Hyperparameter Optimization of Naive Bayes for Supervisor Recommendation in Computer Science

Sinaga, Muhammad Nabil
Kurniawan R, Rakhmat

TIN: TERAPAN INFORMATIKA NUSANTARA
Vol 6 No 5 (2025): October 2025

The increasing number of students in the Department of Computer Science at UIN Sumatera Utara has made the process of selecting thesis supervisors more complex and time-consuming. This study aims to develop a system that automatically recommends the most suitable supervisor based on the similarity between thesis titles and lecturers’ areas of expertise. The proposed model applies text preprocessing techniques such as case folding, tokenization, stopword removal, and keyword extraction to transform thesis titles into meaningful features. These features are then classified using the Naive Bayes algorithm to predict the probability of each lecturer being the most relevant supervisor. The dataset consists of 794 thesis titles and 25 lecturers collected from 2019–2024. The model was evaluated using an 80:20 data split, achieving an accuracy of 87.3% with stable precision and recall scores, demonstrating reliable performance in supervisor recommendations. This enhanced Naive Bayes model can assist academic departments in ensuring a fairer and more efficient supervisor assignment process.