Student performance prediction using modified chicken swarm optimization and improved conditional generative adversarial network -with parallel support vector machine over educational data

Abstract

A massive estimation of computerized data is been created for an extensive combination in the data mining area. The conception of student performance prediction models focused to classify the academic performance of student is the important factor of the growth of Education Data Mining (EDM). It is one of the important and predictable processes that are performed over the huge education background. Data mining methods are employed to build the prediction models for the educational dataset. But it has issue with huge amount of features from the specified dataset hence the prediction accuracy of student's performance is degraded significantly. To overcome the above mentioned problems, in this work, Improved Conditional Generative Adversarial Network -with Parallel Support Vector Machine (ICGAN-PSVM) algorithm is proposed. It is focused to predict students' performance under supportive learning through school and family tutoring which increases the prediction accuracy considerably. This research includes main steps are feature selection and prediction process. The feature selection process is done via Modified Chicken Swarm Optimization (MCSO) algorithm. It is utilized to choose the more appropriate features from the University dataset. The objective function of MCSO algorithm is used to improve the students' performance via selecting best fitness features. Then the prediction is performed by using ICGAN-PSVM algorithm which is used to classify the more accurate student's performance prediction results. From the experimental result, it concludes that the proposed ICGAN-PSVM algorithm gives higher sensitivity, specificity and Area under the Receiver Operating Characteristic Curve (AUC) rather than the existing algorithms