Application of the XGBoost Algorithm in Detecting AI-Generated Assignments: A Novel Approach to Preserving Academic Integrity in Distance Education.

The emergence of novel AI-based language technologies has posed a serious challenge to verifying the authenticity of student assignments in distance education systems. This study proposes a practical solution based on the XGBoost algorithm capable of distinguishing AI-generated texts from human-written ones. The method employs ۲۴ features across two categories-stylometric indicators (lexical, syntactic, and punctuation-related) and behavioral indicators (temporal, revision-based, and interaction-related)-and was evaluated using the standard HC۳ dataset comprising ۱۰۰۰۰ samples. Validation was conducted through ۵-fold cross-validation, and the results demonstrate strong model performance with ۹۶.۱% accuracy, ۹۵.۸% precision, ۹۶.۴% recall, and ۹۶.۱% F۱-score. Feature importance analysis revealed that type-token ratio (TTR) and average sentence length are the most influential factors in the detection process. The proposed framework can serve as an effective tool for distance education instructors to monitor assignment authenticity and safeguard academic integrity in the age of artificial intelligence.

کلیدواژه ها:

Academic Discipline ، E-Learning ، XGBoost ، Text Stylometry ، Writing Behavior Tracking