Towards a Multi Classifier Machine learning Based Approach for Course Cancelation Problem Avoidance

Abstract

The Course section cancelation problem (MCP) is one of the main problems of the university timetabling process as a significant number of course sections are canceled every semester throughout the academic year.  The problem usually occurs when the academic registry at the university must cancel a course section for violating the "minimum number" hard constraint which refers to the minimum number of students enrolled in the course (the threshold). This problem is common within the universities in which the timetable is created prior to the enrollment process.

This paper discusses the development of a multi-classifier machine learning-based approach for course section cancellation risk estimation. The approach analyzes the enrollment historical data of the university to identify the common features of the canceled sections. Such features include the course, section-time slot, the number of students who are eligible to take it, and the lecturer. These features are then associated with section cancellation status. The resulted data set is fed into a multi-classifier component to predict the risk level of the section cancellation. The proposed approach aims to assist the academic departments in preparing the timetable of the upcoming academic term to avoid including the courses or section with the high risk in the timetable which in turn is expected to minimize the number of the canceled section.

Results have shown that the proposed approach has achieved a classifying accuracy of 85% in identifying the cancelation risk level of sections before including them in the timetable. The classifying accuracy is expected to improve with the growth of the data volume. In addition, using different gives the approach the dynamicity to use the most accurate classier to achieve the highest accuracy based on the provided case.