Creating Synthetic Test Data by Generative Adversarial Networks (GANs) for Mobile Health (mHealth) Applications

Mobile health (mHealth) applications have experienced rapid growth, driven by the demand for health monitoring solutions and smartphone adoption. However, evaluating these apps poses challenges due to limited and diverse user data. This study explores the use of Generative Adversarial Networks (GANs) to generate synthetic test data for mHealth applications. The paper introduces the methodology involved in training GANs using real user data obtained from Google Fitbit and showcases the creation of synthetic data mirroring real user profiles and parameters. Statistical comparisons between real and synthetic datasets validate the alignment and similarities in key attributes such as age, BMI, and exercise duration. The paper elucidates the importance of user-centered design methodologies and the role of test data in mHealth app evaluation. User personas and diverse user scenarios are incorporated, showcasing the efficacy of synthetic data in mitigating data limitations. The study emphasizes the potential of synthetic test data to enhance the evaluation and validation of mHealth applications, providing a pathway to address data scarcity challenges. Future research avenues are outlined, including expanding user diversity, refining GAN models, and assessing the impact of synthetic data on machine learning models within mHealth apps. The study advocates for ethical considerations and privacy safeguards in synthetic data generation and usage, suggesting frameworks for responsible implementation. This research contributes to advancing mHealth application testing methodologies by leveraging GANs to create diverse and reliable synthetic test data.