Claim Missing Document
Check
Articles

Found 11 Documents
Search

SISTEM REKOMENDASI SKRIPSI MENGGUNAKAN METODE CONTENT BASED FILTERING Toyibah, Zulfah Binti; Aini, Nuru
Infotech: Journal of Technology Information Vol 11, No 1 (2025): JUNI
Publisher : ISTEK WIDURI

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37365/jti.v11i1.377

Abstract

The management of the thesis repository in the Reading Room of the Faculty of Education, Trunojoyo University still uses Google Spreadsheets for the data collection process, so it is very limited in supporting the search and management of theses. As a result, students have to search for theses manually by looking one by one on the reading rack. In addition, laboratory assistants are often overwhelmed in the data collection process, which causes delays in updating thesis data on the reading rack. Therefore, a thesis recommendation system is developed using the content-based filtering method to facilitate automatic thesis search and thesis management by integrating the thesis collection feature independently. The method used is content based filtering using TF-IDF, SVD, and Cosine Similarity algorithms. This research uses the Research and Development (R&D) method with a waterfall model that has several stages, namely analysis, design, coding, and testing. Based on the results of the algorithm performance expert test, a value of 100% (very feasible) was obtained. System expert validation results show a value of 1, meaning that all features on the recommendation system can function properly. The results of web expert validation obtained a percentage of 96.92% (very feasible). In the user trial results, a percentage result of 91.3% (very feasible) was obtained. In testing the performance of the algorithm, a trial was conducted with 40 keyword samples using confusion matrix at threshold 3, threshold 5, threshold 10, and threshold 15. The recommendation system is said to be effective if the acquisition of a high precision value even though the acquisition value of recall is low. In this study, threshold 3 has the highest precision value compared to testing on other thresholds, namely accuracy 97%, precision 95%, recall 17%, and f1-score of 28%.