This paper talks about how to plan efficient delivery route by analyzing drivers' behavior. Today, as E-commerce services represented by Amazon are growing, demand of delivery service is getting bigger. It is a problem that not only the amount of delivered baggage is expanding, but also the level of delivery service is rising. On the other hand, the number of drivers is reducing because of a low salary and poor working conditions. From these background, it is required to improve efficiency of delivery to keep delivery service. Especially in urban areas, the most time-consuming work in last mile delivery is not moving between customers by a truck but work occurring after parking (including parking). In spite of this, many researches focus on how to minimize the total time of moving between customers, like solving traveling salesman problem (TSP). So this paper has tried analyzing drivers' “parking behavior”, consisting of a successions of behavior such as searching for parking place, parking a truck, taking out baggage from a truck and visiting customers to entrances.On this paper, I have tried analyzing by following steps. First, I have detected each parking behavior from the GPS data of the trucks by 2-stage clustering method, which has resistance to noise. Next, I have matched each parking behavior with each customers. Then, I can analyze how drivers decide parking behavior by customers, and we can plan efficient delivery routes.