MINING STUDENT'S ADMISSION DATA AND PREDICTING STUDENT'S PERFORMANCE USING DECISION TREES

The purpose of data mining is to find out new and possibly useful information from huge amounts of data. Data mining techniques are useful in many application areas like fraud detection, businesses, banking and telecommunications. Educational Data Mining is the application of Data Mining Techniques to educational data. Quality Assurance in education has compelled academia to constantly explore ways to improve overall educational processes. This has led to increasing interest in educational data mining. This paper is a first attempt to retrieve pedagogical information from the data of a public sector engineering university in Pakistan. The data mining techniques are used on the educational data of the undergraduate students in order to predict the performance of students. The study is planned around three research question: Can the students' college marks be used to predict their performance at the undergraduate education? Is the discipline in which they are enrolled significant in predicting their performance? Is any one particular year out of their four years undergraduate studies more decisive than the rest in predicting their performance? To answer these questions, data mining algorithms were applied to identify patterns in the available historical data. The students' marks at college level were examined and mined using the k-means clustering algorithm. The findings revealed a strong correlation in the students' college marks and their marks in individual subjects particularly in Maths, Physics and Chemistry at college level, however, no significant correlation was found between the students' college marks and their overall performance in the undergraduate programme. So the first question is answered negatively which is in agreement with results of different studies conducted in other countries. This analysis suggests that students' performance at university level might be based on the learning and teaching methods of university. The result of clustering pointed out that discipline should be taken into account to predict performance. The application of different decision trees, as classification algorithms, to the examination marks of students from different years of their current degree programme in order to predict their academic achievement in their final year examination indicates that performance in first and second year has a considerably decisive impact in predicting students' final year performance. The study carries important implication for the academic institutions by helping them in providing assistance to students to improve their academic skills at the appropriate level and time.