Abstract:
In recent years, the biggest challenges that educational institutions are facing the explosive growth of educational data and to use this data to improve the quality of managerial decisions. The number of students scoring low performance in Ethiopia Higher Education Entrance Examination (EHEEE) result is increasing from year to year as the reports of NEAEA and Ministry of Education. At the same time, many students who have low performance have joined higher institutions in order to full fill the intake capacity of universities. Moreover, many students have caused the families for more expenses. Even if there are a number of studies regarding these problems, the works were focused only on few selected attributes at preparatory and secondary schools. Those studies are also not understandable by all stakeholders and not easy to be guide by the result. Although there are studies regarding academic performance of students using data mining techniques, they are all about university students. Thus, this study aims to apply data mining techniques to develop a classifier model which predicts the Performance of Students in Ethiopian Higher Education Entrance Examination (EHEEE) Examination in order to help new students early before they face the problem and enables managerial in making different decisions to improve the students’ academic performance. A total of 3013 student’s records and 32 attributes were used to build the predictive model using J48, JRIP, REPTree and PART algorithms and the class label of the record are taken to be Excellent, Very Good, Good, Satisfactory and Fail based on the academic performance assessment system for Grade 12 National Exam. The researcher has developed classifier model by using Hybrid data mining model where PART algorithm which registered the highest accuracy of 95.37%. As the result, 157 rules have been found as the guide to perform better performance. In this case, the educational planers can identify the determinant attributes to give support at each grade level to full fill their gaps. Finally, the system evaluation has found 96% by using the system performance and 91.2% by user acceptance tests. Further study is needed to get the better performance on the real time.
