Speed-density relationships are a classic way of modeling stationary traffic relationships. Besides offering valuable insight into traffic stream flows, such relationships are widely used in dynamic traffic assignment (DTA) systems. In this research, an alternative paradigm for traffic dynamics models, appropriate for traffic simulation models and based on machine-learning approaches such as k-means clustering, k-nearest-neighborhood classification, and locally weighted regression is proposed. Although these models may not provide as much insight into traffic flow theory as speed-density relationships do, they allow for easy incorporation of additional information to speed estimation and hence may be more appropriate for use in DTA models, especially simulation-based models. This paper (with data from a network in Irvine, California) demonstrates that such machine-learning methods can considerably improve the accuracy of speed estimation.