In this paper, we propose a map matching algorithm for car navigation systems that predict user destination. This car navigation system is a novel system that automatically predicts user purpose and destination to present various information based on predicted purpose without user interaction. It requires the correct road where the car drives in real time, and it also need to know the route from the start point to the current point correctly. The proposal map matching method divides the trajectory into equal intervals and calculates the shortest path score for each one. Testing using GPS data for actual car trips showed that its use results in better destination prediction than with conventional methods in most cases. The results were the best for intervals of 5 minutes.