With the spread of electronic commerce, parcel delivery services are expanding. However, delivery businesses are suffering from a chronic shortage of human resources, making it urgent to improve delivery efficiency. Currently, the selection of delivery routes is often left to drivers. Inexperienced drivers rely on car navigation systems to determine their delivery routes, but these systems are for ordinary drivers, primarily assisting with single journeys and reducing driving time. They do not, for example, provide information on safe and suitable places to stop for unloading. On the other hand, a skilled driver will choose an efficient route based on his experience without relying on car navigation. In this paper, we use particle filter optimization to create a learned model replicating the delivery routes of skilled drivers and apply this model to delivery route planning and identification. The model replicated 74% of the delivery route of a skilled driver in the area studied, achieving a 7.0% improvement in accuracy rate over the shortest-path route and, in the non-learned area, achieved 65% replication and 4.4% improvement.