Learning an Interpretable Behavioral Intervention Policy Using MHealth Data

• Activity Number: 254 - Contributed Poster Presentations: Section on Statistical Learning and Data Science
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Xinyu Hu* and Min Qian and Ying Kuen Ken Cheung
• Companies: Columbia University and Columbia University and Columbia University
• Keywords: Mobile health; Behavior intervention; threshold Q-learning

Abstract

In the area of mobile health (mHealth), massive data are collected to monitor users' health condition and provide insights for behavior intervention. However, ongoing support is rarely provided to help users to understand how their behavior contributes to changes in their health status. To tackle this problem, we aim to develop an interpretable policy for physical activity recommendation over a given time horizon. Specifically, we studied the effect of physical activity on psychological stress level. We formulate this problem as a sequential decision-making problem and solve it using a new method that we refer to as threshold Q-learning (TQL). The interpretability of TQL is achieved by making model assumptions and incorporating threshold selection into the learning process. Our simulation results indicate TQL is comparable with the traditional Q-learning and outperforms it under model misspecification. This work serves as a first step toward a computational health coaching solution for mobile device users.