Font Size: a A A

Research On The Optimization Of Support Vector Machine Algorithm Based On Fisher Discriminant Analysis

Posted on:2017-07-09Degree:MasterType:Thesis
Country:ChinaCandidate:X LuoFull Text:PDF
GTID:2348330488466909Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Support vector machine is a supervised learning model, it is based on Statistical Learning Theory. So, it has a solid statistical foundation, it makes support vector machine to have a strong suitability in solving the classification problems in the field of the high dimension, small sample and nonlinear pattern recognition problem.Compared to the other machine learning algorithms, the support vector machine has a good robustness. Meanwhile, the characteristics of the algorithm is simple and feasible, which makes it applicable in the areas of machine learning and pattern recognition, and has a more broad space for development. Support vector machine also has certain disadvantages:when dealing with large-scale sample classification problem, the performance of support vector machine can greatly reduce. This thesis puts forward an optimizing support vector machine algorithm dealing with large samples.This thesis introduces the support vector machine and Fisher discrimination analysis at first, respectively from the linear separable and linear inseparable with the direction of their principle, then gets the correlation between support vector machine and Fisher discrimination analysis:there is a certain relationship between the optimal separating hyperplane of support vector machine and the optimal projection direction of Fisher identification. Then, this thesis proposes a Fisher based discrimination analysis to optimize support vector machine algorithm: by Fisher discrimination analysis, the key support vectors of the support vector machine problems can be quickly chosen, thus the training time in support vector machine sample can be reduced greatly.After putting forward the optimization algorithm, this thesis verifies the algorithm by simulation experiments of linear separable and linear inseparable cases. It indicates that without reducing the classification accuracy, the optimization algorithm can improve the performance of support vector machine.Finally, on the basis of above, by means of the optimization algorithm, the experimental data is analyzed again. The optimization algorithm is applied into Intrusion Detection Systems, results show the performance is good.
Keywords/Search Tags:support vector machine, Fisher discrimination analysis, key support vector, the optimal separating hyperplane, the optimal projection direction, linear separability
PDF Full Text Request
Related items