Predicting and analyzing the performance of students through data mining techniques to improve academic performance

Nowadays, significant advancements in information technology and communication field in different societies are seen. Given that these advancements, universities as a leading institution in the field of science, have moved towards electronic processes in the management of education and educational environments, there are databases with a large amount of information. By analyzing this massive data of educational systems, methods can be provided to improve the educational status of students. Educational data mining has sought to discover the knowledge contained in the data of the educational system. One of the applications of educational data mining is to predict students' academic performance. Predicting students' academic performance and providing useful solutions is of particular importance in the success of educational systems and can help managers make the right decisions to increase the efficiency of the educational system and better student performance. The purpose of this paper is to identify the effective indicators on academic performance, predict students' academic status using data mining techniques, and finally present a new trend for modifying unit selection and educational strategies to increase the efficiency of the education system.

steps of this research are determined according to CRISP model. In current research, Databases containing 9 datasets of specialized courses in industrial engineering were used. The students' grade was bachelor's degree. Indicators affecting student performance have been identified based on previous researches and expert opinions. Demographic data and academic records of undergraduate students are entered in database. After data preprocessing, 13 attributes are selected, different models were proposed to predict student's academic status in the next semester. Then, a comparison between the results of 4 different algorithms has been done.

All 13 attributes are identified to be effective according to information gain and gain ratio. This 13 attributes as follow: GPA, Total passed units, Number of conditional terms, Type of admission, Marital status, Gender, University admission year, Living place , Age, Current semester, Prerequisite course score, instructor of the course, Repeat the course. Between of 4 considered models, the Logit Boost algorithm is known as the best model in categorizing in two class and multi-class according to the accuracy rate and ROC.

Because of acceptable performance of data mining algorithms, the use of these algorithms in predicting student performance is appropriate and the proposed model can be used as a support tool for decision making in educational systems. Finally, according to the obtained results and the opinion of academic experts, the unit selection process was redesigned. The proposed model can be used as a decision support tool in educational systems. Finally, due to the results obtained and the opinions of the academic experts, the process of unit selection was redesigned. The presented process uses the available data in educational systems and data mining science, provides useful knowledge to decision-makers to make the right and appropriate decision. Decision makers can make appropriate decisions by examining the predictions made by the data mining algorithm and obtaining useful information, in order to make the educational system more efficient.