Large Scale Classification Based on Combination of Parallel SVM and Interpolative MDS

(1 Key Laboratory for Computer Network of Shandong Province, Shandong Computer Science Center, Jinan, Shandong, 250014, China 2 School of Informatics and Computing, Pervasive Technology Institute, Indiana University Bloomington, Bloomington, Indiana, 47408, USA) [email protected], [email protected] Abstract: With the development of information technology, the scale of electronic data becomes larger and larger. Data deluge occurs in many kinds of application fields. How to explore the useful information from the large scale dataset is a very important issue. Data mining is just to take on the task. Support Vector Machines (SVM) is a powerful classification and regression tools of data mining. It has been widely studied by many scholars and applied in many kinds of practical fields. But its compute and storage requirements increase rapidly with the number of training vectors, putting many problems of practical interest out of their reach. For applying SVM to large scale data mining, parallel SVM are studied and some parallel SVM methods are proposed. Currently parallel SVM methods are all based on classical MPI model. It is not easy to be used in practical, especial to large scale data-intensive data mining problems. MapReduce is an efficient distribution computing model to process large scale data mining problems. In this paper, parallel SVM based on iterative MapReduce model Twister is studied. Feature extraction is an efficient means to decrease SVM’s computing cost. Some feature extraction methods have been proposed, such as PCA, SOM network, and Multidimensional Scaling (MDS) and so on. But PCA can only measure the linear correlation between variables. The computation cost of SOM network is very expensive. In this paper, MDS is used to reduce the dimension of sample features, and interpolation MDS is used to improve computation speed. Parallel SVM combines with MDS to analyze large scale classification problems. The efficiency of the method is illustrated through analyzing practical problems.