EXPLAINABLE ARTIFICIAL INTELLIGENCE APPLIED TO PREDICTING CHILDHOOD OBESITY
Techniques based on Artificial Intelligence show promise in decision support in the health area, in which several studies show computational efficiency in processing large amounts of data and its effectiveness in contributing to decisions based on the information obtained. However, the lack of understanding of the models' predictive mechanisms has led to the search for new techniques based on Explainable Artificial Intelligence that allow elucidating questions about how predictions are made.In this context, this research sought to develop a method based on Explainable Artificial Intelligence demonstrating the relationship between input data and the predictive response. In the preliminary results, two methods of explainability were considered (Local Interpretable Model-agnostic Explanation and Shapley Additive Explanations) applied to predict childhood obesity, and it was concluded that waist circumference was a risk factor for the outcome, with evidence suggesting greater incorporation into clinical practice.