Support Vector Machine Data Classification

Support Vector Machine (SVM) is a new kind of machine learning technology that appeared in the middle of the 90's century. Different from the traditional neutral network technology(NN), which is based on traditional statistic learning theory, SVM is based on the statistic learning theory (SLT). Traditional statistic learning theory prerequisites enough volume of samples, while statistic learning theory, which provides theoretical frame for machine learning, focuses on the research of statistical regulation and learning methods under the condition of small samples. It has proved in practice that Support Vector Machine based on SLT not only is uncomplicated in structure, but also has evident improvement in technology especially in promotion capacity, so it can solve many small sample-learning problems in practice as a new kind of neutral network technology. At present, SVM has become hot spot in the area of machine learning internationally. The paper tries to discuss Support Vector Machine (SVM) in a deep and systematical way.The main contributions of the paper can be summarized as follows.First, the first chapter introduces the appearance of Support Vector Machine as a result of the development of traditional neural network and its defect in Machine Learning. Besides, this chapter also points out the necessity and importance of the research and application of Support Vector Machine, and the achievement and inadequacy of the research of Support Vector Machine. The second chapter mainly discusses the theoretical foundation of Support Vector Machine. The chapter introduces the definition and the commonly used algorithm of Support Vector Machine, which is followed by the numerical test of the algorithm. Meanwhile, the chapter also reflects the shortcomings that need improvement in this method, lying the base for further improvement of Support Vector Machine.Second, the main point of the paper, smooth technology of Support Vector Machine, is introduced in chapter 3. In view of non-smoothness of Support Vector Machine, a new smooth polynomial is proposed based on the research methods of preceding researchers. As a result, the paper puts forward SSSVM, and proves its advantages compared with other methods in concrete numerical test.Finally, concerning non-linear classification, many scholars propose other simple methods for Support Vector Machine besides the commonly used kernel function. However, this kind of research is still in exploration. To further improve the general utilization of Support Vector Machine and its capacity of promotion, application, recognition and so on, the forth chapter further explores Support Vector Machine according to the structural feature of non-linear data point. In application of sphere structural model in geometry and kernel function of Support Vector Machine, the chapter classifies the non-linear data, resulting in sphere structural Support Vector Machine, and proves its feasibility by algorithm.