Parallel SVM Algorithm Based On Hadoop And Its Application In Mooc Platform

In the process of globalization of education,the Mu class was born.Mu class platform for its many learners,free,open,and many other features,is set off a revolution in education.However,due to many factors,the number of final exams through the Mu class,to obtain a certificate in a small proportion of the overall learner,Mu class platform to promote the development,improve learner rate is a priority to obtain the certificate.Therefore large data in the mass education,learners classify,process monitoring learners can obtain a certificate,and then to provide them with customized services efficiently is an effective way to improve the Mu class certificate acquisition rate.In the classification algorithm,support vector machines because less over-fitting,feature vector dimension disaster obvious advantages are widely used.The traditional SVM algorithm complexity is high because of the time,training for a long time and other factors,not suitable for large data sets.To address these deficiencies,the SVM algorithm parallelism to process large data sets,is an effective solution.The main contents include:(1)A brief description of the background of this study and practical significance,consult the relevant domestic and international research literature,and analyzed and summarized in this paper is to provide an improved important scientific reference and theoretical support Hadoop-based parallel SVM algorithm.(2)Support vector machine theory describes the system,including the definition of support vector machine and its mathematical model,and then introduced the SVM classification typical application in two cases,and finally use the binary PSO and genetic algorithm,SVM's kernel parameter and punishment factor optimization,improve classification accuracy.(3)On the theory of SVM stacked on top of Hadoop parallel algorithm based on SVM it was designed and implemented,and for the shortcomings of its existence,made two suggestions for improvement,and apply it to handle large scale data sets,greatly improve the speed of training SVM.(4)Summarize the work and research that exist shortcomings of this article,and parallel development support vector machine direction was the future.