An Explainable AI Framework with Interactive Visualization for Predicting Adolescent Concern over Unhealthy Food Advertisements
Journal of Sensors, IoT & Health Sciences · 30.06.2026 · 4 (2) · 18-30
DOI: 10.69996/jsihs.2026008 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Sensors, IoT & Health Sciences, online erschienen am 30.06.2026. Verfasst von Kotte Vinaykumar und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Advertising of unhealthy foods has a considerable effect on the eating habits of teenagers and children. Fast food advertising is used to persuade teenagers to choose high-calorie and low-nutrition food (e.g., burgers, candy). The techniques used by advertisers to promote high-calorie foods are often persuasive, such as using celebrity endorsements and giving away toys with the purchase of a Happy Meal. These examples of persuasive advertising dominate the public's perception of the products advertised and the information given to consumers about the advertised product. This paper aims to create a comprehensive explainable AI predictive framework to quantify the level of concern (high, medium, low) of adolescents who are exposed to unhealthy food advertising. This study utilized a dataset of 1030 adolescents (ages 13-18) collected through a survey and utilized a stacked ensemble model with logistic regression as the final phase learner, producing an accuracy of 95.63 percent and a weighted F1 of 0.957. The researchers utilized synthetic minority oversampling technique (SMOTE) to resolve the issue of class imbalance in the data set. In addition to predicting which adolescent was most likely to be concerned with a piece of unhealthy food advertising, the researchers used SHAP (Shapley Additive Explanations) to provide both global and local explainability of the model used to make the prediction. In both cases, celebrity endorsement, free toys, parental education, knowledge of health, and age were the most important variables in predicting concern level among adolescents who were exposed to unhealthy food advertising. An interactive web-based dashboard in Flask was developed to enable users to filter real-time demographic data to view the predictive output per adolescent, view the distribution of concern levels, assess the impact of an advertisement on the individual, assess the impact of an advertisement on an adolescent based on age and gender, view correlation heatmaps, and generate individual SHAP explanations. This work replicates and validates the predictive methodology documented by Titu et al. (2023), but this work provides an interactive visualization layer that can be used to assist in educating the public on healthy eating.
Zitieren
APA
Vinaykumar, K., Burra, S. K. R., Yerra, N. T., Kodam, V., & Chandragiri, J. (2026). An Explainable AI Framework with Interactive Visualization for Predicting Adolescent Concern over Unhealthy Food Advertisements. Journal of Sensors, IoT & Health Sciences, 4(2), 18-30. https://doi.org/10.69996/jsihs.2026008
Vancouver
Vinaykumar K, Burra SKR, Yerra NT, Kodam V, Chandragiri J. An Explainable AI Framework with Interactive Visualization for Predicting Adolescent Concern over Unhealthy Food Advertisements. Journal of Sensors, IoT & Health Sciences. 2026;4(2):18-30. doi:10.69996/jsihs.2026008
BibTeX
@article{Vinaykumar2026,
author = {Vinaykumar, Kotte and Burra, Shiva Krishna Reddy and Yerra, Naga Tharun and Kodam, Varshitha and Chandragiri, Jayanth},
title = {An Explainable AI Framework with Interactive Visualization for Predicting Adolescent Concern over Unhealthy Food Advertisements},
journal = {Journal of Sensors, IoT & Health Sciences},
year = {2026},
volume = {4},
number = {2},
pages = {18--30},
doi = {10.69996/jsihs.2026008}
}
EUVIMED