The Application of Problems Concerning the Imbalanced Data Classification by Means of Support Vector Machines

Support Vector Machine (Support Vector Machine, SVM) demonstrates many unique advantages in solving the small sample, nonlinear and high dimensional pattern recognition, and can promote to the application of the use of the function fitting, and other machine learning problems. In order to solve the problem concerning the imbalanced data classification, researchers at home and abroad put forward various solutions based on support vector machines for unbalanced data classification. Many scholars have proposed categories by constructing a small number of samples to compensate for the gap with larger classes to balance the effect, but the new sample is difficult to ensure the same distribution with the original sample, and increases the burden of training devices, or by reducing the number of samples to achieve balance, although this will speed up the training speed, but will reduce the sample information, and the overall error rate of two samples were not reduced. From the perspective of statistical distribution of the sample, the thesis analyzes the reasons resulting in unsatisfactory classification of unbalanced data by support vector machines, and summarizes the uneven performance evaluation index and the progress of unbalanced data classification by support vector machines.