Boosting a kNN classifier by improving feature extraction for authorship identification of source code
(2020) The 12th meeting of the Forum for Information Retrieval Evaluation (FIRE 2020) — Location: Hyderabad (En ligne) (16.December.2020)
Files
LASTCamera-readyAI-SOCO_NewRef.pdf
Open Access - Adobe PDF
- 861.47 KB
Details
- Authors
- Abstract
- This paper presents the system developed by the LAST to identify the author of a source code. It combines the classical 1-nearest neighbor algorithm with a feature extraction step whose main characteristics are the use of indentation-aware tokenization to gather n-grams and skip-grams, which are weighted by Relevance Frequency. Its performance is over 92% correct identification of the author among 1,000 potential programmers.
- Affiliations
Citations
Bestgen, Y. (2020). Boosting a kNN classifier by improving feature extraction for authorship identification of source code. In Parth Mehta, Thomas Mandl, Prasenjit Majumder, Mandar Mitra (ed.), Proceedings of The 12th meeting of the Forum for Information Retrieval Evaluation - FIRE 2020 (pp. 705-712). CEUR Workshop Proceedings. https://hdl.handle.net/2078.5/220757