Study On Evaluation Theories And Methods Of Small Sample

People often encounter this problem that there are many positive samples and few negative samples in daily life and engineering practice, especially in high-tech fields such as defense, aerospace and nuclear facilities. Therefore, how to extract available information in limited sampling data to analysis and forecast the test process scientifically and effectively is what the theories of small sample need to study and resolve.Common evaluation methods are described and concluded in detail in this paper, including the narration of their sources and mathematical principles. Also the advantages, disadvantages and limitations are described in this paper. Especially the method of support vector machine which have recently formed and developed rapidly in recent years is generally introduced, including the related concepts and development of this method on solving small sample problems.Subsequently, support vector machine is specified from two aspects:classification and regression, including their mathematical principle and modeling mechanism. In the classification part, the method of support vector machine is used to the coal and gas outbursts prediction. Firstly, a three-classification prediction is made on typical support vector machine; the categories of the outburst are safety, small outburst and large outburst. Then the support vector machine is combined with the factor analysis, and the combined model is used to verify the above problem, then the conclusion that the classification accuracy is improved on some degree is drawn. In the regression part, the method of support vector machine is used to the prediction of coal mine safety. Firstly, the GM (1,1) model and typical support vector machine are used to predict the death toll and death rate of per million tons of coal mines in recent years, then the two methods are combined, and the predicting accuracy of the combined model is higher than the former two models.