Multi-tenant-based mileage fee intelligent bidding method and device and computer equipment

Abstract

The invention relates to an intelligent mileage fee bidding method and device based on multiple tenants and computer equipment, and the method comprises the steps: obtaining supply and demand relation information in a target region, the supply and demand relation information comprising online car-hailing position information, passenger distribution data and traffic flow information; setting a mileage fee pricing neural network model based on the supply-demand relationship information; obtaining a pricing baseline template corresponding to each tenant in the target area; inputting each pricing baseline template into a mileage fee pricing neural network model for prediction, and generating a plurality of final price templates; and pushing each final price template to the corresponding tenant to complete bidding. According to the method and the device, the online car-hailing and passenger position information, hot events, supervision policies and other information in the special region are acquired in real time, and then the mileage fee bidding of each tenant is adjusted by using the mileage fee pricing neural network model, so that the reasonability of mileage fee bidding is improved, and the utilization efficiency of online car-hailing in the special region is improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of online car-hailing technology, and in particular to a multi-tenant-based mileage fee intelligent bidding method, device and computer equipment. Background Art

[0002] In the field of online car-hailing business, airports, train stations, parks and other areas have large traffic flows and the flow of people is greatly affected by policies and events. Therefore, each online car-hailing platform will regard such areas as special areas, so that the mileage fees of all online car-hailing businesses leaving or entering the special areas need to be specially set.

[0003] However, the inventors realized that different online ride-hailing platforms have inconsistent standards for mileage fees in special areas, which can easily lead to confusion and dissatisfaction among passengers. In addition, the online ride-hailing fees in special areas may fluctuate dramatically due to peak hours or special events, making it difficult for passengers to estimate travel costs. In addition, in the absence of an effective market regulation mechanism, some online ride-hailing platforms may resort to malicious price cuts to attract passengers, undermining the healthy development of the industry. Finally, the calculation method of mileage fees should depend on the actual mileage of the passengers, but because the mileage fees for online ride-hailing services in special areas are too high compared to ordinary areas, passengers are forced to stay away from special areas before taking a taxi, resulting in reduced utilization efficiency of online ride-hailing services in special areas. Summary of the invention

[0004] In response to the above-mentioned deficiencies or shortcomings, the present application provides a multi-tenant-based intelligent bidding method, device and computer equipment for mileage fees. The present application uses a combination of Geographic Information System (GIS) and real-time data stream processing Apache Kafka technology to obtain real-time information on online ride-hailing vehicles and passenger locations, hot events, regulatory policies, etc. in special areas, and then uses the set mileage fee pricing neural network model to adjust the mileage fee bids of each tenant, avoiding malicious bidding by multiple tenants in special areas, improving the rationality of mileage fee bids, and improving the utilization efficiency of online ride-hailing vehicles in special areas.

[0005] According to a first aspect, the present application provides a multi-tenant-based mileage fee intelligent bidding method, the method comprising:

[0006] Obtaining supply and demand relationship information in the target area, including online car-hailing location information, passenger distribution data, and traffic flow information;

[0007] Set up a mileage fee pricing neural network model based on supply and demand information;

[0008] Obtain a pricing baseline template corresponding to each tenant in the target area, where the pricing baseline template refers to a regional distance price list that is priced based on the length of various online ride-hailing trips starting from the target area;

[0009] Input each pricing baseline template into the mileage fee pricing neural network model for prediction, and determine the final price template corresponding to each pricing baseline template;

[0010] Push each final price template to the corresponding tenant to complete the bidding.

[0011] In some embodiments, the target area is a limited area determined on an electronic map according to business requirements; obtaining a pricing baseline template corresponding to each tenant in the target area includes:

[0012] Determine the platform type identifier corresponding to each tenant in the target area;

[0013] According to the current time period and each platform type identifier, determine the service date type of each tenant, which service date type is a working day or a rest day;

[0014] Based on the platform type identification and service date type corresponding to each tenant, a pricing baseline template corresponding to multiple tenants is generated; wherein each pricing baseline template includes a starting fee price list, a segmented mileage fee price list, and a segmented duration fee price list corresponding to each time period.

[0015] In some embodiments, the supply and demand relationship information is obtained from the global positioning system (GPS) of each online car-hailing vehicle in the target area based on the geographic information system GIS; the mileage fee pricing neural network model is set based on the supply and demand relationship information, including:

[0016] Using the real-time data stream processing technology Apache Kafka, determine a first mapping relationship data set between the identification information of each online-hailing vehicle and the location information of the online-hailing vehicle;

[0017] Perform data cleaning operations on the first mapping relationship data set through the distributed file system Hadoop HDFS (Hadoop Distributed System) and the data warehouse technology ETL (Extract-Transform-Load) to generate a tenant online car-hailing identification and positioning report;

[0018] A mileage fee pricing neural network model is set based on the tenant's online car-hailing identification and positioning report.

[0019] In some embodiments, a mileage fee pricing neural network model is set based on the tenant's online car-hailing identification and location report, including:

[0020] Using TensorFlow technology, define the input layer according to the data type of the tenant's online car-hailing identification and location report;

[0021] Construct the hidden layer of Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN);

[0022] Connect the CNN hidden layer and the RNN hidden layer, so that the output information of the CNN hidden layer is converted and input into the RNN hidden layer;

[0023] Define the output layer based on smart bidding requirements;

[0024] Select the optimizer type and select the activation function for the optimizer. The activation function can be any of ReLU (Rectified Linear Unit, linear rectification function), sigmoid, tanh, and Leaky ReLU (Leaky Rectified Linear Unit, linear rectification function variant);

[0025] Execute the initialization function corresponding to the activation function to complete the bias initialization configuration of the mileage pricing neural network model.

[0026] In some embodiments, the traffic flow information includes the passenger flow of the target area on that day, and the method further includes:

[0027] Using real-time data stream processing technology Apache Kafka to determine a second mapping relationship data set between identity information of each passenger and passenger distribution data;

[0028] Performing a preprocessing operation on the second mapping relationship data set and the traffic flow information to obtain formatted data for time series analysis;

[0029] The fitted differential autoregressive moving average model ARIMA (Autoregressive Integrated Moving Average Model) is used to predict the passenger demand in each future period.

[0030] In some embodiments, the method further comprises:

[0031] Use mean square error (MSE) and mean absolute error (MAE) to evaluate the prediction results of the current ARIMA model.

[0032] Update the optimizer's parameter weights and / or bias initialization configuration based on the loss function gradients of the mean square error (MSE) and mean absolute error (MAE) of the prediction results;

[0033] The mileage pricing neural network model is iteratively trained through the reinforcement learning algorithms DQN (Deep Q-Network) and PPO (Proximal Policy Optimization). MSE and MAE are used to evaluate the current ARIMA prediction results in each iterative training.

[0034] In some embodiments, each pricing baseline template is input into the mileage fee pricing neural network model for prediction, and a final price template corresponding to each pricing baseline template is determined, including:

[0035] Based on the passenger demand in each future period and the input pricing baseline template, multiple payable pricing templates are generated after processing using the mileage fee pricing neural network model;

[0036] According to the real-time policies and conditions of the target area, each payable pricing template is modified to obtain the final price template.

[0037] In some embodiments, the target area is an airport area; and according to the real-time policies and conditions of the target area, each payable pricing template is modified to obtain a final price template, including:

[0038] Use Flink or Storm technology to obtain real-time policies and conditions in the airport area; policies and conditions include flight dynamic information, event events, and weather information;

[0039] Extract the time information of flight dynamics and events, and determine the geo-fence effective area of ​​flight dynamics and events through the Geographic Information System (GIS);

[0040] According to flight dynamics, event occurrence time information, and geographic fence effective area, the set rule engine is used to increase or decrease the price of each payable pricing template to obtain the final price template.

[0041] According to a second aspect, the present application provides a multi-tenant-based mileage fee intelligent bidding device, the device comprising:

[0042] The regional supply and demand relationship acquisition module is used to obtain the supply and demand relationship information in the target area. The supply and demand relationship information includes the location information of online car-hailing vehicles, passenger distribution data and traffic flow information;

[0043] A pricing model building module, used to set up a mileage fee pricing neural network model based on supply and demand information;

[0044] An initial pricing acquisition module is used to obtain a pricing baseline template corresponding to each tenant in the target area. The pricing baseline template refers to a regional distance price list based on the length of various online car-hailing trips starting from the target area;

[0045] A final pricing adjustment module, used to input each pricing baseline template into the mileage fee pricing neural network model for prediction, and determine the final price template corresponding to each pricing baseline template;

[0046] The final bid update module is used to push each final price template to the corresponding tenant to complete the bidding.

[0047] According to a third aspect, the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned multi-tenant-based mileage fee intelligent bidding methods are implemented.

[0048] The method described in the above embodiment can be applied to a SaaS (Software as a Service) cloud platform, wherein there are generally multiple online car-hailing tenants on the SaaS cloud platform. For special areas such as airports, there will also be multiple tenants to handle online car-hailing services for special areas, and these tenants belong to the online car-hailing service providers in the special area. Generally, in the special area, each online car-hailing car has only one online car-hailing driver, and the terminal device (such as a mobile phone) of the online car-hailing driver will be installed with one or more tenants' clients (for example, the mobile phone of the online car-hailing driver a with the online car-hailing identification A is installed with the clients of two tenants, Sunshine Travel and Didi Travel), that is, each online car-hailing identification (such as a license plate number) can be mapped with multiple tenants, for example, the license plate number A corresponds to the tenants of Sunshine Travel and Didi Travel. Further, the SaaS cloud platform first uses the combination of the geographic information system GIS and the real-time data stream processing Apache Kafka technology to obtain the supply and demand relationship information in the target area (i.e., the above-mentioned special area) in real time, and the supply and demand relationship information includes the location information of the online car-hailing car, the passenger distribution data and the traffic flow information, and sets the mileage fee pricing neural network model based on the supply and demand relationship information. Then, the pricing baseline templates corresponding to each tenant in the target area are obtained through Apache Kafka technology, and each pricing baseline template is input into the above-mentioned mileage fee pricing neural network model for prediction, and the final price template corresponding to each pricing baseline template is determined, and finally each final price template is pushed to the corresponding tenant to complete the bid. In this way, malicious bidding by multiple tenants in special areas is avoided, the rationality of mileage fee bids is improved, and the utilization efficiency of online ride-hailing in special areas is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flowchart of a multi-tenant-based mileage fee intelligent bidding method in one or more embodiments of the present application;

[0050] Figure 2 A schematic flow chart of a method for obtaining a pricing baseline template for each tenant in a target area in one or more embodiments of the present application;

[0051] Figure 3 This is an example diagram of a pricing baseline template for "Sunshine Travel Economy" in one or more embodiments of the present application;

[0052] Figure 4 A flowchart of a method for setting a mileage fee pricing neural network model according to supply and demand relationship information in one or more embodiments of the present application;

[0053] Figure 5 A flowchart of a method for setting a mileage fee pricing neural network model based on a tenant's online car-hailing identification and location report in one or more embodiments of the present application;

[0054] Figure 6 A flowchart of a method for predicting passenger demand in each future time period using ARIMA in one or more embodiments of the present application;

[0055] Figure 7 A flowchart of a method for iteratively training a mileage fee pricing neural network model in one or more embodiments of the present application;

[0056] Figure 8 This is an example diagram of a final price template for a tenant in one or more embodiments of the present application;

[0057] Fig. 9 A schematic flow chart of a method for determining a final price template corresponding to each pricing baseline template in one or more embodiments of the present application;

[0058] Fig.10 A flowchart of a method for obtaining a final price template after modifying each payable pricing template in one or more embodiments of the present application;

[0059] Fig.11 A schematic diagram of a process of a SaaS platform making a unified bid for n tenants in an airport area in one or more embodiments of the present application;

[0060] Fig.12 A schematic diagram of the structure of a multi-tenant based mileage fee intelligent bidding device in one or more embodiments of the present application;

[0061] Fig.13A schematic diagram of the structure of a computer device in one or more embodiments of the present application. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] According to the first aspect, the present application provides a multi-tenant based mileage fee intelligent bidding method, such as Figure 1 As shown, the method includes:

[0064] S110: Obtaining supply and demand relationship information in the target area, where the supply and demand relationship information includes online car-hailing location information, passenger distribution data, and traffic flow information;

[0065] Specifically, the method can be applied to a SaaS cloud platform, on which there are multiple online car-hailing tenants to handle online car-hailing services for the target area. In the target area, each online car-hailing vehicle can be installed with one or more tenants, for example, each online car-hailing vehicle license plate number corresponds to one or more tenants. First, the SaaS cloud platform first obtains the online car-hailing vehicle and passenger location information, hot events, regulatory policies and other information in the target area in real time through a combination of GIS and Apache Kafka technology as the above-mentioned supply and demand relationship information.

[0066] In some embodiments, the target area is a limited area determined on an electronic map according to business needs; obtaining a pricing baseline template corresponding to each tenant in the target area, such as Figure 2 As shown, including:

[0067] S210: Determine the platform type identifier corresponding to each tenant in the target area;

[0068] S220: Determine the service date type of each tenant according to the current time period and each platform type identifier, where the service date type is a working day or a rest day;

[0069] S230: Generate pricing baseline templates corresponding to multiple tenants according to the platform type identifier and service date type corresponding to each tenant; wherein each pricing baseline template includes a starting fee price list, a segmented mileage fee price list, and a segmented duration fee price list corresponding to each time period.

[0070] Specifically, the SaaS cloud platform can obtain the platform type identifier corresponding to each tenant in real time through Apache Kafka technology and then generate pricing baseline templates corresponding to multiple tenants.

[0071] For example, Figure 3As shown, the target area can be the virtual airport area fence determined on the AutoNavi map according to the airport diversion needs, and the current time period can be 8:30 p.m. on Sunday, that is, 20:28. It can be determined that the information in the pricing baseline template with the platform type identified as "Sunshine Travel Economy" includes: the current starting fee is 15 yuan for the first 3 kilometers, the segmented mileage fee is 1.4 yuan per kilometer, and the segmented time fee is 0.45 yuan per minute. In this way, the pricing baseline template corresponding to each tenant can be generated in real time, which improves the reliability and timeliness of the subsequent mileage fee pricing neural network model during the setting and training process.

[0072] S120: Setting a mileage fee pricing neural network model based on supply and demand relationship information;

[0073] In some embodiments, the supply and demand relationship information is obtained from the global positioning system GPS of each online car-hailing vehicle in the target area based on the geographic information system GIS; a mileage fee pricing neural network model is set based on the supply and demand relationship information, such as Figure 4 As shown, the method includes:

[0074] S310: Determine a first mapping relationship data set between identification information of each online-hailing vehicle and location information of the online-hailing vehicle using real-time data stream processing technology;

[0075] Specifically, the first mapping relationship data set may include a one-to-one mapping relationship between all online-hailing vehicle license plates in the target area and their corresponding longitude and latitude positions.

[0076] S320: Performing data cleaning operations for the first mapping relationship data set through the distributed file system Hadoop HDFS and the data warehouse technology ETL to generate a tenant online car-hailing identification and positioning report;

[0077] S330: Setting a mileage fee pricing neural network model based on the tenant's online car-hailing identification and positioning report.

[0078] Specifically, first, the first mapping relationship data set also includes the driving data, user information, order details, etc. of the online car-hailing service, and then these data are stored in Hadoop HDFS, which is suitable for storing large-scale data sets and can be horizontally expanded. Then, the ETL tool is used to process the data stored in HDFS. The process includes: reading the original data from HDFS, cleaning and converting the data, including removing duplicate records, correcting erroneous data, standardizing the format, filtering out incomplete records, etc. Finally, a specific BI (Business Intelligence) tool or custom script is used to extract the cleaned data from the ETL data warehouse to generate a tenant online car-hailing identification and positioning report. The content of the report may include: the frequency and time period of the tenant's processing of the online car-hailing service, the destinations and driving routes frequently visited by the online car-hailing service, the distribution and use of the online car-hailing service, etc. Finally, the mileage fee pricing neural network model is set based on the report. In this way, real-time reporting feedback on the supply and demand relationship information of online car-hailing services in special areas is realized, laying the foundation for the subsequent setting of the mileage fee pricing neural network model.

[0079] In some embodiments, a mileage fee pricing neural network model is set based on the tenant's online car-hailing identification and location report, such as Figure 5 As shown, the method includes:

[0080] S410: Using TensorFlow technology, define the input layer according to the data type of the tenant's online car-hailing identification and positioning report;

[0081] Specifically, assume that the tenant online car-hailing identification includes the tenant identification type and license plate number, and the positioning report contains the latitude and longitude coordinates, timestamp, and speed of each online car-hailing vehicle. TensorFlow technology can be used to perform feature engineering based on these data types to extract or construct useful features. For example, extract time data such as date and hour from the timestamp, and calculate vector data of distance and direction from longitude and latitude. Then convert the time data and vector data to a unified scale to achieve data standardization, so as to adapt to the subsequent convolutional neural network CNN.

[0082] S420: Construct the hidden layer of convolutional neural network (CNN) and the hidden layer of recurrent neural network (RNN);

[0083] S430: Connect the CNN hidden layer and the RNN hidden layer, so that the output information of the CNN hidden layer is converted and input into the RNN hidden layer;

[0084] Specifically, a CNN model and its CNN hidden layer can be defined first to extract the above-mentioned standardized time data and vector data. After being processed by the CNN hidden layer, feature vector data of a fixed size is output. Then the hidden layer of the CNN model is connected to the hidden layer of the newly constructed RNN model, and these fixed-size feature vector data are adjusted or formatted and input into the hidden layer of the RNN model.

[0085] S440: define an output layer according to smart bidding requirements;

[0086] S450: Select an optimizer type and select an activation function for the optimizer, where the activation function is any one of ReLU, sigmoid, tanh, and Leaky ReLU;

[0087] Specifically, if ReLU is selected as the activation function, the optimizer has the advantages of being simple and fast, and can meet the needs of mileage fee bidding with high timeliness requirements, but there is a dead zone problem, which is not conducive to the convergence of bidding results. If Sigmoid is selected, the optimizer is suitable for binary classification problems and can meet the needs of mileage fee bidding that needs to exclude abnormal services. If Tanh is selected, although the optimizer also has the advantages of being simple and fast, the calculation is relatively complex and not as fast as ReLU.

[0088] For example, assuming that the smart bidding requirement is: "the timeliness of bidding is high, and the bidding needs to be stably converged", the optimizer needs to be simple and fast, and avoid the dead zone problem. Then the selection function can use a variant of ReLU: Leaky ReLU.

[0089] S460: Execute the initialization function corresponding to the activation function to complete the bias initialization configuration of the mileage pricing neural network model.

[0090] For example, if Leaky ReLU is selected as the activation function, the corresponding He initialization function needs to be executed. In this way, a mileage pricing neural network model with deep learning potential is set up through a tenant online car-hailing identification and positioning report, which improves the rationality of subsequent smart bidding.

[0091] In some embodiments, the traffic flow information includes the passenger flow of the target area on that day, such as Figure 6 As shown, the method also includes:

[0092] S510: Determine a second mapping relationship data set between identity information of each passenger and passenger distribution data using real-time data stream processing technology;

[0093] Specifically, GIS or GPS can be used to collect the real-time location and identity information of passengers, and then the collected data can be integrated with the passenger's booking records, historical travel patterns, etc. The key features of the integrated information can be further identified, such as passenger identity information, current and historical locations, timestamps, common routes, etc. Finally, the mapping relationship between each passenger identity information and passenger distribution data is determined based on these key features.

[0094] S520: performing a preprocessing operation on the second mapping relationship data set and the traffic flow information to obtain formatted data for time series analysis;

[0095] Specifically, the missing values ​​of the second mapping relationship data set can be removed or filled, and the outliers or outliers of the second mapping relationship data set can be deleted. Then the non-homologous data (such as passenger distribution data and traffic flow information) in the second mapping relationship data set are merged into one data set, and then useful feature data (such as timestamp, number of passengers, traffic density) are extracted from the data set. Then all data are aligned according to a uniform time interval (such as every minute, every hour or every day). Finally, the aligned data is converted into a format suitable for time series analysis, such as converting the date time into hours, weeks, etc.

[0096] S530: Use the fitted differential autoregressive moving average model ARIMA to predict the passenger demand in each future period.

[0097] Specifically, when fitting ARIMA, it is necessary to first determine the ARIMA model parameters, which include: autoregressive term p, difference order d, and moving average term q. Among them, p is used to reflect the correlation between the current value and the previous value. d is used to reflect the number of differences required to convert a non-stationary series into a stationary series. q is used to reflect the degree of smoothness between the current value and the previous residual. These parameters are then used to fit the ARIMA model for subsequent prediction steps. In this way, real-time prediction of passenger demand in special areas in various future time periods is achieved. These passenger demands are used as the basis for iterative training of the mileage fee pricing neural network model, which improves the rationality of subsequent intelligent bidding.

[0098] In some embodiments, Figure 7 As shown, the above method also includes:

[0099] S610: Use mean square error MSE and mean absolute error MAE to evaluate the prediction result of the current difference autoregressive moving average model ARIMA;

[0100] Specifically, we can first obtain the actual passenger demand data during the forecast period, and then use the forecast value and the actual value to calculate the MSE and MAE. Generally speaking, the smaller the MSE and MAE values, the better the ARIMA forecast effect. MSE imposes a greater penalty on large forecast errors, while MAE treats all forecast errors equally.

[0101] S620: Update the parameter weights and / or bias initialization configuration of the optimizer according to the loss function gradients of the mean square error MSE and the mean absolute error MAE of the prediction results;

[0102] Specifically, in the reinforcement learning process of the mileage pricing neural network model, the loss function is usually based on the difference between the predicted result and the actual result. For DQN, this is usually the time difference (Time Division, TD) loss, while for general PPO, the loss function will include the policy gradient and the loss of the value function. Use MSE and MAE as part of the loss function to calculate the gradient of the current policy or value function parameters. Use the calculated gradient to complete the update of the parameter weight and / or bias initialization configuration through an optimizer (such as Adam or RMSprop)

[0103] S630: The mileage pricing neural network model is iteratively trained through the reinforcement learning algorithms DQN and PPO. MSE and MAE are used to evaluate the current ARIMA prediction results in each iterative training.

[0104] Specifically, the reward signal can be designed based on the MSE and MAE evaluation values ​​of the current ARIMA-based prediction results. Among them, lower MSE and MAE evaluation values ​​mean higher rewards. Then use the collected reward signal data to update the target network of DQN, and then train it through the experience replay buffer to stabilize the learning process. Use the truncated policy gradient method to update the policy network to ensure that the amplitude of the policy update is within an acceptable range to avoid large policy changes. Finally, repeat the above operations until the mileage pricing neural network model converges or reaches a predetermined number of iterations.

[0105] S130: Obtain a pricing baseline template corresponding to each tenant in the target area, where the pricing baseline template refers to a regional distance price list that is priced based on the length of various online ride-hailing trips starting from the target area;

[0106] For example, Figure 3 As shown, the type of pricing baseline template can be "Sunny Travel Economy". Assuming that the current time is 8:30 p.m. on Sunday, that is, 20:28, it can be seen from the figure: the current starting fee is 15 yuan for the first 3 kilometers, the segment mileage fee is 1.4 yuan per kilometer, and the segment time fee is 0.45 yuan per minute.

[0107] S140: Inputting each pricing baseline template into the mileage fee pricing neural network model for prediction, and determining a final price template corresponding to each pricing baseline template;

[0108] Specifically, in the final price template corresponding to each pricing baseline template, one or more items of the starting fee price list, the segmented mileage fee price list or the segmented duration fee price list will change.

[0109] For example, Figure 8 As shown, the starting fee price list, segmented mileage fee price list and segmented duration fee price list of the final price template have all changed.

[0110] In some embodiments, each pricing baseline template is input into the mileage fee pricing neural network model for prediction, and a final price template corresponding to each pricing baseline template is determined, such as Fig. 9 As shown, the method includes:

[0111] S710: generating multiple payable pricing templates based on the passenger demand in each future period and the input pricing baseline template using the mileage fee pricing neural network model;

[0112] Specifically, each payable pricing template refers to a theoretical bid template calculated based on the passenger demand forecast results, so each payable pricing template should also take into account the impact of policies and unexpected situations.

[0113] S720: According to the real-time policies and conditions of the target area, each payable pricing template is modified to obtain a final price template.

[0114] For example, an airport adjusted its parking fee policy for online ride-hailing vehicles, resulting in higher parking fees, which required drivers to pass on this cost to passengers, resulting in a significant increase in the prices of the final price template. In this way, the final price templates used for mileage fee bids can be adjusted according to real-time policies and situations, which improves the rationality of mileage fee bids and improves the utilization efficiency of online ride-hailing vehicles in special areas.

[0115] In some embodiments, the target area is an airport area; according to the real-time policies and conditions of the target area, each payable pricing template is modified to obtain a final price template, such as Fig.10 As shown, the method includes:

[0116] S810: Obtain real-time policies and conditions in the airport area through Flink or Storm technology; policies and conditions include flight dynamic information, event events, and weather information;

[0117] Specifically, Flink can process and analyze the real-time policies and situations in the airport area by subscribing to data in Kafka, and then store the results in data warehouses such as Hologres. Storm has similar data stream processing capabilities as Flink, and Storm's topology allows SaaS cloud platform users to define the flow of data streams.

[0118] S820: extracting the occurrence time information of flight dynamics and events, and determining the geographic fence effective area of ​​the flight dynamics and events through the geographic information system GIS;

[0119] Specifically, flight dynamics and event information can be collected from processed and analyzed data sources of policies and situations. These data sources may include airlines, airports, weather services, etc., and the occurrence time of flight dynamics and events can be extracted from the collected data sources, which usually include the estimated departure time, actual departure time, estimated arrival time, actual arrival time, etc. of the flight. In addition, before GIS analysis, it is necessary to determine the starting and ending points of the flight, as well as any related stopovers. Then use GIS software to define a geographic fence, which can be a specific area around the airport or a key point or area on the route. At the same time, GIS tools are also used to perform geospatial analysis of flight dynamics and events to determine whether they occur within the defined geographic fence, and then the time information is associated with the results of the geographic fence analysis to determine which flight dynamics and events occur within a specific geographic fence area.

[0120] S830: According to the flight dynamics, the time of occurrence of the event, and the effective area of ​​the geographic fence, the set rule engine is used to increase or reduce the price of each payable pricing template to obtain the final price template.

[0121] Specifically, when setting up the rule engine, it is necessary to clearly define the rules for price increases or decreases. These rules can be based on time (such as peak hours), location (such as specific airports or routes), event type (such as weather delays, special events), etc. Then set up geographic fences in the GIS system to determine which areas of flight dynamics and events will trigger price adjustments, and input these situations into the rule engine. Then, the current flight dynamics and event occurrence time information, and the geographic fence effective area are input into the rule engine again to increase or reduce the prices of each payable pricing template. In this way, it can be ensured that price adjustments are in line with business strategies and can flexibly respond to market changes and special circumstances, thereby improving business competitiveness and customer satisfaction.

[0122] S150: Push each final price template to the corresponding tenant to complete the bidding.

[0123] In some embodiments, Fig.11 As shown, the mileage fee intelligent bidding method can be applied to the SaaS cloud platform. It is known that in the airport area, there are n online car-hailing tenants in the SaaS cloud platform. The SaaS cloud platform first uses the combination of geographic information system GIS and real-time data stream processing Apache Kafka technology to obtain the supply and demand relationship information in the airport area in real time. The supply and demand relationship information includes the location information and pricing baseline templates of the online car-hailing tenants under n online car-hailing tenants, as well as the passenger distribution data and traffic flow information in the airport area, and stores this information in Hadoop HDFS, and uses ETL technology to perform data cleaning operations. Then the pricing baseline template corresponding to each tenant is input into the above-mentioned mileage fee pricing neural network model for prediction, and the payable price template corresponding to each pricing baseline template is determined. At this time, it is necessary to obtain the policy and real-time situation of the airport area in real time through Flink and Storm technology and input it into the rule engine to adjust the payable price template and generate multiple final price templates. Finally, each final price template is pushed to the corresponding tenant to complete the bid. In this way, malicious bidding by multiple tenants in special areas is avoided, the rationality of mileage fee bids is improved, and the utilization efficiency of online ride-hailing vehicles in special areas is improved.

[0124] It should be noted that, with respect to the various steps included in the multi-tenant-based mileage fee intelligent bidding method provided in any of the above embodiments, unless otherwise clearly stated in this document, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least part of these steps may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0125] According to the second aspect, the present application provides a multi-tenant based mileage fee intelligent bidding device, such as Fig.12 As shown, the device comprises:

[0126] The regional supply and demand relationship acquisition module 110 is used to obtain the supply and demand relationship information in the target area, and the supply and demand relationship information includes the location information of the online car-hailing vehicle, the passenger distribution data and the traffic flow information;

[0127] A pricing model building module 120, for setting a mileage fee pricing neural network model based on supply and demand relationship information;

[0128] An initial pricing acquisition module 130 is used to acquire a pricing baseline template corresponding to each tenant in the target area, where the pricing baseline template refers to a regional distance price list that is priced based on the length of various online ride-hailing trips starting from the target area;

[0129] The final pricing adjustment module 140 is used to input each pricing baseline template into the mileage fee pricing neural network model for prediction, and determine the final price template corresponding to each pricing baseline template;

[0130] The final bid updating module 150 is used to push each final price template to the corresponding tenant to complete the bidding.

[0131] In some embodiments, the initial pricing acquisition module 130 is also used to determine the platform type identifier corresponding to each tenant in the target area; determine the service date type of each tenant based on the current time period and each platform type identifier, and the service date type is a weekday or a weekend; generate pricing baseline templates corresponding to multiple tenants based on the platform type identifier and service date type corresponding to each tenant; wherein each pricing baseline template includes a starting fee price list, a segmented mileage fee price list, and a segmented duration fee price list corresponding to each time period.

[0132] In some embodiments, the pricing model building module 120 is also used to determine a first mapping relationship data set between the identification information and location information of each online-hailing vehicle using real-time data stream processing technology; perform data cleaning operations for the first mapping relationship data set through a distributed file system and data warehouse technology to generate a tenant online-hailing vehicle identification and positioning report; and set a mileage fee pricing neural network model based on the tenant online-hailing vehicle identification and positioning report.

[0133] In some embodiments, the pricing model construction module 120 is also used to use TensorFlow technology to define the input layer according to the data type of the tenant's online car-hailing identification and positioning report; construct a convolutional neural network hidden layer and a recurrent neural network hidden layer; connect the convolutional neural network hidden layer and the recurrent neural network hidden layer, so that the output information of the convolutional neural network hidden layer is converted and input into the recurrent neural network hidden layer; define the output layer according to the smart bidding requirements; select the optimizer type and select the activation function for the optimizer, and the activation function selects any one of ReLU, sigmoid, tanh and Leaky ReLU; execute the initialization function corresponding to the activation function to complete the bias initialization configuration of the mileage pricing neural network model.

[0134] In some embodiments, the pricing model building module 120 is also used to determine a second mapping relationship data set between the identity information of each passenger and the passenger distribution data using real-time data stream processing technology; perform preprocessing operations on the second mapping relationship data set and traffic flow information to obtain formatted data for time series analysis; and use the fitted differential autoregressive moving average model to predict passenger demand in each future time period.

[0135] In some embodiments, the pricing model building module 120 is also used to use the mean square error and the mean absolute error to evaluate the prediction results of the current differential autoregressive moving average model; update the parameter weights and / or bias initialization configuration of the optimizer according to the loss function gradient of the mean square error and the mean absolute error of the prediction results; and iteratively train the mileage pricing neural network model through the reinforcement learning algorithms DQN and PPO, and use the mean square error and the mean absolute error to evaluate the prediction results of the current differential autoregressive moving average model in each iterative training.

[0136] In some embodiments, the final pricing adjustment module 140 is also used to generate multiple payable pricing templates based on the passenger demand in each future time period and the input pricing baseline template, using the mileage pricing neural network model for processing; and to obtain the final price template after modifying each payable pricing template based on the real-time policies and conditions of the target area.

[0137] In some embodiments, the final pricing adjustment module 140 is also used to obtain real-time policies and conditions in the airport area through Flink or Storm technology; the policies and conditions include flight dynamics information, event activities, and weather information; extract the occurrence time information of flight dynamics and events, and determine the geo-fence effective area of ​​the flight dynamics and events through the geographic information system; according to the occurrence time information of flight dynamics and events and the geo-fence effective area, use the set rule engine to increase or reduce the price of each payable pricing template to obtain the final price template.

[0138] For the specific limitations applicable to the multi-tenant-based mileage fee intelligent bidding device, please refer to the above limitations on the multi-tenant-based mileage fee intelligent bidding method, which will not be repeated here. The various modules in the above-mentioned multi-tenant-based mileage fee intelligent bidding device can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0139] According to a third aspect, the present application provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned multi-tenant-based mileage fee intelligent bidding methods are implemented.

[0140] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig.13 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to mileage fee smart bidding. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, any of the above-mentioned multi-tenant-based mileage fee smart bidding methods is implemented.

[0141] Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (RamCUs), direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0142] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0143] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

[0144] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

Claims

1. A multi-tenant based mileage fee intelligent bidding method, characterized in that: include: Obtaining supply and demand relationship information in a target area, wherein the supply and demand relationship information includes online car-hailing location information, passenger distribution data, and traffic flow information; Setting a mileage fee pricing neural network model based on the supply and demand relationship information; Obtaining a pricing baseline template corresponding to each tenant in the target area, wherein the pricing baseline template refers to a regional distance price list that is priced according to the length of various online car-hailing trips starting from the target area; Inputting each of the pricing baseline templates into the mileage fee pricing neural network model for prediction, and determining a final price template corresponding to each of the pricing baseline templates; Each of the final price templates is pushed to the corresponding tenant to complete the bidding.

2. The method according to claim 1, characterized in that The target area is a limited area determined on an electronic map according to business requirements; obtaining a pricing baseline template corresponding to each tenant in the target area includes: Determine the platform type identifier corresponding to each of the tenants in the target area; Determine the service date type of each tenant according to the current time period and each platform type identifier, where the service date type is a working day or a rest day; Based on the platform type identifier and service date type corresponding to each of the tenants, multiple pricing baseline templates corresponding to the tenants are generated; wherein each of the pricing baseline templates includes a starting fee price list, a segmented mileage fee price list, and a segmented duration fee price list corresponding to each time period.

3. The method according to claim 2, characterized in that The supply and demand relationship information is obtained from the global positioning systems of each online-hailing vehicle in the target area based on a geographic information system; The mileage fee pricing neural network model is set based on the supply and demand relationship information, including: Determine a first mapping relationship data set between identification information of each online-hailing vehicle and location information of the online-hailing vehicle using real-time data stream processing technology; Performing data cleaning operations for the first mapping relationship data set through a distributed file system and data warehouse technology to generate a tenant online car-hailing identification and positioning report; The mileage fee pricing neural network model is set based on the tenant's online car-hailing identification and positioning report.

4. The method according to claim 3, characterized in that The mileage fee pricing neural network model is set based on the tenant's online car-hailing identification and positioning report, including: Using TensorFlow technology, an input layer is defined according to the data type of the tenant's online car-hailing identification and positioning report; Construct the hidden layer of convolutional neural network and the hidden layer of recurrent neural network; Connecting the convolutional neural network hidden layer and the recurrent neural network hidden layer, so that the output information of the convolutional neural network hidden layer is converted and input into the recurrent neural network hidden layer; Define the output layer based on smart bidding requirements; Select an optimizer type and select an activation function for the optimizer, wherein the activation function is any one of ReLU, sigmoid, tanh, and Leaky ReLU; Execute the initialization function corresponding to the activation function to complete the bias initialization configuration of the mileage pricing neural network model.

5. The method according to claim 4, characterized in that The traffic flow information includes the passenger flow of the target area on that day; the method further includes: Determine a second mapping relationship data set between the identity information of each passenger and the passenger distribution data using real-time data stream processing technology; Performing a preprocessing operation on the second mapping relationship data set and the traffic flow information to obtain formatted data for time series analysis; The fitted differential autoregressive moving average model is used to predict the passenger demand in each future period.

6. The method according to claim 5, characterized in that The method further comprises: Using mean square error and mean absolute error to evaluate the prediction results of the current differential autoregressive moving average model; updating the parameter weights and / or bias initialization configuration of the optimizer according to the loss function gradients of the mean square error and the mean absolute error of the prediction results; The mileage pricing neural network model is iteratively trained by using the reinforcement learning algorithms DQN and PPO. During each iterative training, the mean square error and the mean absolute error are used to evaluate the prediction results of the current differential autoregressive moving average model.

7. The method according to claim 6, characterized in that Inputting each of the pricing baseline templates into the mileage fee pricing neural network model for prediction, and determining a final price template corresponding to each of the pricing baseline templates, including: According to the passenger demand in each future period and the input pricing baseline template, a plurality of payable pricing templates are generated after processing by the mileage fee pricing neural network model; According to the real-time policies and conditions of the target area, each of the payable pricing templates is modified to obtain the final price template.

8. The method according to claim 7, characterized in that The target area is an airport area; and the final price template is obtained by modifying each of the payable pricing templates according to the real-time policies and conditions of the target area, including: Obtain the real-time policies and conditions of the airport area through Flink or Storm technology; the policies and conditions include flight dynamic information, event events, and weather information; Extract the occurrence time information of flight dynamics and events, and determine the geo-fence effective area of ​​flight dynamics and events through the geographic information system; According to the flight dynamics, the time of occurrence of the event, and the effective area of ​​the geographic fence, the set rule engine is used to increase or reduce the price of each payable pricing template to obtain the final price template.

9. A multi-tenant based mileage fee intelligent bidding device, characterized in that: The device includes: A regional supply and demand relationship acquisition module is used to obtain supply and demand relationship information in a target area, wherein the supply and demand relationship information includes online car-hailing location information, passenger distribution data, and traffic flow information; A pricing model building module, used to set a mileage fee pricing neural network model based on the supply and demand relationship information; An initial pricing acquisition module, used to acquire a pricing baseline template corresponding to each tenant in the target area, wherein the pricing baseline template refers to a regional distance price list based on the length of various online car-hailing trips starting from the target area; A final pricing adjustment module, used for inputting each of the pricing baseline templates into the mileage fee pricing neural network model for prediction, and determining a final price template corresponding to each of the pricing baseline templates; The final bid updating module is used to push each of the final price templates to the corresponding tenant to complete the bidding.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images