The use of geographically referenced point data, such as that obtained from global positioning systems (GPS), is rapidly increasing. However, due to error and uncertainty inherent in most geographic datasets, the ability to accurately associate these point locations with other layers of geographic data is still a challenge. One difficulty in particular is how to associate spatially and temporally referenced point-based observations of a network activity with a network topology such that a continuous network path can be best inferred. In this article, an optimization method for inferring a network path from a temporal sequence of point observations of location is presented. An application to GPS data is provided to highlight various characteristics of the proposed modeling approach relative to several other available techniques.