Researches On Algorithm For Confidence Evaluation Of SVM

At present, the emergence of vast amounts of data, urgently need to be converted into useful information and knowledge.This promoted application of data mining, and this technology rapidly developed and improved.Data mining is a active research field involving disciplines such as databases, artificial intelligence and statistic. Among them, as an important aspect of data mining technology, classification technology has always been concerned about by researchers.In recent years, support vector machine is one hot technology of machine learning research fields, and also the most important classification technologies,a powerful tool to solve classification problems using optimization theory. It showed good performance in many practical applications.This paper focuses on the research of algorithm to estimate confidence measure and decision modification of support vector machine (SVM).Totally 4 algorithms using parameters to reflect the confidences of recognition results are presented in this paper, and experiment shows that one algorithm of them is the best algorithm. This algorithm computes the distance from testing sample to the optimal hyperplane of SVM, and the probability that the testing sample and its k nearest neighbors belong to the same class as the decision of Libsvm for the testing sample.The algorithm rejects the classification results of samples whose confidence measures are smaller than the threshold corresponding to a given rejection rate.Experiments show that the performance of the Libsvm classifier has been well improved using this algorithm.