People's eating habits are influenced by psychological, social, cultural, and behavioral factors. Research shows that certain personality types expose people to risky eating behaviors. Given the complexity of nutrition-related factors and the limitations of traditional statistical methods, the use of new approaches such as artificial intelligence and machine learning can play an effective role in analyzing multidimensional data and identifying complex patterns. This cross-sectional pilot study aimed to predict food addiction among university students by integrating demographic, anthropometric and personality data with machine learning methods. The data consisted of 210 samples, which were first preprocessed to ensure data quality and integrity. Tomek Links and SMOTE techniques were used to remove class imbalance. Feature selection was performed using the twelve different algorithms to identify the most important features related to food addiction prediction. Then, ten different machine learning models were implemented, including Logistic Regression (LR), K-Nearest Neighbors (KNN), Gaussian Naive Bayes (GNB), Support Vector Classifier (SVC) with probability estimation, Decision Tree (DT), Random Forest (RF), AdaBoost, Gradient Boosting Classifier (GBC), CatBoost and LightGBM. The models were trained on the training dataset and their performance was evaluated using the accuracy, precision, recall, F1-Score and AUC metrics on the test dataset. In addition, the SHAP (SHapley Additive exPlanations) method was used to analyze the importance of features and interpret the advanced models to determine the impact of each psychological and behavioral feature on the prediction of food addiction. The results showed that more advanced models, especially ensemble methods such as Random Forest and CatBoost, have high power in identifying complex patterns and accurately predicting food addiction behaviors. SHAP analysis also showed that psychological characteristics such as feelings of worthlessness, impulsivity, anger, psychological distress, rigid cognitive styles, weight and height, body mass index (BMI) were related the most important factors affecting prediction. Although limitations such as small sample size, focusing on a specific student population, and the use of self-report instruments reduce the generalizability of the results, the innovation of this study in combining psychological and artificial intelligence approaches for early identification of high-risk individuals is remarkable. Overall, the integration of personality profiles with advanced computational models can form the basis for the development of artificial intelligence-based screening tools and targeted interventions to improve nutritional behaviors in young populations.
Rahimnezhad et al. (Fri,) studied this question.