Svm Classification Based On The Imbalanced Datasets For Problems Of Psychodiagnostics

This article discusses the aspects of application of the different psychodiagnostics tests in the educational sphere, in particular, in the schools and universities, to predict some events. Also, the problem of the results classification of the psychodiagnostics tests of individuals in the educational sphere has been considered. The application of the SVM classification based on the imbalanced datasets to this problem has been discussed. The data imbalance is inherent in the results of many tests, for example, intellectual tests. It is shown that the SMOTE (Synthetic Minority Oversampling Technique) and its modification famous as the bSMOTE (boundary SMOTE) algorithm can be used for the data rebalancing. It allows improving the classification results for the boundary objects. A herewith, the novel approach to the search of the parameters’ values of the bSMOTE algorithm has been analyzed. It allows minimizing the time expenditures for development of the best SVM classifier. It is shown, that the Python toolkids allow accelerating the process of the program development, that can be useful for the problem of the SVM classification based on the imbalanced datasets. The analysis of the experimental results confirms the efficiency of the SVM classification, when it is necessary to predict the assessment of individual on the base of the psychodiagnostics tests' results. © 2017 Published by Future Academy www.FutureAcademy.org.UK