DATA MINING APPROACH FORPREDICTING HIGHER EDUCATION STUDENT'S PERFORMANCE AND DROPOUT CHARACTERISTIC

Abstract

Decision trees are employed to make important1design decisions, and they1are utilized in this study to explain interdependence between dropout student characteristics and student outcomes. Using tree analysis, logistic1regression is utilized to determine likelihood of outcomes & dropout features based1on a range of parameters. In this article, data mining methodologies are studied to discover how accurately they predict school achievement and dropout rates. Several algorithms are discussed, including decision tree classification, random forest, OneR, logistic regression, & the naïve bayes1classifier. Finally, consequences of developing classification1models using clustered populations1are examined. Models for predicting student performance and dropouts are developed using clustered population and classification approaches. When classification models are evaluated on a validation set and a randomly selected cluster1population, a sense of their generalizability emerges.