Learning Algorithms And Applications Of The SVM For Regression

Support Vector Machine (SVM) based on the statistical learning theory is a new machine learning method, which was developed by Vapnik and his team in 1995, the SVM has become the hotspot in the field of machine learning because of its excellent learning performance. Support Vector Machine for regression (SVR) has recently attracted growing research interest due to its obvious advantages such as nonlinear function approximation with arbitrary accuracy, and good generalization ability, unique and globally optimal solutions. But its theoretical system has much room for improvement and the realization of the learning algorithm has many problems for solution. How to design SVR learning algorithms which are both fast and valid has become the bottleneck in practical applications of SVR. Moreover, its research in applications needs to be enhanced.In this dissertation, some existing SVR learning algorithms are firstly introduced. By inference and via improved mathematical formulas, a new SVR learning algorithm—Lagrangian Support Vector Machine for Regression (LSVM-R) is presented. Then some applications are studied. The SVR and LSVM-R based methods are presented to determine serum cholesterol levels from the measurements of spectral content of a blood sample in medical science, which are compared with the BP-network-based method. In addition, the SVR-based method is presented to curvefit the experiment data of the interference and diffraction of light in physics experiments, which is compared with the Least-Square-Algorithm-based method. The experimental results show that theproposed methods are better than the other two methods in whole properties.