Enhancing Adaptive Interventions with Generative and Neural Networks

• Makayla Pham, Computer Science, University of Delaware

• Keith Decker, Computer & Information Sciences, University of Delaware

This research explores the application of artificial neural networks (ANN) and generative adversarial networks (GANs) to generate synthetic data, aimed at enhancing trends observed in a Just-In-Time Adaptive Intervention (JITAI) study. By using models that curate simulated data based on an existing experimental dataset, the primary objective is to better learn when to generate adaptive interventions to increase physical activity. While previous research has proposed the use of JITAI to positively influence physical behavior, there is little real-world JITAI data, and use of synthetic data to supplement these interventions is underexplored. Neural networks, designed to mimic the human brain’s interconnected neuron structure, were created to recognize patterns and trends within the data. These networks consist of multiple layers, including an input layer, hidden layers, and an output layer, each composed of artificial neurons that process and transmit information. Through training, the ANNs identify relationships within the data, improving the predictive accuracy of the models. Generative adversarial networks consist of two competing neural networks: a generator and a discriminator. The generator creates synthetic data, while the discriminator evaluates its authenticity. This adversarial process continues until the generator produces highly realistic simulation data. By leveraging GANs, the study generated simulated datasets. Analysis from these programs show that timely adaptive interventions correlate with increased physical activity. Notably, interventions are more effective depending on the time of day, current location, and current state of activity. These findings suggest that strategically timed interventions, informed by real and synthetic data trends, can enhance JITAI’s effectiveness. This research underscores the importance of personalized, timely nudges in promoting physical activity, indicating that optimized intervention strategies can lead to reduced sedentary behavior. This study contributes to the field by demonstrating the valuable role of synthetic data in refining and enhancing adaptive intervention strategies.