Third-Party Library Recommendations Through Robust Similarity Measures

Abhinav Jamwal, Sandeep Kumar

2025

Abstract

This research systematically investigates the impact of different similarity measurements on third-party library (TPL) recommendation systems. By assessing the metrics of average precision (MP), average recall (MR), average F1 score (MF), average reciprocal rank (MRR) and average precision (MAP) at different levels of sparsity, the research demonstrates the significant impact of similarity measurements on recommendation performance. Jaccard similarity consistently outperformed the measurements tested and performed better in low-order and high-order app-library interactions. Its ability to reduce the number of sparse data sets and achieve a balance between precision and recall makes it the optimal measurement for the TPL recommendation. Other measurements, such as Manhattan, Minkowski, Cosine, and Dice, exhibited limitations to a certain extent, most importantly under sparse conditions. This research provides a practical understanding of the strengths and weaknesses of similarity measurements, which provides a basis for optimizing the TPL recommendation system in practice.

Paper Citation

in Harvard Style

Jamwal A. and Kumar S. (2025). Third-Party Library Recommendations Through Robust Similarity Measures. In Proceedings of the 20th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE; ISBN 978-989-758-742-9, SciTePress, pages 635-642. DOI: 10.5220/0013366600003928

in Bibtex Style

@conference{enase25,
author={Abhinav Jamwal and Sandeep Kumar},
title={Third-Party Library Recommendations Through Robust Similarity Measures},
booktitle={Proceedings of the 20th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE},
year={2025},
pages={635-642},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013366600003928},
isbn={978-989-758-742-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 20th International Conference on Evaluation of Novel Approaches to Software Engineering - Volume 1: ENASE
TI - Third-Party Library Recommendations Through Robust Similarity Measures
SN - 978-989-758-742-9
AU - Jamwal A.
AU - Kumar S.
PY - 2025
SP - 635
EP - 642
DO - 10.5220/0013366600003928
PB - SciTePress