Counts provide the foundation for measuring nonmotorized travel along a link or a network and are also useful for monitoring trends, planning new infrastructure, and for conducting safety, health, and economic analyses. Most agencies still use manual counting methods, however over the last decade, several automated technologies have been developed to count bicyclists and pedestrians. Due to cost and other resource considerations, permanent counts are still limited to small subsets of networks. Although counts provide direct measure of activity at a location, they are not useful for understanding activity across the network. The lack of widely available pedestrian count data precludes safety studies and analysis of trends, which has become critically important especially with the nationwide increase in pedestrian crashes over the last decade. The emergence of crowdsourced data such as Strava and StreetLight has allowed for the collection of large-scale datasets over broad areas of the network. A recent research study evaluated how to derive pedestrian count estimates from these data sources using traditional modeling methods. This study seeks to apply machine learning methods for pedestrian volume estimation to improve the accuracy and resolution of initial predictions.