On predicting school dropouts in Egypt: A machine learning approach

Abstract Compulsory school-dropout is a serious problem affecting not only the education systems, but also developmental progress of any country as whole. Identifying risk dropping out, and characterizing its main determinants, could help decision-makers to draw eradicating policies for this persisting reducing social economic negativities over time. Based on substantially imbalanced Egyptian survey dataset, paper aims develop Logistic classifier capable early predicting students at-risk out. Training with an usually weaken performance especially when it comes false negative classification. Due fact, extensive comparative analysis conducted investigate variety resampling techniques. More specifically, based eight under-sampling techniques four over-sampling ones, their mutually exclusive mixed pairs, forty-five experiments dataset are build best possible classifier. The contribution provide explicit predictive model school dropouts in Egypt which be employed identifying vulnerable who continuously feeding chronic problem. key factors vulnerability suggested identified student diseases, co-educational, parents' illiteracy, educational performance, teacher caring. These matching those found by many research previously similar countries. Accordingly, authorities confidently monitor these tailor suitable actions intervention.