Simulation method and system based on supply and demand order

Abstract

The embodiment of the invention discloses a simulation method and system based on supply and demand order formation. Through behavior simulation and order forming scheme reproduction, the influence of the order forming scheme on the user experience can be simulated in a low-cost, low-risk, high-efficiency and high-quality mode, so that iterative optimization of the order forming scheme is promoted.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a simulation method and system based on supply and demand ordering. Background Art

[0002] In a two-sided market environment, continuously improving the order efficiency and experience of both supply and demand sides (such as drivers and passengers) is the core task of online-to-offline service platforms (such as online ride-hailing platforms), and the order solution is one of the important factors affecting the order efficiency and experience of both supply and demand sides.

[0003] The general process of iterative optimization of order placement schemes is: first obtain an optimized version of a heuristic scheme or a model offline (i.e., a new order placement scheme), conduct a control experiment in a real market environment, and then collect the experimental results to evaluate the effect of the new order placement scheme. This process has the following problems: (1) When the order placement scheme is changed significantly, it may have a negative impact on the efficiency and experience of drivers and passengers in a real environment; (2) The control experiment in a real environment generally has a long cycle, such as a time slice rotation experiment, which generally takes two weeks, resulting in a slow iterative optimization of the order placement scheme; (3) Some order placement schemes have many hyperparameters that need to be adapted in different supply and demand environments. Due to the limitations of factors such as the long control experiment cycle and the changing supply and demand environment, it is difficult to obtain the optimal hyperparameters, which affects the effect of the new order placement scheme.

[0004] In view of this, we hope to provide a simulation solution based on supply and demand (such as the transaction of online car-hailing orders) to simulate the impact of the order-making solution on user experience in a low-cost, low-risk, high-efficiency, and high-quality manner, thereby promoting the iterative optimization of the order-making solution. Summary of the invention

[0005] The purpose of this application is to provide a simulation solution based on supply and demand ordering.

[0006] The first aspect of the embodiments of this specification provides a simulation method based on supply and demand ordering. The method includes: obtaining the scene data of the target space-time unit; performing ordering simulation based on the scene data of the target space-time unit to obtain the ordering simulation result. The ordering simulation includes: simulating the behavior of each service requester and each service provider in the target space-time unit to obtain the behavior simulation result; determining the ordering plan of the order simulation allocation result according to the target ordering plan; and, based on the behavior simulation result and the order simulation allocation result, updating the simulation state of each order and each service provider in the target space-time unit, and the scene data of the target space-time unit is used to initialize the simulation state of each order and each service provider in the target space-time unit.

[0007] In some embodiments, acquiring the scene data of the target space-time unit includes: extracting the scene data of the target space-time unit from a database, wherein the database stores scene data of a plurality of space-time units.

[0008] In some embodiments, the scenario data includes snapshot data and event data, the snapshot data of each space-time unit reflects the real state of each order and each service provider at the initial moment of the space-time unit, and the event data of each space-time unit reflects the real behavior of each order and each service provider in the space-time unit. The scenario data of each space-time unit is used to initialize the simulation state of each order and each service provider in the space-time unit. The database also stores model data and hyperparameter data, and the method further includes: obtaining the model data and the hyperparameter data to perform order simulation based on the scenario data of the target space-time unit, the model data and the hyperparameter data. The model data includes a behavior simulation model and feature data, the feature data reflects the features of each service requester and the features of each service provider, and the behavior simulation model is used to simulate the behavior of each service requester and each service provider based on the feature data. The hyperparameter data includes the hyperparameters of the behavior simulation model and the hyperparameters of the order solution.

[0009] In some embodiments, the hyperparameters are calibrated by narrowing the gap between the estimated value and the true value of the order efficiency of the historical order-making scheme.

[0010] In some embodiments, the multiple space-time units correspond to multiple target space fences in the spatial dimension, and the multiple target space fences are obtained by performing aggregation processing on multiple initial space fences based on regional similarity, and the response closure of the multiple target space fences meets the set requirements, and the response closure is the proportion of responded orders whose departure location of the order and the response location of the service provider are within the same target space fence to all responded orders.

[0011] In some embodiments, the multiple space-time units are divided into multiple supply and demand scenarios. The extracting the scenario data of the target space-time unit from the database includes: determining the target supply and demand scenario among the multiple supply and demand scenarios according to the simulation requirements; and determining the scenario data of the space-time unit belonging to the target supply and demand scenario from the scenario data of the multiple space-time units as the scenario data of the target space-time unit. In some embodiments, the division basis of the multiple space-time units includes order density and / or demand-supply ratio, the order density of each space-time unit reflects the number of orders per unit area in the space-time unit, and the demand-supply ratio of each space-time unit reflects the ratio of the number of orders in the space-time unit to the online time of the service provider.

[0012] In some embodiments, the method further includes: evaluating the order efficiency of the target order solution based on the order simulation result to obtain an evaluation result; determining whether the evaluation result meets the preset requirements; in response to determining that the evaluation result meets the preset requirements, applying the target order solution to online-to-offline services.

[0013] In some embodiments, the service requester is a passenger and the service provider is a driver.

[0014] The second aspect of the embodiments of this specification provides a simulation system based on supply and demand ordering. The system includes: a reading module for acquiring the scene data of the target space-time unit; an execution module for performing ordering simulation based on the scene data of the target space-time unit to obtain ordering simulation results. The ordering simulation includes: simulating the behavior of each service requester and each service provider in the target space-time unit to obtain a behavior simulation result; determining an ordering plan based on the target ordering plan to simulate the distribution results; and, based on the behavior simulation results and the order simulation distribution results, updating the simulation status of each order and each service provider in the target space-time unit, and the scene data of the target space-time unit is used to initialize the simulation status of each order and each service provider in the target space-time unit.

[0015] The third aspect of the embodiments of this specification provides a simulation device based on supply and demand order formation. The device includes a processor and a storage device, the storage device is used to store instructions, and when the processor executes the instructions, the simulation method based on supply and demand order formation as described in any embodiment of this specification is implemented.

[0016] A fourth aspect of the embodiments of this specification provides a computer-readable storage medium. The storage medium stores computer instructions, and when a computer executes the instructions, a simulation method based on supply and demand ordering as described in any embodiment of this specification is implemented.

[0017] The embodiments of this specification provide a simulation method and system based on supply and demand ordering. Through behavior simulation and ordering scheme reproduction, the impact of ordering schemes on user experience can be simulated in a low-cost, low-risk, high-efficiency, and high-quality manner, thereby promoting iterative optimization of ordering schemes. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0019] Figure 1 is a schematic diagram of a scenario of online-to-offline services according to some embodiments of this specification;

[0020] Figure 2 is an exemplary module diagram of a simulation system based on supply and demand ordering according to some embodiments of this specification;

[0021] Figure 3 is an exemplary flow chart of a simulation method based on supply and demand ordering according to some embodiments of this specification;

[0022] Figure 4 is a schematic diagram of space division according to some embodiments of this specification;

[0023] Figure 5 A collection of illustrative supply and demand scenarios is presented;

[0024] Figure 6 An exemplary data classification of the database is presented;

[0025] Figure 7 is a schematic diagram of interactions within a simulation system according to some embodiments of this specification;

[0026] Figure 8 An exemplary driver-passenger state transition is shown;

[0027] Fig. 9 An exemplary driver-passenger ordering scheme is shown;

[0028] Fig.10 is an engineering architecture diagram of an exemplary simulation system;

[0029] Fig.11 An exemplary simulation system performance evaluation system is presented;

[0030] Fig.12 It is an exemplary flow chart of an order-making plan evaluation method under a target supply and demand scenario as shown in some embodiments of this specification. DETAILED DESCRIPTION

[0031] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0032] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0033] As shown in this specification, unless the context clearly indicates an exception, the words "a", "an", "a kind" and / or "the" do not refer to the singular, but also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0034] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0035] The technical solutions in this specification, if they involve the processing of personal information, will be processed on the premise of having a legal basis (such as obtaining the consent of the subject of personal information, or being necessary for the performance of a contract, etc.), and will only be processed within the prescribed or agreed scope. The user's refusal to process personal information other than the necessary information required for basic functions will not affect the user's use of basic functions. The collection, storage, use, processing, transmission, provision and presentation of relevant information shall comply with the provisions of relevant laws and regulations, do not infringe on the privacy of others, and do not violate public order and good customs.

[0036] In this specification, without causing confusion, the terms "user end" and "user" can be used interchangeably, the terms "behavior" and "event" can be used interchangeably, and the terms "server end" and "platform" can be used interchangeably.

[0037] It should be noted that although this manual mainly uses online car-hailing orders as an example, the simulation system and method provided in this manual are applicable to supply and demand orders for various online-to-offline services, including but not limited to designated driver services, moving services, same-city delivery services, towing services, and rescue services. As a matchmaker between the supply and demand parties (service requesters and service providers), the service platform needs to continuously iterate and optimize the ordering plan to improve the ordering efficiency and experience of both parties.

[0038] Figure 1 1 is a schematic diagram of an online-to-offline service scenario according to some embodiments of the present specification. Scenario 100 includes multiple user terminals 110 , a server terminal 120 , and a network 130 .

[0039] The client 110 is used to initiate a service request to the server 120 and receive a service response from the server 120. The user may refer to the server requester (such as a passenger) or the service provider (such as a driver) on both sides of the supply and demand.

[0040] The service requester (such as a passenger) can initiate an order (such as an online car-hailing order) on its user terminal 110, and receive relevant information (such as vehicle model, license plate number) from the service provider (such as a driver) matched (assigned) for the order on the service terminal 120.

[0041] The service provider (such as a driver) can receive the order assigned by the service end 120 at its user end 110, and accordingly, the order status is updated from "waiting for response" to "answered".

[0042] The user can also cancel the order on the user terminal 110. For example, the passenger can cancel the order on the user terminal 110 while waiting for the order to be answered. For another example, the driver can cancel the answered order on the user terminal 110 before picking up the passenger.

[0043] The user terminal 110 can record user behaviors related to order placement (for example, behaviors related to status updates of orders and service providers), and upload them to the server 120 to promote the advancement of the order placement process and the improvement of order placement records. For example, in response to a driver clicking "departure" or "receipt of vehicle" on his user terminal 110, the driver's user terminal 110 can record the driver's departure behavior (event) or receipt of vehicle behavior (event), and upload it to the server 120. For another example, the order itself can be regarded as a call event (indicating that the service requester "calls" the service provider). For another example, in response to a user canceling an order, the user terminal 110 can record the order cancellation behavior (event) and upload it to the server 120.

[0044] In some embodiments, the user terminal 110 may include one or more of a car computer (such as 110 - 1 ), a smart phone (such as 110 - 2 ), a wearable device (not shown), a tablet computer (not shown), etc.

[0045] The server 120 (also referred to as the platform) is used to process received service requests. For example, the server 120 can match (assign) a suitable service provider for an order initiated by a service requester. Order assignment can be achieved by executing a predetermined order-making scheme. For the specific implementation of the order-making scheme, please refer to Fig. 9 and its related description.

[0046] The server 120 can update the status of each order and each service provider according to the user behavior (event) uploaded by the client 110, for example, the status of each order and each driver (referred to as driver status). For the relationship between events and status updates (transfers), please refer to Figure 8 and its related description.

[0047] The server 120 can save various types of data, including but not limited to user feature data (for example, user ID, membership level, credit level, vehicle information, etc.), snapshot data (for example, recording driver and passenger status every half hour), event data (such as driver and passenger events such as calling, dispatching, and receiving vehicles), order-making plans, prediction models (such as response time prediction models, order cancellation behavior prediction models), and order-making efficiency (for example, the order completion rate each month).

[0048] The server 120 may be a single server or a server group.

[0049] The network 130 can facilitate information and / or data exchange. For example, the client 110 can send a service request to the server 120 via the network 130, and the server 120 can return a service response to the client 110 via the network 130. For another example, the server 120 can collect information necessary to provide basic functions (e.g., order-related services) from the client 110 via the network 130, such as location information (e.g., the latitude and longitude of the driver and the passenger).

[0050] The network 130 may include a wireless network and / or a wired network.

[0051] In some embodiments, the scene 100 may also include a storage device. The storage device may be used to store data, instructions, and / or any other information. For example, the storage device may be used as a database (e.g., Figure 6 The database 205 stores driver and passenger scenario data, model data, hyperparameter data, etc. In some embodiments, the storage device may include a large-capacity memory, a removable memory, etc., or any combination thereof. In some embodiments, the storage device may be a part of the server 120.

[0052] Figure 2 According to the exemplary module diagram of the simulation system based on supply and demand ordering shown in some embodiments of this specification. The simulation system 200 can be implemented on the server 120, or on other computer devices. In some embodiments, at least a part of the simulation system 200 can be a computer operating system (i.e., kernel) with simulation function, and is therefore also referred to as a simulation kernel. Figure 2 As shown, the simulation system 200 includes a reading module 210 and an execution module 220 .

[0053] The reading module 210 is used to obtain (e.g., read from the database 205) the scene data of the target spatiotemporal unit. In some embodiments, the simulation system 200 also includes a database 205, which is used to store various types of data that the execution module 220 needs to call, such as the scene data of the target spatiotemporal unit. In some embodiments, the simulation system 200 includes a database 205 and a simulation kernel 215, and the simulation kernel 215 includes a reading module 210 and an execution module 220.

[0054] The execution module 220 is used to perform order simulation based on the scenario data of the target space-time unit to obtain order simulation results, and the order simulation includes: simulating the behavior of each service requester and each service provider in the target space-time unit to obtain behavior simulation results; determining an order simulation allocation result order plan according to the target order plan; and, based on the behavior simulation results and the order simulation allocation results, updating the simulation status of each order and each service provider in the target space-time unit. The scenario data of the target space-time unit is used to initialize the simulation status of each order and each service provider in the target space-time unit.

[0055] In some embodiments, the simulation system 200 further includes an evaluation module 230 (or referred to as an assessment module 230). The evaluation module 230 can be used to evaluate the target ordering solution, for example, to evaluate the feasibility of the target ordering solution, ordering efficiency, etc. The evaluation module 230 can also be used to evaluate the performance of the simulation system 200 (such as the simulation kernel 215).

[0056] For more details about the simulation system 200 and its modules, please refer to Figure 3 and its related description.

[0057] It should be understood that Figure 2The system and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer executable instructions and / or included in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the system and its modules of this specification. Not only can the hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc. be implemented, it can also be implemented with software such as executed by various types of processors, and it can also be implemented by a combination of the above hardware circuits and software (for example, firmware).

[0058] It should be noted that the above description of the system and its modules is only for convenience of description and does not limit this specification to the scope of the embodiments. It is understandable that for those skilled in the art, after understanding the principle of the system, it is possible to arbitrarily combine the modules or form a subsystem connected with other modules without deviating from this principle. For example, in some embodiments, the reading module 210 and the execution module 220 can be two modules or combined into one module. Such variations are all within the scope of protection of this specification.

[0059] Figure 3 300 is an exemplary flow chart of a simulation method based on supply and demand ordering according to some embodiments of this specification. Process 300 may be executed by the server 120 or other computer devices. Specifically, process 300 may be executed by the simulation system 200 implemented on the server 120 or other computer devices. Figure 3 As shown, the process 300 includes the following steps.

[0060] Step 310 , acquiring scene data of the target spatiotemporal unit. In some embodiments, step 310 is performed by the reading module 210 .

[0061] In this specification, a spatiotemporal unit refers to a unit with a certain spatial range and a certain time range. For example, the target spatiotemporal unit may refer to Beijing on Friday. For another example, a map may be divided into multiple spatial fences (or grids), and the whole day may be evenly divided into multiple time slices (such as half an hour as a time slice), then each spatiotemporal unit may refer to a specific spatial fence of a specific time slice (such as 18:00-18:30).

[0062] The scenario data is data generated in a real environment, rather than simulated data generated in a simulation environment (such as the simulation kernel 215). In some embodiments, the scenario data includes snapshot data and event data, wherein the snapshot data reflects the real status of each order and each service provider (for example, the driver dispatches, the passenger calls, the passenger waits for a response, the response has been received, the order is cancelled, the pick-up, the delivery, the order is completed, the driver is offline, and other driver and passenger status), and the event data reflects the real behavior of each service requester and each service provider (for example, the driver cruises with an empty car, cancels an order, and the driver collects the car, and other driver and passenger events).

[0063] It should be noted that although the scene data often contains noise and may even contain errors, it is still considered to be true. In some embodiments, the scene data may also be cleaned to reduce or even remove noise and / or errors in the scene data, thereby improving the reliability of the scene data.

[0064] In some embodiments, the simulation system 200 can perform order simulation in local space and time. Specifically, the database 205 stores the scene data of multiple space-time units, and the reading module 210 reads the scene data of the target space-time unit from the database 205, which is used for the order simulation of the target space-time unit. The effect of the ordering scheme of supply and demand ordering in the local space and time of different supply and demand situations is different. Therefore, the evaluation problem of the ordering scheme (such as the evaluation of ordering efficiency) is split into the "atomic unit" scene, which can more specifically improve the effect of the ordering scheme under each "atomic scene" (such as ordering efficiency). This embodiment proposes to perform ordering simulation in local space and time. Under the premise of ensuring the effective evaluation of the ordering scheme, the amount of data that the ordering simulation depends on is greatly reduced, thereby greatly reducing the time-consuming of the ordering simulation. That is, this embodiment provides a short-term simulation scheme based on supply and demand ordering.

[0065] In some embodiments, the plurality of spatiotemporal units correspond to a plurality of target space fences in the spatial dimension, and the plurality of target space fences are obtained by performing aggregating processing based on region similarity on a plurality of initial space fences. Figure 4As shown, the initial spatial fence is a hexagonal grid, and the target spatial fence is a composite grid of the same color (composed of several hexagonal grids). The initial spatial fence can be determined by artificial division, and the size of the initial spatial fence can be dynamically adjusted according to the size of the overall area (such as the size of the city), that is, the larger the overall area, the larger the initial spatial fence. Regional similarity can refer to the feature similarity between spatial fences, for example, the similarity of features such as distance, order volume (the number of orders initiated), order response rate, and order completion rate. In this way, while ensuring the connectivity of the aggregated area (target spatial fence), the similarity of the area (initial spatial fence) (such as the similarity of issuing orders) can be taken into account. It can be understood that the control experiment is based on the idea of ​​controlling variables, and it is necessary to ensure that the order-making plan is targeted at a homogeneous population and tested at the same time. Aggregation processing based on regional similarity helps to ensure the homogeneity of the population in the aggregated area (target spatial fence).

[0066] It is worth noting that in the process of determining the target space fence, special attention can be paid to the response closure of the target space fence, that is, the response closure of the multiple target space fences meets the preset requirements. Response closure refers to the proportion of answered orders whose departure location (such as the boarding point) of the order and the response location of the service provider (such as the driver) are within the same target space fence to all answered orders. The determination of the target space fence expects that each target space fence is regarded as an isolated space-time (similar to the concept of a connected domain). Specifically, it is expected that the response closure of the target space fence is at a relatively high level. Generally speaking, the larger the target space fence, the higher its response closure, but it will reduce the ability to finely characterize supply and demand, which is not conducive to analyzing the effect of the order-making scheme under different supply and demand categories. As an example, the average value of the response closure of the multiple target space fences can be determined by technicians based on experience or based on historical data, and this application is not limited to this.

[0067] In some embodiments, the multiple space-time units are divided into multiple supply and demand scenarios. Accordingly, the reading module 210 determines the target supply and demand scenario among the multiple supply and demand scenarios according to the simulation requirements. Furthermore, the reading module 210 determines the scene data of the space-time unit belonging to the target supply and demand scenario from the scene data of the multiple space-time units as the scene data of the target space-time unit. Based on this, the effect of the order-making scheme (such as order-making efficiency) can be evaluated under different supply and demand scenarios, thereby promoting the iterative optimization of the order-making scheme for different supply and demand scenarios.

[0068] The division of supply and demand scenarios can be single-dimensional or multi-dimensional.

[0069] In some embodiments, the division of the plurality of space-time units is based on order density (or order density) and / or demand-supply ratio. The order density of each space-time unit reflects the number of orders per unit area in the space-time unit (for example, the number of orders divided by the area of ​​the target space fence). The demand-supply ratio of each space-time unit reflects the ratio of the number of orders (number of orders) in the space-time unit to the online time of the service provider (such as a driver). For example, Figure 5 As shown in the table, the multiple space-time units can be divided into 10 supply and demand scenarios. Different levels of demand-supply ratio / order density correspond to different value ranges. The supply and demand scenario numbered -1 only requires the order density level to be "very low", and there is no need to pay attention to the level of the demand-supply ratio (indicated by a horizontal bar).

[0070] In some embodiments, the multiple time-space units can also be divided into non-extreme natural supply and demand scenarios, extreme natural supply and demand scenarios, and parameter reorganization supply and demand scenarios. The scenario data of the natural supply and demand scenario comes from the real environment, for example, the actual call event (including the call time, the latitude and longitude of the starting point), the event in the order process (cancellation before answering, cancellation after answering, order completion), the driver's departure event and the car collection event (departure time, departure latitude and longitude, car collection time), etc.

[0071] In practical applications, a group of typical cities can be selected regularly (such as quarterly) or irregularly for natural supply and demand scenarios. The selection of typical cities follows the screening logic of efficiency experiments in real environments. Generally, all cities are clustered into multiple categories (also called multiple clusters) according to the response rate and the number of orders issued. In each category, several (a preset number, such as 3) cities closest to the center point (a certain city) are selected as typical cities. Only the scenario data of typical cities is imported into the database 205, which greatly reduces the data storage cost while ensuring the effective evaluation of the order-making plan.

[0072] Extreme / non-extreme natural supply and demand scenarios can be defined artificially. It can be understood that as long as the extreme natural supply and demand scenarios are defined first, the non-extreme natural supply and demand scenarios can be determined accordingly. For example, time and space units that have experienced extreme weather (such as heavy rain, heavy snow, typhoons, high temperatures, sandstorms) or major holidays (such as Labor Day, National Day, and Spring Festival) can be defined as extreme natural supply and demand scenarios. Correspondingly, non-extreme natural supply and demand scenarios are time and space units that have not experienced extreme weather and are not major holidays. Non-extreme natural supply and demand scenarios serve as sufficient testing scenarios for order-making solutions, and are used to provide an evaluation of the effectiveness of order-making solutions under normal circumstances. Extreme natural supply and demand scenarios serve as necessary testing scenarios for order-making solutions, and are used to provide an evaluation of the effectiveness of order-making solutions under extreme circumstances.

[0073] By reorganizing the scenario data of known (e.g., defined) supply and demand scenarios (e.g., known natural supply and demand scenarios) or randomly generating scenario data, scenario data of parameter reorganization supply and demand scenarios can be generated. Parameter reorganization supply and demand scenarios have the characteristics of infinity, scalability, batching, and automation, and can be used as a supplementary test scenario for order-making solutions to provide effect evaluation of order-making solutions in more situations.

[0074] It should be noted that although the scenario data of the parameter reorganization supply and demand scenario is not generated in a real environment, it can be regarded as data generated in a real environment. For example, the snapshot data and event data in the scenario data of the reorganization supply and demand scenario can be regarded as reflecting the real driver and passenger status and real driver and passenger events.

[0075] like Figure 6 As shown, for online car-hailing orders, the database 205 stores driver and passenger scenario data, model data, and hyperparameter data for multiple supply and demand scenarios. In one division method, the multiple supply and demand scenarios include non-extreme natural supply and demand scenarios, extreme natural supply and demand scenarios, and parameter reorganization natural supply and demand scenarios. In another division method, the driver and passenger scenario data include passenger event (behavior) data, passenger snapshot data, driver event (behavior) data, and driver snapshot data. Model data includes a behavior simulation model (not shown), passenger feature data, and driver feature data. Hyperparameter data includes hyperparameters of the behavior simulation model and hyperparameters of the platform (such as the online car-hailing platform) ordering scheme. The driver and passenger scenario data is the initial state input of the driver and passenger behavior simulation, and the model data is the input of the behavior simulation model. During the simulation process, the simulation kernel 215 loads the hyperparameters in the hyperparameter data to simulate the driver and passenger behavior and execute the target ordering scheme. For more details about snapshot data, event data, model data, and hyperparameter data, please refer to the following.

[0076] In some embodiments, when the database 205 stores scenario data of multiple space-time units, and the scenario data includes snapshot data and event data, the snapshot data of each space-time unit reflects the real state of each order and each service provider at the initial moment of the space-time unit, and the event data of each space-time unit reflects the real behavior of each order and each service provider in the space-time unit. In addition to the scenario data of the target space-time unit, the data used for order simulation also includes model data and hyperparameter data. That is, the database 205 also stores model data and hyperparameter data, and the simulation kernel 215 can read the model data and hyperparameter data from the database 205 to perform order simulation based on the scenario data, model data and hyperparameter data of the target space-time unit.

[0077] Among them, the model data includes a behavior simulation model and feature data. Among them, the feature data reflects the features of each service requester and the features of each service provider, for example, the features of the driver and the passenger. The feature data may include statistical data on the user's historical behavior, for example, the number of times the driver has completed / cancelled orders in the last month, and the number of times the passenger has completed / cancelled orders in the last month. The behavior simulation model is used to simulate the behavior of each service requester and each service provider (such as the driver's behavior) based on the feature data (such as the driver's and the passenger's features). The hyperparameter data includes the hyperparameters of the behavior simulation model and the hyperparameters of the order-making scheme, and the hyperparameters refer to the parameters set before the model training or the iterative optimization of the order-making scheme. The hyperparameters can be in the form of thresholds or weights. As an example only, the behavior simulation model can predict the probability of occurrence of user behavior (such as canceling order behavior). When the probability exceeds the set threshold, the behavior simulation result output by the behavior simulation model indicates that the user behavior will occur. The hyperparameters in the hyperparameter data can be reused. For example, the hyperparameters of the order-making scheme in the hyperparameter data can be the hyperparameters of the historical order-making scheme. In the subsequent simulation, the target order-making scheme can be executed according to the hyperparameters of the historical order-making scheme data.

[0078] In some embodiments, reference Figure 7 , the hyperparameter data is calibrated. For the convenience of description, the process of calibrating the hyperparameter data is called the baseline task, and the calibrated hyperparameter data is called the baseline task data. The baseline task is essentially a simulation task (see step 320 for execution details), but the purpose of the baseline task is to calibrate the hyperparameter data. The calibration of the hyperparameter data can be achieved by narrowing the gap between the evaluation value and the true value of the order efficiency (such as the transaction rate) of the historical order-making scheme. In subsequent simulation tasks, the hyperparameters in the baseline task data can be called to simulate user behavior (such as driver and passenger behavior) and execute the target order-making scheme. Furthermore, the evaluation module 230 can evaluate the order efficiency of the target order-making scheme based on the order-making simulation results (for example, to evaluate whether the transaction rate has been significantly improved).

[0079] The same behavior simulation model can be used in different spatiotemporal units (or different supply and demand scenarios), or different behavior simulation models can be used. For example, a common behavior simulation model can be used in all spatiotemporal units. For another example, a customized behavior simulation model can be prepared for each supply and demand scenario.

[0080] like Figure 7 As shown in the figure, taking online car-hailing orders as an example, the data that the simulation task relies on include driver and passenger scenario data, model data and baseline task data.

[0081] The preparatory work for order simulation includes the data preparation stage and the data import stage. In the data preparation stage, local space-time division (determine the multiple space-time units and divide the supply and demand scenarios), driver and passenger scenario data cleaning (for example, processing invalid values ​​and outliers, unique hot encoding of type features, bucketing of continuous values, etc.), model offline training and evaluation, and hyperparameter data calibration can be performed. It should be noted that the parameters adjusted in the offline training of the model are not the hyperparameters of the behavior simulation model, but the model parameters (such as weights, biases, etc.). In the data import stage, the driver and passenger scenario data, model data, and baseline task data produced in the data preparation stage can be imported into the database 250.

[0082] On the one hand, the database 205 provides standard scenarios (such as Figure 5 The 10 supply and demand scenarios in the simulation system 200 are used to evaluate the effects of different order-making solutions in a unified test environment. On the other hand, by completing the data tasks that the simulation tasks depend on in advance (including the preparation of model data and baseline task data), the order-making solution evaluation process of the simulation system 200 is simplified, and the simulation speed and ease of use of the simulation system 200 are improved.

[0083] Step 320 , performing a single simulation based on the basic data of the target spatiotemporal unit to obtain a single simulation result. In some embodiments, step 320 is performed by the execution module 220 .

[0084] refer to Figure 7 , the simulation kernel 215 may receive the target supply and demand scenario as input. Accordingly, the simulation kernel 215 may sample M samples (i.e., M target spatiotemporal units) in the target supply and demand scenario, where M is a positive integer not less than 2. Each sample corresponds to a simulation task. That is, the simulation kernel 215 may perform M simulation tasks. For example only, the target supply and demand scenario may be Figure 5 Any of the scenarios shown may also be other supply and demand scenarios (such as a combination of supply and demand scenario 0 and an extreme natural supply and demand scenario).

[0085] Specifically, step 320 includes sub-steps 322 to 326.

[0086] Step 322, simulating the behavior of each service requester and each service provider in the target space-time unit to obtain a behavior simulation result.

[0087] The scene data of the target spatiotemporal unit is used to initialize the simulated state of each order and each service provider in the target spatiotemporal unit. In some embodiments, when the database 205 stores the scene data of multiple spatiotemporal units, the scene data of each spatiotemporal unit is used to initialize the simulated state of each order and each service provider in the spatiotemporal unit. Taking the online car-hailing order as an example, the execution module 220 can determine the real driver and passenger state (snapshot data from the scene data) at the initial moment (such as 18:00) of the target spatiotemporal unit (such as a certain area from 18:00 to 18:30) as the simulated driver and passenger state at the initial moment. The execution module 220 can also determine the simulated driver and passenger state (such as an idle driver) at the subsequent moment based on the real behavior related to the initial state at the subsequent moment (such as 18:10) (such as passenger call order, driver departure). It should be understood that the update of the simulation state does not depend on the scene data (such as the real cancellation of order behavior, car collection behavior, etc.), that is, only the state and termination state in the order process are predicted.

[0088] State transitions are determined by behaviors (events). Figure 8 Exemplary driver and passenger state transitions are shown. Figure 8 Middle: Rounded boxes represent the driver and passenger status, including passenger calling, passenger waiting for answer, (answered), pick-up, delivery, order cancellation, order completion, driver dispatched, driver offline); rectangular boxes represent driver and passenger behaviors, including driver cruising with an empty car, canceling before passenger answers, canceling after passenger answers, canceling after driver answers, passenger boarding, and driver collecting the car. For simplification, Figure 8 Some behaviors are omitted, such as passenger calling for orders and driver dispatching. The starting state of an order is "passenger calling for orders", and the ending state is "order canceled" or "order completed". The starting state of a driver is "driver dispatching", and the ending state is "driver offline". Cancellation before a passenger responds, cancellation after a passenger responds, and cancellation after a driver responds will cause the order to enter the "order canceled" state. After an order enters the "order completed" state, the driver will generally start the "driver cruising with an empty car" behavior, and may then enter the "responding" state, or the "driver offline" state based on the "driver collecting the car" behavior.

[0089] In some embodiments, the simulation kernel 215 simulates the behavior of each service requester and each service provider in the target space-time unit through a behavior simulation model. The behavior simulation model uses the simulation state of each order and each service provider as state input (including initial state input and intermediate state input) and outputs the behavior simulation result. In some embodiments, the behavior simulation model includes a trained machine learning model, and the machine learning model includes but is not limited to a decision tree, a support vector machine, a Bayesian model, and a neural network. Specifically, the behavior simulation model may include multiple prediction models corresponding to multiple states / behaviors. For an order or service provider in a certain (simulated) state, the simulation kernel 200 may call the prediction model corresponding to the state to predict whether the service requester of the order will perform a behavior that is time-related to the state. For example, refer to Figure 8 For orders (or passengers) in the "answer" state, the simulation kernel 215 can call the prediction model corresponding to the "answer" state / cancel order behavior to predict whether the driver (or the passenger) will cancel the answer.

[0090] In some embodiments, the simulation kernel 215 calls the behavior simulation model to process the user's feature data to obtain a behavior simulation result. It is understood that each type of behavior can correspond to a behavior prediction model to obtain a prediction result of the type of behavior. Each prediction model can be a trained machine learning model, including but not limited to a decision tree, a support vector machine, a Bayesian model, and a neural network.

[0091] In some embodiments, the simulation kernel 215 can dynamically update the user's characteristic data based on the user's simulation status, for example, the number of times the driver completed / cancelled orders in the last month, and the number of times the passenger completed / cancelled orders in the last month.

[0092] The behavior simulation result may include one or more of information such as whether the behavior occurs, the time when the behavior occurs (such as the pick-up time), and the location where the behavior occurs (such as the pick-up location). For example, for the behavior of canceling an order, the behavior simulation result may indicate whether the user (such as the driver or the passenger) will cancel the order. When the behavior simulation result indicates that the user (such as the driver or the passenger) will cancel the order, the behavior simulation result may also include the predicted time of canceling the order.

[0093] Step 324, determine the order simulation allocation result according to the target order solution.

[0094] The order-making scheme takes the state data as input and outputs the matching result of the order and the service provider (i.e., the order allocation result). As an example, the order-making scheme (such as the target order-making scheme) can obtain the orders waiting for response (referred to as waiting for response orders) and the idle service providers (such as drivers) as input, and output the order allocation result (such as the driver-passenger matching result).

[0095] Fig. 9 An exemplary driver-passenger order-making scheme is shown. The order-making scheme includes a recall layer, a scoring layer, and a matching layer. The recall layer includes a basic recall scheme and a platform filtering model recall scheme. The basic recall scheme may include a vehicle type filtering scheme (filtering drivers of vehicles that are not selected by passengers), an order radius filtering scheme (filtering drivers who are not within the order radius), etc. The platform filtering model recall scheme is a supplement to the basic recall scheme. Unfiltered driver-passenger matching relationships will enter the scoring layer, which includes a scoring scheme (a passenger perspective scoring scheme and a driver perspective scoring scheme) and a scoring adjustment scheme (a passenger rights scoring adjustment scheme and a driver rights scoring adjustment scheme) under the optimal driver-passenger scheme. The scoring output by the scoring layer is sent to the matching layer. The matching layer executes the optimal driver-passenger matching scheme based on the scoring situation and outputs the driver-passenger matching result, for example, assigning the order to the driver with the highest score.

[0096] The order allocation result affects the state transition during the order process. Taking online ride-hailing orders as an example, the order allocation result means that the status of successfully matched orders and drivers will be updated to the "answered" state.

[0097] Step 326: based on the behavior simulation results and the order simulation allocation results, update the simulation status of each order and each service provider in the target space-time unit.

[0098] Order allocation (execution of the target order-making plan) can be regarded as the behavior of the matchmaker (server 120). The behavior of the supply and demand parties and the behavior of the matchmaker jointly affect the state transfer of the supply and demand parties (such as the driver and passenger state). Therefore, based on the behavior simulation results and the order simulation allocation results, the simulation kernel 215 can update the simulation state of each order and each service provider in the target space-time unit. In some embodiments, the simulation kernel 215 updates the simulation state of each order and each service provider in the target space-time unit through a state machine. The state machine is composed of a state register and a combinational logic circuit, and can perform state transfer according to a pre-set state based on a control signal. It is a control center that coordinates related signal actions and completes specific operations. In step 326, the state machine takes the behavior simulation results and the order simulation allocation results as input.

[0099] refer to Figure 6 or Figure 7 The updated simulation state can be fed back to the subsequent behavior simulation and order allocation process. Specifically, the simulation kernel 215 can determine the set of orders waiting for response and the set of idle service providers (such as the set of drivers) according to the updated simulation state to continue the order allocation. The simulation kernel 215 can also predict the behavior associated with the state (such as canceling the behavior after the response) according to the updated simulation state (such as the "response" state) to continue the behavior simulation.

[0100] In some embodiments, the time range of the target spatiotemporal unit is equally divided into multiple time windows (for example, each time window is 2 seconds), and the execution of each simulation task includes the simulation state update of multiple time windows (or steps). In each time window, the simulation kernel 215 can perform a behavior simulation and an order matching (allocation). This specification does not specifically limit the priority between the behavior simulation and the order matching in each time window, as long as the priority between the behavior simulation and the order matching is consistent in the multiple time windows. For example, in each time window, the simulation kernel 215 can first perform a driver-passenger behavior simulation, update the driver-passenger simulation state based on the behavior simulation result, and then output the driver-passenger matching result based on the updated driver-passenger simulation state and the target order-making plan, and update the driver-passenger state again based on the driver-passenger matching result. For another example, in each time window, the simulation kernel 215 can first perform an order allocation, update the driver-passenger simulation state based on the order allocation result, and then perform a behavior simulation based on the updated driver-passenger simulation state, and update the driver-passenger simulation state again based on the behavior simulation result.

[0101] For the engineering architecture and working principle of the simulation kernel 215, please refer to Fig.10 and its related description.

[0102] Step 330, based on the order simulation result, evaluate the target order solution. In some embodiments, step 330 is performed by the evaluation module 230. In alternative embodiments, manual evaluation may be adopted.

[0103] The single simulation results may include any data generated during the single simulation process, such as behavioral simulation results and state transfer processes.

[0104] The evaluation module 230 can determine the evaluation indicators of the target order-making solution based on the simulation requirements, for example, order-making efficiency indicators such as response rate, transaction rate, GMV (transaction amount), and pick-up distance. Furthermore, the evaluation module 230 can calculate the evaluation value of the order-making efficiency of the target order-making solution based on the order-making simulation results. Taking the transaction rate as an example, the evaluation module 230 can extract all completed orders (with the termination status of "order completed") from the order-making simulation results, and then calculate the evaluation value of the transaction rate.

[0105] In some embodiments, the evaluation module 230 evaluates the order efficiency of the target order solution based on the order simulation result to obtain an evaluation result. Furthermore, the evaluation module 230 can determine whether the evaluation result meets the preset requirements, for example, whether the evaluation result shows that the target order solution can significantly improve the order efficiency (such as the transaction rate). In response to determining that the evaluation result meets the preset requirements, the evaluation module 230 can apply the target order solution to online-to-offline services (such as online car-hailing services), that is, launch the target order solution in a real environment. In response to determining that the evaluation result does not meet the preset requirements, the simulation system 200 can obtain a new (for example, updated) target order solution, re-execute the order simulation, and obtain a new evaluation result. The above process can be repeatedly performed until the simulation system 200 determines that the evaluation result meets the preset requirements of the target order solution, and launches the target order solution finally obtained in a real environment.

[0106] In some embodiments, the performance of the simulation system 200 may be evaluated. For the evaluation of the performance of the simulation system, please refer to Fig.11 and its related description.

[0107] like Figure 7 As shown, the evaluation module 230 can evaluate the effect of the order-making scheme under the target supply and demand scenario in multiple supply and demand scenarios and output the evaluation result. Fig.12 and its related description.

[0108] Fig.10 is an exemplary engineering architecture diagram of the simulation system. The engineering architecture construction of the simulation system 200 is the basis for behavioral simulation and order solution execution, which may include distributed system architecture construction, model automatic scaling construction, online order solution synchronization mechanism construction, performance stability guarantee mechanism construction and simulation platform construction. In this way, the speed and stability of the simulation system 200 can be greatly improved. Fig.10 As shown, the architecture (simulation system 200) includes a task management layer, a task execution layer, a state promotion layer, a single solution layer and a data storage layer (database 205).

[0109] The task management layer (manager) is responsible for the background management and maintenance of the simulation system 200. The task management layer provides external communication interfaces (such as http interfaces) to support functions such as the submission of single solutions, the submission or suspension of simulation tasks, and data query. The task management layer provides internal thrift interfaces for the task execution layer to obtain the task units to be executed and synchronize the task execution status.

[0110] The task execution layer (worker) is responsible for the computation control, scheduling and runtime data management of a single simulation task. The computation control of simulation tasks refers to obtaining the simulation tasks to be calculated from the task management layer and executing the simulation tasks to be calculated, regularly synchronizing the task execution status to the task manager, and receiving task pause signals. The scheduling execution of simulation tasks refers to controlling the advancement of the time window of the simulation task, and is responsible for calling the state advancement layer and the single solution layer to process the simulation calculation within each time window, and recording the calculation results; the runtime data management of simulation tasks refers to maintaining the runtime data of the simulation task, and regularly persisting the data for breakpoint continuation.

[0111] The state transfer layer (transfer) is responsible for the simulation state transfer of orders and service providers within each time window of the simulation task. Based on the behavior simulation results (such as model prediction results) and order allocation results, the state transfer layer completes the simulation state update of orders and service providers. The state transfer layer can be implemented based on a state machine.

[0112] The order-making scheme layer includes a unified allocation layer (matcher) and a scheduling decision module (master). The unified allocation layer is responsible for the order allocation (referred to as order splitting) within a single time window of the simulation task. The order-making scheme layer accepts the call of the task execution layer. The order-making scheme layer completes the supply and demand allocation calculation of a single step based on the set of orders waiting for response and the set of idle service providers (such as the set of drivers). The filtering module filters and recalls the order set and the set of service providers (such as the set of drivers). The scheduling decision module is responsible for providing decisions about scheduling service providers (such as drivers) to coordinate the supply and demand relationship.

[0113] The unified allocation layer includes the filtering module (cmr) and the scene module (scene).

[0114] The scene module includes sub-modules such as stg and pqueue. The scene module is used to support the expansion of order splitting scenarios.

[0115] The data storage layer (database), also known as the unified storage layer, is responsible for storing the data that the simulation task depends on (such as scene data, model data, and baseline task data) and the simulation calculation data of each step (such as behavior simulation results and simulation status).

[0116] The services that the simulation system 200 relies on may include bridger (bridge service), model (model service), redis (remote dictionary service), mysql (a relational database service) and vrp (universal routing service).

[0117] Fig.11 An exemplary simulation system performance evaluation system is shown. Fig.11As shown, the performance evaluation items of the simulation system 200 include accuracy, speed, ease of use and stability, and these evaluation items can be used to guide the iterative optimization work of the database 205 (such as hyperparameter data) and the simulation system 200 (such as the simulation kernel 215).

[0118] Accuracy includes true restoration degree and deduction credibility. The true restoration degree is used to evaluate the restoration degree of the order efficiency (such as response rate, transaction rate, GMV, pick-up distance) of the same order-making plan in a simulation environment relative to the order efficiency in a real environment (such as the ratio between the two). The deduction credibility is used to evaluate the difference between the change in the order efficiency of the same order-making plan in a simulation environment and the change in the order efficiency in a real environment. For example, the target order-making plan increases the response rate by 1pp in a simulation environment, but the target order-making plan decreases the response rate by 1pp in a real environment, indicating that the deduction credibility is poor.

[0119] The speed includes single task time, data import time and evaluation time. Among them, single task time refers to the time taken to execute a simulation task, data import time refers to the time taken to complete the data import work, and evaluation time refers to the time taken to complete a single solution evaluation task (such as executing process 1200 once).

[0120] Usability includes system usage cost and problem location cost. The system usage cost refers to the learning cost of relevant personnel (such as R&D personnel) using the simulation system 200 to perform order simulation and order solution evaluation, such as familiarity with the simulation process and development of order solutions. Problem location cost refers to the cost of locating problems when a simulation task fails or the evaluation result is illogical (for example, not in line with experience or preset rules).

[0121] Stability includes the execution success rate, which refers to the proportion of simulation tasks that are successfully executed and the simulation results that meet expectations (such as the transaction rate meets the target).

[0122] Fig.12 1200 is an exemplary flow chart of a method for evaluating a target ordering solution in a target supply and demand scenario according to some embodiments of this specification. Process 1200 is an evaluation method based on a t-test, and the evaluation module 230 calculates the confidence evaluation result of the target ordering solution based on the results of multiple simulation tasks. Fig.12 As shown, process 1200 includes the following steps.

[0123] Step 1210, determine simulation requirements.

[0124] The simulation requirements may include the target supply and demand scenario (such as scenario 0), order efficiency indicators (such as transaction rate), type I error threshold (such as α = 0.05), type II error threshold (such as β = 0.8), and the minimum expected improvement in order efficiency indicators (such as an increase in transaction rate of at least 0.5%). A type I error is the incorrect rejection of a correct null hypothesis, also known as a false positive error. A type II error is the incorrect failure to reject an incorrect null hypothesis, also known as a false negative error.

[0125] Step 1220, determine the simulation accuracy level.

[0126] The simulation system 200 samples N samples in the target supply and demand scenario to perform the baseline task. After completing the baseline task, the simulation system 200 calculates the simulation accuracy error distribution of the order efficiency of the baseline task. The distribution is generally a normal distribution with a mean near 0 and a standard deviation denoted as δ. The simulation accuracy level is the standard deviation δ of the error distribution.

[0127] Step 1230, determine the minimum sample size required for single solution evaluation.

[0128] Assuming that the order efficiency error of the baseline task (used to calibrate the hyperparameter data) is identically distributed with the order efficiency error of the simulation task (referred to as the evaluation task) used to evaluate the order solution, the minimum sample size (denoted as M) is calculated as follows:

[0129] Among them, Q α is the percentile function value of the first-class error threshold α, Q β is the percentile function value of the type II error threshold β, δ comes from step 1220, MDE (Minimum Detectable Effect) comes from step 120, It is the mean of the order efficiency index in the baseline task (the mean of the order efficiency of N samples).

[0130] Step 1240 , perform significance assessment based on the minimum sample size.

[0131] The simulation system 200 samples M samples in the target supply and demand scenario and obtains a significance evaluation result. The significance evaluation result indicates whether the target order-making solution can produce a significant impact, for example, whether it can significantly improve the transaction rate. The significance evaluation result includes a t value and a p value. The calculation method of the t value is as follows:

[0132] in, is the mean of the order efficiency index in the evaluation task. Based on the t value, the evaluation module 230 can calculate the p value. In some embodiments, when the p value is less than a preset value (such as 0.05), it can be considered that the target order solution can have a significant impact on the order efficiency index. Further, when Greater than When the p value is less than the preset value, it can be considered that the target order-making scheme can significantly improve the order-making efficiency index; otherwise, it can be considered that the target order-making scheme can significantly reduce the order-making efficiency index. Greater than And when the p value is less than the preset value (ie, the evaluation result meets the preset requirements), the evaluation module 230 can apply the target order-making solution to the online-to-offline service.

[0133] It should be noted that the above description of the relevant process is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the process under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0134] The beneficial effects that may be brought about by the embodiments of this specification include but are not limited to: (1) Through behavioral simulation and order-making scheme reproduction, a simulation method and system based on supply and demand order-making is provided, which can promote the iterative optimization of order-making schemes at low cost, low risk, high efficiency and high quality; (2) It is proposed to perform order-making simulation in local time and space, and on the premise of ensuring the effective evaluation of order-making schemes, the amount of data relied on by order-making simulation is greatly reduced, thereby greatly reducing the time consumption of order-making simulation; (3) On the one hand, the database provides a standard scenario data set for the simulation system, which is convenient for evaluating the effects of different order-making schemes in a unified test environment; on the other hand, by pre-completing the data tasks that the simulation tasks depend on (including the preparation of model data and baseline task data), the order-making scheme evaluation process of the simulation system is simplified, and the simulation speed and ease of use of the simulation system are improved. It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

[0135] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation on the embodiments of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to the embodiments of this specification. Such modifications, improvements and corrections are suggested in the embodiments of this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0136] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment" and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different locations in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0137] In addition, it will be appreciated by those skilled in the art that the various aspects of the embodiments of this specification may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, the various aspects of the embodiments of this specification may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, the various aspects of the embodiments of this specification may be represented as a computer product located in one or more computer-readable media, the product including computer-readable program code.

[0138] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0139] The computer program coding required for the operation of each part of the embodiment of this specification can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, VisualBasic, Fortran2003, Perl, COBOL2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program coding can be run completely on the user's computer, or run on the user's computer as an independent software package, or partly run on the user's computer and partly run on the remote computer, or run completely on the remote computer or processing equipment. In the latter case, the remote computer can be connected to the user's computer by any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., by the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0140] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in the embodiments of this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of the embodiments of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing processing device or mobile device.

[0141] Similarly, it should be noted that in order to simplify the description of the embodiments disclosed in this specification, thereby helping to understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the objects of the embodiments of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0142] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification is hereby incorporated by reference in its entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this application (currently or later attached to this application) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0143] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of the embodiments of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A simulation method based on supply and demand ordering, characterized in that: include: Obtain scene data of the target spatiotemporal unit; Performing a single simulation based on the scene data of the target spatiotemporal unit to obtain a single simulation result, wherein the single simulation includes: Simulate the behavior of each service requester and each service provider in the target space-time unit to obtain a behavior simulation result, wherein the scene data of the target space-time unit is used to initialize the simulation state of each order and each service provider in the target space-time unit; Determine the order simulation allocation results according to the target order plan; and Based on the behavior simulation results and the order simulation allocation results, the simulation status of each order and each service provider in the target space-time unit is updated, and the scene data of the target space-time unit is used to initialize the simulation status of each order and each service provider in the target space-time unit.

2. The method according to claim 1, characterized in that The step of acquiring scene data of a target spatiotemporal unit includes: The scene data of the target space-time unit is read from a database, where the database stores scene data of a plurality of space-time units.

3. The method according to claim 2, characterized in that The scenario data includes snapshot data and event data. The snapshot data of each space-time unit reflects the real state of each order and each service provider at the initial moment of the space-time unit, and the event data of each space-time unit reflects the real behavior of each order and each service provider in the space-time unit. The scenario data of each space-time unit is used to initialize the simulation state of each order and each service provider in the space-time unit. The database also stores model data and hyperparameter data, and the method further includes: acquiring the model data and the hyperparameter data to perform a single simulation based on the scene data of the target spatiotemporal unit, the model data and the hyperparameter data; The model data includes a behavior simulation model and feature data, the feature data reflects the features of each service requester and each service provider, and the behavior simulation model is used to simulate the behavior of each service requester and each service provider based on the feature data; the hyperparameter data includes the hyperparameters of the behavior simulation model and the hyperparameters of the order-making plan.

4. The method according to claim 3, characterized in that The hyperparameters are calibrated by narrowing the gap between the estimated value and the true value of the order efficiency of the historical order-making scheme.

5. The method according to claim 2, characterized in that The multiple space-time units correspond to multiple target space fences in the spatial dimension. The multiple target space fences are obtained by performing aggregation processing on multiple initial space fences based on region similarity, and the response closure of the multiple target space fences meets the set requirements. The response closure is the proportion of responded orders whose departure location of the order and the response location of the service provider are within the same target space fence to all responded orders.

6. The method according to claim 2, characterized in that The plurality of space-time units are divided into a plurality of supply and demand scenarios; the step of extracting the scenario data of the target space-time unit from the database includes: Determining a target supply and demand scenario among the multiple supply and demand scenarios according to the simulation requirements; From the scene data of the multiple space-time units, the scene data of the space-time unit belonging to the target supply and demand scene is determined as the scene data of the target space-time unit.

7. The method according to claim 1, characterized in that The method further comprises: Evaluate the order efficiency of the target order solution based on the order simulation result to obtain an evaluation result; Determining whether the evaluation result meets the preset requirements; In response to determining that the evaluation result meets preset requirements, the target order-making solution is applied to the online-to-offline service.

8. The method according to claim 1, characterized in that The service requester is a passenger, and the service provider is a driver.

9. A simulation device based on supply and demand ordering, characterized in that: It comprises a processor and a storage device, wherein the storage device is used to store instructions. When the processor executes the instructions, the simulation method based on supply and demand ordering as described in any one of claims 1 to 8 is implemented.

10. A simulation system based on supply and demand ordering, characterized in that: include: A reading module is used to obtain scene data of a target spatiotemporal unit; An execution module is used to perform a single simulation based on the scene data of the target spatiotemporal unit to obtain a single simulation result, wherein the single simulation includes: Simulate the behavior of each service requester and each service provider in the target space-time unit to obtain a behavior simulation result, wherein the scene data of the target space-time unit is used to initialize the simulation state of each order and each service provider in the target space-time unit; Determine the order simulation allocation result order plan based on the target order plan; Furthermore, based on the behavior simulation results and the order simulation allocation results, the simulation status of each order and each service provider in the target space-time unit is updated, and the scene data of the target space-time unit is used to initialize the simulation status of each order and each service provider in the target space-time unit.

Images