Prediction of Students Dropout Using Data Mining

Abstract

As a developing country, dropout is the dominant obstacle for our educational sectors. Therefore, it is important to develop the logical method for prediction of the students at the risk point of dropping out, allowing a proactive process to reduce this problem. This research work develops a prototype which can automatically recognize either the students will continue his/her study or dropout, using classification rules. Data were performed at one of the famous and prestigious university named Daffodil international university, with the main goal to reveal the high prospective of data mining applications. Data were collected from the students mainly focused on their personal and Family problems and university-performance. The responsible factors for dropping out were found through the Association technique using Apriori algorithm. Pre-processed factors were applied on the running students who were already completed one years or 3rd semester of their study. Classification method can be highly supportive in predicting student’s dropout reasons. Selected 10 best attributes using CFS which were directly affected on the analysis. Finally, decision was making based onC4.5, Naïve Bayes algorithms that a running student would continue their study or would drop out. C4.5 algorithms was found the best classifier with 86.014% accuracy whereas Naïve Bayes was 76.22% of accuracy.