| Sumario: | Background: Food addiction, characterized by the compulsive consumption of highly palatable foods, poses significant health risks, particularly among university students. This study investigates the prevalence of food addiction among Bangladeshi university students and its associations with mental health (depression, anxiety, stress, and insomnia) and behavioral factors (smoking, drug, alcohol use, and pornography consumption). Machine learning (ML) models were applied to enhance predictive accuracy. Methods: A cross‐sectional survey was conducted among 1697 participants across two Bangladeshi universities. Food addiction was assessed using the Modified Yale Food Addiction Scale 2.0 (mYFAS 2.0). Associations were examined using logistic regression and subgroup analyses by gender. Six ML models—K‐nearest neighbors (KNN), support vector machine (SVM), random forest (RF), gradient boosting machine (GBM), XGBoost, and CatBoost—were employed to improve classification performance. Results: Overall, 13% of students met the criteria for food addiction, with higher prevalence among males (14.8%) than females (10.4%). In adjusted models, anxiety (AOR = 2.44, 95% CI: 1.43–4.16), stress (AOR = 1.74, 95% CI: 1.18–2.58), and pornography use (AOR = 1.74, 95% CI: 1.12–2.69) were significant predictors. Subgroup analyses showed that anxiety, stress, and pornography use were significant predictors only among males. Among ML models, KNN achieved the highest accuracy (85.3%), while RF demonstrated the best AUC‐ROC (0.697), confirming their utility in identifying at‐risk individuals. Conclusions: Food addiction affects a notable proportion of Bangladeshi university students and is strongly linked with anxiety, stress, and pornography use, particularly among males. Interventions should include cognitive‐behavioral therapy and stress management programs, digital hygiene education, and nutritional counseling tailored to student populations. ML‐based predictive models, such as RF and CatBoost, may be integrated into campus health systems to support early identification and personalized interventions.
|
|---|