Order Recommendation For Ride-hailing Based On The Influence Factors Of Drivers’ Willingness To Orders

In the online car-hailing platform,as the number of drivers and the number of orders continues to increase,the problem of information overload is becoming more and more serious.How to make effective order recommendations for drivers has become an urgent problem to be solved.Traditional order recommendation is mainly based on demand matching.The platform’s order dispatching algorithm is difficult to meet the individual needs of drivers,resulting in a decrease in driver satisfaction and a low order response rate.In order to solve the above problems,through the research on the recommendation method and the factors that influence the willingness of online car-hailing drivers to take orders,this paper proposes an order recommendation method in the online car-hailing dispatch scenario that considers the factors that affect the willingness of drivers to take orders,and proposes mitigation The method of order cold start.First,construct a driver model that considers the factors affecting the willingness to take orders.Based on the idea of collaborative filtering recommendation,analyze the characteristics of the driver’s order-taking behavior,and express the influencing factors of the driver’s willingness to take orders as the driver’s regional preference,expected profit and order-taking ability,and combine the driver’s historical order-taking behavior and related descriptions of the order The information measures various dimensions and builds a driver model.Then,on the basis of the driver model,K-nearest neighbors are determined through the calculation of similarity.Use cosine similarity to calculate the similarity of drivers in the dimensions of regional preference,expected return,and order-taking ability respectively,and use the fusion coefficient to calculate the comprehensive similarity of the driver,and then determine the driver’s K-nearest neighbor.Finally,a top-N order recommendation list is generated based on the interaction records of K neighbors.At the same time,for the cold start problem of new orders that cannot be solved by the collaborative filtering algorithm,this research uses the content-based recommendation algorithm idea to calculate the similarity between the order feature and the driver’s preference vector to complete the recommendation of the new order.Finally,experiments are conducted on the real data set provided by Didi Chuxing.The results show that it is reasonable to establish a driver model in this study considering the factors affecting the willingness of drivers to take orders,and the method of order recommendation for drivers based on this model is effective;at the same time,The built driver model can also be used to solve the cold start problem of the order.This research combines the influencing factors of driver’s willingness to take orders with order recommendation,which not only enriches the research of online car-hailing order recommendation,expands the application of recommendation methods in online car-hailing dispatching business,but also helps solve problems in online car-hailing platforms.Information overload and improvement of driver satisfaction are of practical significance.