Predicting student dropouts using random forest

Abstract

Abstract Among the other problems in the learning process, student dropout is an acute problem that needs to be taken care of by the educationist and policymakers. This paper is based on 330 students admitted to the Jawahar Navodaya Vidyalaya (JNV) school in the 6th class in five successive batches. The dataset has ten attributes out of which eight variables are categorical, and two are numerical. The paper addresses the research question as to what factors are important vis-a-vis the dropout students. Further, we have applied a random forest classifier to predict the school dropouts after five years. The results show that performance in the 6th class, income, father’s education, and gender are factors that influence the school dropouts. The random forest classifier achieves 86 per cent accuracy, 41 percent sensitivity and 98 percent specificity. We need to take data from more schools to further generalize the study.