Principle Component Analysis And Recursive Feature Elimination Based Support Vector Machine Classification Methods Research

With the advancement of computer technology, it has become an inevitable trend that big data will be based on intelligent algorithms. Support vector machine is a widely used intelligent algorithm because of its decent generalization ability, small sample decision-making ability and the ability of nonlinear data processing. Statistical learning theory is good at processing small-sample problems and support vector machine is a learning machine based on the statistical learning theory. In recent years, support vector machine developed very quickly and generated many algorithms based on it, such as LS-SVM and LIB-SVM. These new algorithms make its range of application more wide, such as time series modelling, feedforward control, optimal control and so on. Moreover, some intelligent algorithms is defective, for example neural network intelligent control algorithm may be non-convergence and local optimization generates no unique solution and the structure is hard to select. Therefore, support vector machine has decent generalization ability and strong ability to process non-linear modeling problems. It is always applied to deal with large timedelay and non-linear systems. However, support vector machine is not perfect now and it could become more perfect in the future. Biomedical data is non-linear and very complicated and it is necessary to select appropriate intelligent algorithms to classify this data and apply this result to assist doctors analyze patients’ conditions. Since biomedical data has high dimensions, principle component analysis and recursive feature elimination are combined with support vector machine to classify it.Firstly, fundamental theory and modeling methods of support vector machine, principle component analysis and recursive feature elimination are introduced. To process the high-dimensional biomedical data, principle component analysis are employed to preprocess it and decrease its dimensions so as to improve the classification accuracy of support vector machine. Moreover, recursive feature elimination is another dimension reducing method, which is employed to preprocess the data and it is as efficient as principle component analysis. Both of them can extract the most important variables from high-dimensional data, which can improve support vector machine’s computational complexity, computational time and classification accuracy. Lastly, support vector machine, principle component analysis and recursive feature elimination are applied to classify the biomedical data and their simulation results are compared. From the decent simulation results, we can draw a conclusion that support vector machine is an efficient classification algorithm of biomedical data.