A decision support system for evaluating international student applications

In today’s transnational admission environment, evaluating applicant qualifications is becoming increasingly challenging. While standardized tests can be helpful, studies have shown that they are rather noisy predictors of performance. Predicting educational outcome is a viable alternative in such heterogeneous environments. Performance prediction models can be built by applying data mining techniques to enrollment data. In this paper we present an approach to using Bayesian networks to predict graduating cumulative Grade Point Average based on applicant background at the time of admission. While such prediction models can be helpful, their recommendations may not be followed by departmental faculty members making admission decisions if they are presented as black boxes. We thus present a novel approach to deriving a case-based retrieval mechanism from the Bayesian network prediction model in such a way that the similarity measure used by the case-based system is consistent with the predictive model. The case-based component retrieves the past student most similar to the applicant being evaluated. The Bayesian network model is evaluated using stratified ten-fold cross validation.