MACHINE LEARNING FRAMEWORK FOR ASSESSING ADOLESCENT CONCER TOWARD UNHEATHY FOOD ADVERTISEMENT

Abstract

For the purpose of raising health awareness and directing public policy, it is essential to forecast teenage concern over marketing for harmful foods. XGBoost, a gradient boosting machine learning model, is used in this work to forecast teenagers' degrees of concern based on behavioural and demographic characteristics. Age, parental education, and types of commercial exposure, such as free toys and celebrity endorsements, were among the survey data gathered from 1030 teenagers. To deal with unbalanced classes, the model is trained via synthetic oversampling and hyperparameter adjustment. By interpreting feature importance using explainable AI approaches (LIME and SHAP), it is possible to determine which elements have the greatest impact on teenage concern. The findings show that XGBoost achieves great prediction accuracy, providing a practical and understandable way to comprehend and lessen the effects of marketing for unhealthy foods.