Route Feature Analysis And Recommendation Model For Truckers Seeking Goods Online

Today,my country’s online freight trading platforms is still unable to achieve large-scale matching of vehicle-to-cargo.The current online transaction mode is mainly based on truck drivers actively seek goods on the platform to obtain orders.The source route is the decisive factor for truck drivers to search for goods.However,the efficiency of drivers’ active search for goods is relatively low,and the existing recommendation algorithm are inefficient.Therefore,based on the driver’s online behavior data,this paper analyzes the drivers’ route preference characteristics,mines potential order-taking routes,and recommends supply routes for drivers based on the driver’s area,so as to increase the efficiency of drivers looking for orders online and improve the order response rate.In order to study the preferred routes of drivers,this paper has done the following four main tasks.The first step is to compare the driver’s online search behaviors and transaction behaviors,and determine the unity of the driver’s search source route and the order delivery route.The second step is to analyze the spatiotemporal characteristics of drivers’ search routes based on the drivers’ online search data.The results show that the search time,urban geographical features,and changes in the driver’s location all have a significant impact on the driver’s route selection.The third step is to design a potentially favored route mining algorithm based on the urban geographic features of the drivers looking for routes,combined with the OD hotspot and the Voronoi algorithm,and demonstrate its effectiveness by comparing it with the user-based collaborative filtering model(User-based CF).User-based CF considering latent routes effectively improves recall by 3% and precision by 16%.The fourth step is to build a dynamic route recommendation model based on time demand and driver location,and the model’s performance is verified by using the online goods-seeking behavior data of 646 drivers on the domestic online vehicle-to-cargo matching platform within one month.Compared with the existing methods,the recall rate increased by 37.66% on average,the precision rate increased by 120.07% on average,and the F value increased by 77.82%on average.At the same time,the recommendation effects at different times and different behavior modes are significantly improved and the differences are small,compared with the existing rule judgment methods,the interpretability and stability are slightly improved.The dynamic recommendation model based on user needs and location changes to recommend source routes with higher transaction rate for drivers,which can support the vehicle-to-cargo order matching logic layer of the comprehensive vehicle-to-cargo matching system in the future,and make up for the platform’s lack of search mode to efficient vehicle-to-cargo matching.Using the recommendation model to solve the problem of information overload and recommend source orders for trucks will help improve the efficiency of domestic vehicle-to-cargo matching,increase the user stickiness of the online freight trading platform,and effectively improve the platform transaction rate and user retention rate.