Akademik Veri Yönetim Sistemi
JOURNAL OF INTERPERSONAL VIOLENCE, cilt.0, sa.0, ss.1-5, 2026 (SSCI, Scopus)
This study aimed to identify key risk factors and predict digital intimate partner violence (DIPV) exposure and perpetration among university students using machine learning (ML) algorithms. A cross-sectional online survey was conducted with 1,764 university students (age range = 18–41 years, M = 20.8; 87.2% female, 12.8% male) selected through snowball sampling from a large public university in Türkiye. The survey included sociodemographic, lifestyle, and relationship variables, along with the Digital Intimate Partner Violence Scale. Six ML models were used: Logistic Regression (LR), XGBoost, Gradient Boosting (GB), Random Forest (RF), LightGBM, and Support Vector Machines (SVM). Model performance was evaluated using accuracy, precision, recall, F1 score, and receiver operating characteristics-area under the curve (ROC-AUC). XGBoost achieved the highest performance (AUC = 0.996), followed closely by RF and LightGBM (AUC = 0.995). LR and GB also performed well (AUC = 0.992), while SVM had slightly lower performance (AUC = 0.989). SHapley Additive exPlanations analysis revealed that domestic violence history, urban residence, father’s low education, short relationship duration, and frequent digital communication were risk factors. High income perception and non-smoking reduced DIPV risk. ML models, particularly XGBoost, effectively predict DIPV. Socioeconomic and psychosocial factors should be targeted in prevention efforts, alongside digital literacy and support services.