Fast Algorithms Of Multi-class Classification Problem

Support vector machine (SVM) as one of the latest and most extensive data miningmethod, has been widely applied to the areas of machine learning, such as patternrecognition, image processing, etc., the study is endless. The traditional support vectormachine mainly aimed at two kinds of classification problem, in most of the time, weencountered multi-class classification problem. How to solve the problem of large-scalemultiple categories of data has become one of the focus of the field of machine learningresearch at present.Firstly on the development of machine learning stage to do a simple introduction, andresearch status of multi-class classification algorithm to do a brief elaboration, analyzessome typical methods to introduce. Then based on the limitations of multi-classclassification algorithm put forward two new methods applied to class classificationproblem, the method can effectively avoid solving the quadratic programming problem,shorten the operation time. At last, through the empirical analysis to illustrate the newmethod is effective and desirable, through empirical evidence, in the case of classifyingthe same set of data, the speed and accuracy of this method are high.Firstly, the article introduces the research significance and research status at homeand abroad, then introduces the two types of classification problem and multi-classclassification problem and its related algorithm. Secondly, it elaborates the minimumhyper-sphere of class center and the definition of a fast classification algorithm based onminimum super class center sphere, and then introduces the smallest super class center infeature space sphere and the definition of class center in feature space minimum sphere ofquick sort algorithm. Finally, it contains multi-class classification problem fast algorithmand its empirical analysis and summary, the quick sort algorithm and combine multi-classclassification algorithm, get fast algorithm for multi-class classification problems, usingthe fast algorithm for multi-class classification problems and existing algorithms onseveral data sets in some machine learning database class classification, and comparingeach method, and draw the conclusion as follows: when the fast algorithm of radial basiskernel function used in multi-level classification problem, its running time and theclassification results are better than existing algorithms.