Estimating shortest paths on large networks is a crucial problem for dynamic route guidance systems. The present paper proposes a statistical approach for approximating fastest paths on urban networks. The network data for statistical analysis is generated using a macroscopic traffic flow based simulation software. The input to the software are the input flows and the arc loads or the number of cars in each arc and the outputs from the software are the various paths joining the origins and the destinations of the network. The network data obtained from the simulation software is subjected to hybrid clustering followed by canonical correlation analysis. The hybrid clustering comprises of two methods namely k-means and ward's hierarchical agglomerative clustering. The results of the data analysis are decision rules containing arc loads and input flows that govern the fastest paths on the network. These rules are used for predicting the paths to follow while arriving at the entrances of the network. Before entering the network, the arc loads and input flows provided by the rules are checked inside the network. If agreement is found, then the path obtained from the data analysis is the fastest path otherwise the shortest path is chosen as the fastest path.