Implementation of Course Recommender System for Virtual University of Pakistan

TL;DR

This paper presents a web-based course recommender system for Virtual University of Pakistan that predicts student grades using collaborative filtering, aiding students in selecting courses aligned with their interests and competencies.

Contribution

It introduces a novel course recommender system tailored for Virtual University, utilizing user-based collaborative filtering and rating prediction methods.

Findings

System tested on 470 courses and 2600 students.

Expected marks depend on student's past performance and similar students' grades.

Mean Absolute Error indicates acceptable prediction accuracy.

Abstract

Universities working in Pakistan are offering a comprehensive set of degree programs for different levels. Virtual University of Pakistan is country's first institution completely based on modern information and communication technologies. It offers education in many different majors and various areas of study are available. Multiple courses are offered in each program that satisfy several general requirements of degree. Selection of courses that align with competency and interest can become an important factor in determining final score (CGPA) of student. For this purpose, a web-based course recommender system specifically designed for courses offered at Virtual University is developed. User-based collaborative filtering and rating-prediction approach is used for calculation of expected marks and grades. System is tested against 470 currently available courses and simulated data of 2600 students. Test results showed that expected marks are somehow dependent on student's average marks in already studied courses and average marks of similar students in target course. Accuracy of implemented system is measured using Mean Absolute Error for 100 observations. MAE value came out to be in acceptable range.