Combination of rough set and cosine similarity approaches in student graduation prediction

Abstrak

Higher education institutions must deliver high-quality education that produces graduates who are knowledgeable, skilled, creative, and competitive. In this system, students are a vital asset, and their timely graduation rate is an important factor to consider. In the department of computer science, a challenge arises in distinguishing between students who graduate on time and those who do not. With a low on-time graduation rate of just 1.90% out of 158 graduates, this issue could negatively affect the institution's accreditation evaluation. This research employs the Case-Based Reasoning method, enhanced with an indexing process using rough sets and a prediction process utilizing cosine similarity. The testing, conducted using k-fold validation with 60%, 70%, and 80% of the data, produced average accuracy rates of 64.2%, 66.3%, and 65.6%, respectively. The test results indicate that the highest average accuracy of 66.3% was achieved with 70% of the cases.