International Journal of Scientific Research and Engineering Development

Plagiarism in research papers has become hard to detect due to the rise in the use of online materials and the development of artificial intelligence-assisted writing tools. Traditional plagiarism detection techniques through text matching tend to perform poorly, especially if the plagiarized text is paraphrased, partially plagiarized, or rewritten in different styles. This paper proposes a plagiarism detection methodology focusing on stylometric analysis. Instead of relying on text matching, the methodology utilizes machine learning classification based on feature extraction. The methodology analyzes the research paper from word, syntax, and structure perspectives. to grasp the distinct writing patterns of the research paper author and pinpoint irregular patterns indicative of plagiarized or foreign text. Utilized features include word frequency, Part of Speech, punctuation, and sentence structure, which are measured and normalized as vectors. These vectors help train traditional classifiers such as K-Nearest Neighbors, Naive Bayes, Decision Trees, ensemble models like XGBoost, and stacking classifiers. These trained classifiers have the ability to distinguish between author style and suspicious pieces of content with high precision. The tool provides both document-level and section-level features and is capable of identifying fully plagiarized documents, as well as hybrid documents that contain only some sections of suspicious content. This tool assists supervisors, reviewers, and institutions in maintaining academic integrity because of the in-depth style insights that lie beyond mere similarity scanning.

Dr.M.Ayyappa Chakravarthi, K.L.Roshini, K.S.R.S.Vaishnavi, Ch.Aparna, E.SaiVarshika,"Stylometric Feature Extraction and Machine Learning Classification for Research Paper Plagiarism Detection" International Journal of Scientific Research and Engineering Development, V9(2): Page(965-970) Mar-Apr 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.