Predictive Analytics in Higher Education: Model Choices A comparison of six student-completion models finds that careful sample and predictor design can matter more than choosing a complex algorithm. Similar accuracy scores also conceal meaningful differences in who gets classified as high risk. Read More →