Using synthetic data and machine learning to predict heart rate and enable lifestyle recommendations

Background: Heart rate (HR) is an essential indicator for cardiovascular (CV) health and is known to be influenced by unhealthy lifestyle habits like smoking, alcohol and caffeinated drinks consumption. However, to date the potential degree of impact of multiple lifestyle factors on HR is unknown. The main goal in this study is to train a machine learning (ML) model to capture future HR over a 10-day period and subsequent generation of lifestyle recommendations based on predicted HR. Methods: The proposed system consists of generating synthetic data, building HR random forest ML predictive model, and evaluating the performance of the model versus real participants’ data. Results: We applied the system to 25 male participants. The study results validate the system’s ability to demonstrate effectiveness in predicting HR variations in response to lifestyle factors, including smoking, alcohol, and caffeinated drinks during different times of the day. Accurate predictions were observed for baseline HR readings, compared to the poorest performance for the prediction model at noon as reflected in the mean absolute error closer to zero and high R2 score of 0.953, capturing around 95.3% of the variance in the actual data. Conclusions: The study presents the foundation for further studies to investigate the applicability of other ML techniques for predicting HR using synthetic data generation. Additionally, the model can be further explored to improve its accuracy by including additional behavioural factors that influence the HR variability.