EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Mentor Recommendation System Using Content-Based Filtering with TF-IDF and Cosine Similarity

Regan Putra Ramadhan, Retno Mumpuni, Muhammad Muharrom Al Haromainy

bit-Tech · 10.08.2026 · 9 (1) · 587-598

DOI: 10.32877/bt.v9i1.3945 ↗

Kurz gefasst

Fachartikel in der Zeitschrift bit-Tech, online erschienen am 10.08.2026. Verfasst von Regan Putra Ramadhan und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Conventional alumni platforms typically rely on static directories, which often lead to information overload and high mismatch rates in mentor selection. Furthermore, the prevalent use of bilingual technical terms in user profiles often distorts semantic matching. To address these critical issues, this research presents the design and preliminary implementation of a career mentor recommendation system that dynamically matches student competencies with alumni track records. The contribution of this work lies in the integration of a custom bilingual preprocessing pipeline with a Content-Based Filtering (CBF) engine, specifically tailored for Indonesian–English alumni ecosystems under cold-start conditions. Methodologically, the system integrates a custom bilingual text preprocessing pipeline featuring stemming and lemmatization to normalize mixed professional terms. Feature extraction is executed through the Term Frequency-Inverse Document Frequency (TF-IDF) algorithm, while profile alignment is computed using the Cosine Similarity metric. Developed iteratively using the Agile methodology, the matchmaking engine was evaluated on a beta-testing corpus consisting of 30 enriched alumni profiles and 10 student test cases. Using actual student mentor requests to determine relevance labels, the experimental results yielded a Mean Average Precision (MAP) score of 0.825 (82.50%) for the Top-5 mentor recommendations. Furthermore, an ablation study demonstrated that the custom bilingual lemmatizer improved the system's Recall@5 from 80.00% to 100.00%, providing solid empirical evidence of its efficacy in mitigating semantic fragmentation. The practical implication is a foundation for a more effective professional collaboration ecosystem within the alumni platform.

Zitieren

APA

Ramadhan, R. P., Mumpuni, R., & Haromainy, M. M. A. (2026). Mentor Recommendation System Using Content-Based Filtering with TF-IDF and Cosine Similarity. bit-Tech, 9(1), 587-598. https://doi.org/10.32877/bt.v9i1.3945

Vancouver

Ramadhan RP, Mumpuni R, Haromainy MMA. Mentor Recommendation System Using Content-Based Filtering with TF-IDF and Cosine Similarity. bit-Tech. 2026;9(1):587-598. doi:10.32877/bt.v9i1.3945

BibTeX
@article{Ramadhan2026,
  author = {Ramadhan, Regan Putra and Mumpuni, Retno and Haromainy, Muhammad Muharrom Al},
  title = {Mentor Recommendation System Using Content-Based Filtering with TF-IDF and Cosine Similarity},
  journal = {bit-Tech},
  year = {2026},
  volume = {9},
  number = {1},
  pages = {587--598},
  doi = {10.32877/bt.v9i1.3945}
}
Kostenlos starten →