Route planning plays an important role in intelligent transportation systems (ITS). Its objective is to find the optimal route for users with consideration to the efficiency and safety of the route as well as theusers' preferences. The existing methods for finding this optimal route are mainly based on computing the shortest geographic route between the source and destination locations. However, traveling only along the shortest path route may be sub-optimal in terms of travel time because the route ignores prior users' experience and other environmental factors, such as road capacity and historical traffic patterns. In this paper, we present a smart driving route recommendation system based Geolife GPS trajectory dataset from Microsoft Research, generated by 182 users in a period of over five years under various weather conditions. The approach consists of three steps. First, each trajectory is segmented into small routes according to stationary points (places where users spends a significant amount of time) and intersections (with other trajectories) on their current route. Second, we extract the features of each route from all of the trajectory data, such as the average and standard deviation of the speeds, and the confidence of the GPS trajectory data matching the real map using on-line map services, e.g. OpenStreetMap(OSM) and Google Map. These features are utilized to model the probability of this route being a good route using Bayesian theory. Correlation between these features and the historical weather data is studied as additional factor of the road conditions. Lastly, an improved Johnson's algorithm is employed to calculate the optimal driving route to the destination. In our method, the edge weight relies on not only the distance between two locations, butalso the route evaluation and the estimated impact of the weather. Results show that the proposedmethod has better performance compared to the traditional path-planning methods.