A Data Mining Approach for Exploring Correlates of Self-Reported Comparative Physical Activity Levels of Urban Latinos.

We applied data mining techniques to a community-based behavioral dataset to build prediction models to gain insights about physical activity levels as the foundation for future interventions for urban Latinos. Our application of data mining strategies identified environment factors including having a convenient location for physical activity and psychological factors including depression as the strongest correlates of self-reported comparative physical activity among hundreds of variables. The data mining methods were useful to build prediction models to gain insights about perceptions of physical activity behavior as compared to peers.