Recently, "A matching method, device, computer equipment and storage medium for service business" and "Processing method, device, computer equipment, storage medium and vehicle" applied by Shanghai Junzheng Network Technology Co., Ltd. The two inventions of "Scheduling System" have both received invention patent authorization notices issued by the State Intellectual Property Office. It is understood that the two invention patents are respectively applied to the Hello Hitchhiking and Hello Taxi-hailing businesses, which will respectively improve the driver-rider matching efficiency of Hitchhiking and the capacity dispatching capabilities of online ride-hailing services.
Hello Brain Matching Engine 2.0 adds “Intelligent Recommendation Assistant”
Enjoyable travel has penetrated into people’s lives, and the model of equal and mutual assistance and shared rides has not only improved people’s The convenience of travel also reduces travel costs for both drivers and passengers. However, when drivers and passengers use ride-hailing services, they usually have a large number of the same itineraries. When the number of orders is large in the same time period, the back-end server needs to calculate the matching party corresponding to each order in time. For For frequent driver users, there will be a certain lag in service matching, which indirectly affects the user experience.
The invention of "a matching method, device, computer equipment and storage medium for service business" aims to analyze and obtain the travel characteristics and dynamics of both parties through the historical ticket issuance behavior of both parties. Calculate the travel behavior at the next time, and intelligently recommend car owners or passengers on the road in two directions, which is equivalent to an "intelligent recommendation assistant" for platform users.
It is understood that the Hello Hitch backend currently receives massive amounts of itinerary data every day. This patent allows the backend to effectively filter out highly relevant itineraries through geographical information; and uses the Hello Brain matching engine to calculate the matching results in 2.0 seconds and make personalized recommendations to both drivers and passengers.
The prospectus submitted by Hello Travel on April 24 shows that in 2020, the total transaction volume of Hello Ride was 7 billion yuan, making it the second largest ride-hailing trading platform in China. By the end of 2020, HelloHitch has accumulated 26.1 million transaction users and nearly 10 million registered drivers.
Predict the grid supply and demand ratio and carry out transportation capacity scheduling
The invention of "processing methods, devices, computer equipment, storage media and vehicle dispatching systems" is to solve the different problems of Hello Taxi The problem of regional supply and demand balance can be based on deep learning to predict supply and demand and balance transportation capacity, so that users in various regions can get a taxi faster. Currently, Hello's taxi-hailing business has been launched in four cities: Zhongshan, Huizhou, Heyuan and Shanwei in Guangdong Province.
In this patent, Hello Taxi’s backend divides the operating area into a number of grids. Through neural network models and deep learning, it considers the orders, transportation capacity and road conditions of different grids at a certain time to predict The supply and demand ratio of each grid in a certain period of time in the future, and certain supply and demand intervention behaviors will be carried out based on the supply and demand ratio, such as price intervention, capacity scheduling, marketing intervention, etc.
This patent matches supply and demand from multiple dimensions. During the supply and demand matching process, the driver’s location is reported in real time and immediately uploaded to the Hello Brain Matching Engine 2.0. At a certain moment, a passenger initiates a travel request from A to B. When the backend gets this data, it will immediately match it with the driver selected based on certain conditions in the pool. Factors that will be considered for matching include distance, convenience, real-time traffic conditions, passenger characteristics and preferences, driver registration characteristics and platform behavioral characteristics. The entire matching process is completed within 200 milliseconds.
Driver-rider matching also includes personalized feature matching. For example, individual users may only like a certain type of car and a certain type of driver. Female users may have higher requirements for car models and drivers at 11 or 2 o'clock in the night, which require personalized matching.
Author: Zhang Xiaoming