<p>Stream classification plays an important role in the study and management of freshwater ecosystems. Many classification schemes exist that focus predominantly on physical habitat, hydrology, and thermal regimes, but few frameworks explicitly include river network connectivity. Because river network connectivity directly affects the movement of water, nutrients, sediments, and aquatic species throughout a watershed, including connectivity metrics in river classification provides opportunities for advancing riverine research at large spatial extents. We developed a robust framework and dataset that incorporates a network connectivity-based stream classification system for the conterminous United States (NetConUS), utilizing the National Hydrography Dataset Plus version 2 (NHD). Connectivity classes are differentiated by degree centrality, eigen-vector centrality, clustering co-efficient, closeness centrality and betweenness centrality and comprise five types: central streams, peripheral streams, mainstem streams, cluster stream and convergent streams. The streams classification was validated using Bayesian Neural Networks (BNNs) to account for uncertainty in the assignment of network connectivity classes. The dataset captures the structural roles of stream segments within river networks and provides opportunities to include network connectivity metrics alongside geophysical descriptors in analyses of freshwater fauna at the extent of the conterminous US.</p>