Exploring Machine Learning Methods to Estimate Effects of Aging and Physical Activity Levels in the Dog Aging Project

This study follows a One Health approach in examining the relationship between physical activity measures and aging through analysis of data from the Dog Aging Project with companion animal dogs as the subject model. Physical activity outcomes include lifestyle, intensity, and average daily total active time. In addition to linear regression and multinomial logistic regression, machine learning methods including random forest and extreme gradient boosting are implemented. This study found that the machine learning methods can provide slightly higher predictive abilities with key explanatory variables of dog age, environment type, dog weight, and owner age being highly ranked in importance.