On Improving Students' Performance through Combined Statistics and Data Mining Analysis
On Improving Students' Performance through Combined Statistics and Data Mining Analysis

Summary/Abstract: The learning process is (or should be) periodically monitored and evaluated in order to improve its results and efficiency. Unfortunately, most of the times, only the students' results are taken into consideration when making various reports, statistics or recommendations. However, it is obvious that, besides the instant or a period based student's performance, there are other external factors that affect in an objective manner the overall performance. Traditional investigation methods are based almost exclusively on statistical instruments and are not capable to reveal all aspects that lead to more or less good results for the students. For this reason, the authors aim to make a combined analysis based both on statistics and Data Mining instruments. While the approach based solely on statistics methods offers undeniable numerical results, the Data Mining approach comes with possible explanations for the plain figures. Used together, these tools are capable to offer not only a relevant track of students' performance but can also provide recommendations for improving the results. Various elements are taken into consideration increasing the objectivity of the analysis and offering a new perspective that allows searching and identifying the existence of relations between some not so obvious parameters such as student's load, scholar timetable, sequence of courses. The study is based on traditional statistics methods and Data Mining techniques in an attempt to validate teaching methods, on one hand, and to identify the aspects that may be improved, on the other. The relevance of the study is high since more than 2000 records distributed over a period of 8 years were used. Independent evaluations from practical activities (laboratory, tests, individual projects) and final exams cross-validated the overall results. The statistical approach is represented by time series and histograms which reveal numerical results for the students' performance over the entire period. The results are presented in ways that represent the start of Data Mining analysis which was performed afterwards. The Data Mining section outlines the most important aspects, such as: data preprocessing, implementation of various models and methods, results analysis. The overall approach is capable to identify some causes for non-optimal performance of the students and also to make some recommendations for improving it.

• Details
• Contents