The Improvement Research Of The Support Vector Machine Pretreatment On SMO Algorithm

Block algorithm and decomposition algorithm are two main iteration algorithm of support vector machine,and sequential minimal optimization algorithm is a special decomposition algorithm that fixed the number of samples for a working set of two,so the result is the increase number of iteration. When the number of the original training samples is too much,sequential minimal optimization algorithm need to improve.To solve this problem,using the compression can improve the algorithm,and the sampling sequential minimal optimization algorithm is based on this kind of thought.But because the sampling is random and complex,how to effectively reduce sampling range is the main direction of improving this algorithm.All the support vector are on the boundary and the non support vector have on effect on the result, so through some criteria to find the boundary samples set that contains all of the support vector and reducing the sampling range to this set,can reduce the time of the sampling sequential minimal optimization algorithm to a great degree.At the same time the theoretical analysis and the experimental verification explains that the capability of the improved algorithm is better.