Research On Taxi Cruise Route Planning By Fusing Geographic Information

With the development of Mobile Internet and Big Data,taxi trajectory data is easier to be collected as a log data of taxis than before.It could be used in related works about route plan for taxi drivers.By mining taxi trajectory data,researchers can perform tasks such as pickup points extraction,pickup areas recommendation,and route plan for taxi.However,it is not enough that planning taxi route according to the trajectory data of the taxi singly,because it is difficult to accurately model the real situation.Therefore,we conduct related research on taxi route plan by integrating geographic information including POI information and road network information into taxi trajectory information additionally.The main works of this paper are as follows:(1)Fusing geographic information for pickup areas recommendation.The model reduces shortcomings caused by data sparsity by integrating the objective geographical environment information around the taxi driver into the process of user-grid matrix decomposition.At the same time,according to the real-time spatial location information of the taxi,different pickup areas are recommended to taxis in different places.The experiment proves that comparing with the common methods,the average absolute error between the recommended results and the truth is reduced by 23.4%,and the root mean square error is reduced by 19.8%.(2)Recommending taxi pickup areas based on experienced drivers.The model provides opinions for the common drivers to seeking passenger by referring to the behavior of the experienced drivers.First,the revenue of each driver is obtained by analyzing the trajectory data basing on the profit function.Then,the drivers with higher income are marked as an experienced taxi drivers,the behavior information of selection of driving direction of experienced drivers is regard as the training data.Then,based on the improved KNN classification algorithm,the driving direction selector generated,which is a tool of taxi driving direction decision.Finally,we recommend pickups according to the directions.Experiments show that the accuracy of the direction selection of the model is about 5% higher than the commonly used classification model,and mae is 3% lower than other methods.(3)Route plan for taxi based on road direction.The model should overcome two challenges.First,there are different profit between different sides of road.Considering the sides of road when analyzing road profit is very important.Second,in the process of searching path,Combining local revenue and the global information is important.Based on the distribution of all pickup points within the driver’s current search range on different side of roads,this paper provides global heuristic information for the driver to carry out heuristic route planning.Experiments show that compared with traditional path planning method,our method has an improvement of profit.Pickup areas recommendation recommends some pickup places with high profit for taxi drivers.Taxi drivers choose some areas to go according to the results.Before drivers start to go,Route planning finding high profit route,which can improve the profit of cruise.Totally,we take two steps to explore our work,which are pickup areas recommendation and route plan basing on road direction.