (SRM) theory in statistical learning theory and show better performance. Meanwhile, due to the introduction of nuclear technology,it enhances processing capacity for high-dimensional data,and also solves the curse of dimensionality.Currently, the reseaches of support vector machines mainly focuse on algorithm research, kernel function, and kernel parameters,and with two objectives -- that is, to minimize time-consuming and improve application performance.Our work is to improve support vector machines and apply it in our project about the quality management.The core work includes parameters optimization and modeling of comb-kernel support vector machines and applied research in quality management.My work divided into the following several aspects:about data preprocessing, in order to meet the needs of classification applications, improve the original kernel principal component analysis and use the method of maximizing class information to reduce dimensions; about kernel parameters, proposed an approach that used kernel-mapping space to optimize parameters, and establish the corresponding discriminant to do analysis and comparison;about comb-kernel support vector machine, the paper studied the modeling methods, and presented an approach to model step by step; about applications of quality management,for the features of data missing and different reliability ,and so on, proposed a comprehensive approach of applications.