Research On Parallel SVM Algorithm Based On Spark

Support Vector Machine(SVM)algorithm is widely used in various fields because of its good classification effect and simple practicality,but support vector machine is calculated by two-time programming.The solution of two-order programming will calculate the n-th matrix,and when the amount of data processed is large,the calculation and storage of n-order matrix will make the optimization speed become very slow,even cause memory overflow and cannot continue operation.In this paper,the improved Support vector machine(SVM)algorithm based on spark large data platform can solve the above problems,but it is not competent to face the problem of multiple classification.In this paper,a parallel one-to-many SVM algorithm based on Spark is implemented by constructing multiple classifiers,using spark framework and combining the classification characteristics of support vector machines.And through the UCI data set to carry on the contrast experiment,using the spark large data computing platform to improve the performance of a One-to-many support vector machine under large data is significantly better than a single support vector machine in the stand-alone environment,and this algorithm has a good speedup.