Data processing method and device, equipment and storage medium

Abstract

The application discloses a data processing method and device, equipment and a storage medium, and relates to the technical field of data processing. The method comprises the following steps: acquiring a first observable object and a second observable object in a target region and a target period; acquiring a real target place of the first observable object in the target period according to target data of the first observable object in the target period; performing target place speculation on the second observable object in the target period to obtain a speculated target place of the second observable object in the target period; and performing state updating on each object in the target region and the target period based on the real target place or the speculated target place of each observable object to obtain a full-quantity object set after state updating. The application can be applied to scenes such as artificial intelligence, intelligent transportation and auxiliary driving, and can realize automatic state updating of full-quantity objects in the target region and the target period, thereby improving the state updating efficiency and delicacy of the objects.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data processing method, apparatus, device, and storage medium. Background Technology

[0002] Currently, the dynamic updates of vehicles play a crucial role in providing data support for traffic management.

[0003] However, the vehicle dynamic update methods provided by related technologies require a lot of manpower, material resources, and costs to collect and update vehicle dynamics. The vehicle dynamic update is very time-consuming and labor-intensive, and the update cycle is long. Summary of the Invention

[0004] This application provides a data processing method, apparatus, device, and storage medium that can automatically update the status of all objects within a target area and a target time period, and shorten the update cycle of the status of all objects to the daily (or weekly) level, thereby improving the efficiency and granularity of object status updates. The technical solution may include the following.

[0005] According to one aspect of the embodiments of this application, a data processing method is provided, the method comprising:

[0006] Obtain the set of observable objects within the target area and target time period;

[0007] From the set of observable objects, a first set of observable objects and a second set of observable objects are determined. Each first observable object in the first set of observable objects corresponds to target data within the target time period, and each second observable object in the second set of observable objects does not have target data within the target time period.

[0008] Based on the target data within the target time period corresponding to each of the first observable objects, the actual target location within the target time period corresponding to each of the first observable objects is obtained;

[0009] For each of the second observable objects, the target location within the target time period is inferred to obtain the inferred target location within the target time period corresponding to each of the second observable objects;

[0010] Based on the actual target locations within the target time period corresponding to each of the first observable objects, and the inferred target locations within the target time period corresponding to each of the second observable objects, the state of each object within the target area and the target time period is updated to obtain a full set of objects with updated state; wherein, the full set of objects with updated state includes the observable object set.

[0011] According to one aspect of the embodiments of this application, a data processing apparatus is provided, the apparatus comprising:

[0012] The observable object acquisition module is used to acquire a set of observable objects within a target area and a target time period;

[0013] An observable object segmentation module is used to determine a first observable object set and a second observable object set from the observable object set. Each first observable object in the first observable object set corresponds to target data within the target time period, and each second observable object in the second observable object set does not have target data within the target time period.

[0014] The real target location acquisition module is used to acquire the real target location within the target time period corresponding to each of the first observable objects based on the target data within the target time period corresponding to each of the first observable objects.

[0015] The target location acquisition module is used to infer the target location of each of the second observable objects within the target time period, and obtain the inferred target location of each of the second observable objects within the target time period respectively;

[0016] The full state update module is used to update the state of each object within the target area and the target time period based on the actual target location corresponding to each of the first observable objects within the target time period and the inferred target location corresponding to each of the second observable objects within the target time period, so as to obtain a full set of objects with updated state; wherein, the full set of objects with updated state includes the observable object set.

[0017] According to one aspect of the embodiments of this application, a computer device is provided, the computer device including a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the above-described data processing method.

[0018] According to one aspect of the present application, a computer-readable storage medium is provided, wherein a computer program is stored in the storage medium, the computer program being loaded and executed by a processor to implement the above-described data processing method.

[0019] According to one aspect of the embodiments of this application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the data processing method described above.

[0020] The technical solutions provided in this application embodiment may have the following beneficial effects:

[0021] By inferring the target location within a target time period for observable objects without target data, and updating the status of all objects within the target area and target time period based on the inferred or actual target location corresponding to each observable object, the system achieves automatic status updates for all objects within the target area and target time period. This shortens the update cycle of the status of all objects to the daily or weekly level, thereby improving the efficiency and granularity of dynamic object updates. This expands the applicable scenarios for dynamic object statistics and enhances business statistical capabilities. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the implementation environment of a solution provided in one embodiment of this application;

[0024] Figure 2 This is a flowchart of a data processing method provided in one embodiment of this application;

[0025] Figure 3 This is a flowchart of a method for obtaining a presumed target location according to an embodiment of this application;

[0026] Figure 4 This is a flowchart of a method for obtaining a presumed target location provided in another embodiment of this application;

[0027] Figure 5 This is a schematic diagram of an agent model provided in one embodiment of this application;

[0028] Figure 6 This is a schematic diagram of an agent provided in one embodiment of this application;

[0029] Figure 7This is a schematic diagram of a method for obtaining a dwell point according to an embodiment of this application;

[0030] Figure 8 This is a schematic diagram comparing the related technologies and the effects of one embodiment of this application;

[0031] Figure 9 This is a block diagram of a data processing apparatus provided in one embodiment of this application;

[0032] Figure 10 This is a block diagram of a data processing apparatus provided in another embodiment of this application;

[0033] Figure 11 This is a block diagram of a computer device provided in one embodiment of this application. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0035] Please refer to Figure 1 This diagram illustrates an implementation environment provided by one embodiment of the present application. This implementation environment can be implemented as a data processing system architecture. The implementation environment may include: terminal 10 and server 20.

[0036] Terminal 10 can be an electronic device such as a mobile phone, computer, smart voice interaction device, smart home appliance, multimedia playback device, PC (Personal Computer), or in-vehicle terminal. A client application for the target application can be installed on terminal 10. For example, the target application can be a vehicle dynamic statistics application, a map application, a navigation application, a traffic application, a simulation learning application, a data analysis application, etc.

[0037] Server 20 provides background services to the client of the target application (such as a vehicle dynamic statistics application) in terminal 10. For example, server 20 can be the background server of the aforementioned application (such as a vehicle dynamic statistics application). Server 20 can be a single server, a server cluster consisting of multiple servers, or a cloud computing service center.

[0038] Terminal 10 and server 20 can communicate with each other via network 30. This network 30 can be a wired network or a wireless network.

[0039] The technical solutions provided in this application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving. The technical solutions provided in this application can achieve automatic updates of the status of all objects (such as vehicles) within a target area (e.g., nationwide) and a target time period (e.g., daily), thus shortening the automatic update cycle of the status of all objects to the daily (or weekly) level.

[0040] In one example, taking a vehicle dynamic statistics application client as an example, server 20 obtains observable vehicles (i.e., observable objects) within a target area and a target time period, as well as target data (such as driving data) of the observable vehicles. For the first observable vehicle without target data, the server infers the target location (such as the place of stay) within the target time period, obtaining the inferred target location corresponding to the first observable vehicle. Then, based on the inferred target location corresponding to the first observable vehicle and the actual target location corresponding to the second observable vehicle with target data, the server updates the status of all vehicles within the target area and the target time period. Then, based on the vehicle status, inferred target location, and actual target location, the server updates business statistics data (such as place of stay, permanent residence, migration status, etc.). Finally, based on the updated set of all objects, the server controls, manages, or provides services to the objects. Optionally, the above process can also be completed by the client of the vehicle dynamic statistics application, and this embodiment of the application does not limit this.

[0041] Please refer to Figure 2 It illustrates a flowchart of a data processing method provided in an embodiment of this application, wherein the execution entity of each step of the method may be... Figure 1 The terminal 10 or server 20 in the implementation environment of the scheme shown may include the following steps (201-205).

[0042] Step 201: Obtain the set of observable objects within the target area and the target time period.

[0043] The target area can refer to a region that is set and adjusted according to actual usage needs. For example, the target area can refer to something like the whole country, province, city, county, district, town, township, or city. The target time period can refer to a time period that is set and adjusted according to actual usage needs. For example, the target time period can refer to something like an hour, day, week, month, or year.

[0044] In this application embodiment, an object can refer to any dynamically changing thing, such as a vehicle, mobile device, animal, or other target thing. An observable object refers to an object whose dynamic changes can be observed. For example, taking an observable vehicle as an example, an observable vehicle can be identified based on the vehicle's autonomously uploaded driving data, periodically uploaded driving data, statistical equipment (such as monitoring equipment), payment equipment, etc. For some observable objects, their corresponding target data may not be available within the target time period, such as when these observable objects do not upload target data within the target time period. The observable object set may include some objects within the target area and within the target time period. Target data can refer to things such as vehicle driving data, mobile device movement data, animal migration data, etc., and this application embodiment does not limit this.

[0045] For example, taking observable vehicles as the observable object, in step 201, all observable vehicles within the target area on a certain day can be obtained.

[0046] Step 202: From the set of observable objects, determine the first set of observable objects and the second set of observable objects. Each first observable object in the first set of observable objects corresponds to target data within the target time period, while each second observable object in the second set of observable objects does not have target data within the target time period.

[0047] Optionally, for a target observable object in the set of observable objects, if target data of the target observable object is detected within the target time period, the target observable object can be identified as the first observable object. If no target data of the target observable object is detected within the target time period, the target observable object can be identified as the second observable object.

[0048] Step 203: Based on the target data within the target time period corresponding to each first observable object, obtain the real target location within the target time period corresponding to each first observable object.

[0049] Optionally, the target data within the target time period may include the object's trajectory, location points, target points, and target locations within the target time period. The trajectory may consist of multiple location points, which can be acquired at fixed distances or fixed time intervals. Optionally, the target points and target locations in this embodiment can be set and adjusted according to actual usage needs. Taking an observable vehicle as an example, a target point can refer to a vehicle's stopping point, which can be a location point where the vehicle stays for more than a threshold duration. A target location can refer to the vehicle's stopping location, which is one type of stopping point. For example, if the target time period is a day, stopping points that meet a first condition during the night of that day (which can be extracted based on the day's driving data) can be determined as the stopping locations for that day. This first condition can be set according to actual usage needs; for example, if the object's stopping point does not change between midnight and 5 AM, then that stopping point can be determined as the stopping location for that day. This embodiment does not limit the method for obtaining stopping locations. When the target time period is a week, the stopping locations for each day can be obtained based on driving data.

[0050] Optionally, the true destination refers to the destination determined based on the target data. For example, if the target data shows that object A remains unchanged within region B between midnight and 5 a.m., then region B can be determined as the true destination of object A for that day.

[0051] It should be noted that the target data has been completely de-identified and anonymized. Service providers cannot use algorithms or technology to reconstruct the target data into specific real objects (such as vehicles) in any way. Service providers can only observe the completely de-identified virtual objects and target data, so there is no privacy risk.

[0052] Step 204: Infer the target location within the target time period for each second observable object, and obtain the inferred target location within the target time period for each second observable object.

[0053] The inferred destination can refer to a destination inferred from a set of historical trajectories. Each historical trajectory in this set can refer to a trajectory corresponding to a previous target time period. For example, taking a target time period as a day, the historical trajectory set can include historical trajectories from the previous 30 days. Optionally, the time span corresponding to the historical trajectory set can be set and adjusted according to actual usage needs; this embodiment does not limit this. If there is no corresponding trajectory for a certain historical target time period, the historical trajectory corresponding to that historical target time period can be set to empty.

[0054] In one example, reference Figure 3 Step 204 may include the following sub-steps.

[0055] Step 204a: For each target second observable object among the second observable objects, obtain the set of historical trajectories of the target second observable object within the historical time period; wherein, the historical time period refers to the time period before the target time period, and each historical trajectory in the set of historical trajectories is composed of location points.

[0056] The target second observable object can refer to any one of the various second observable objects. The historical time period can include multiple historical target time periods, and the historical trajectory set can include the historical trajectories corresponding to each of the multiple historical target time periods.

[0057] For example, refer to Figure 4 After obtaining target data (such as driving logs) of the second observable object in the historical period, multiple historical trajectories 401 are extracted from the target data.

[0058] Step 204b: Aggregate the various historical trajectories to obtain the aggregated trajectory.

[0059] Optionally, each location point in the historical trajectory corresponds to both temporal and spatial information. Location points in various historical trajectories can be aggregated based on the spatial information. For example, refer to... Figure 4 Based on spatial information, multiple historical trajectories 401 are aggregated to obtain the aggregated trajectory 402.

[0060] The location points in a single historical trajectory are relatively sparse. The density of location points in a historical trajectory can be increased by aggregating the location points from multiple historical trajectories.

[0061] Step 204c: Extract at least one target point from the aggregated trajectory.

[0062] From the aggregated trajectory, clusters of frequently occurring location points can be extracted, and the regions corresponding to these location point clusters can be identified as target points (multiple target points are possible). This application does not limit the method for extracting target points; for example, clustering methods such as DBSCAN (Density-based Spatial Clustering of Applications with Noise, a classic density-based clustering analysis algorithm) and K-means can be used to extract target points. For example, refer to... Figure 4 From the aggregated trajectory 402, two target points 403 can be extracted.

[0063] In one example, taking the target time period as days, the process of extracting target points can be described as follows:

[0064]

[0065] in, This is used to characterize the location coordinates and time of each target point extracted from the second observable object u over the past n days (the target points can be sorted by time). Let be the historical trajectory of the second observable object u on the nth day in the past. Let represent the high-frequency location clusters extracted after the historical trajectory is aggregated and used as the target points.

[0066] Step 204d: Based on at least one target point, obtain the inferred target location within the target time period corresponding to the second observable object of the target.

[0067] Optionally, a multi-head attention mechanism can be used to infer the target location within a target time period based on at least one target point. The specific details can be as follows:

[0068] 1. Map at least one target point to obtain a mapping matrix corresponding to at least one target point. The mapping matrix includes a time matrix in the time dimension and a space matrix in the space dimension.

[0069] The mapping matrix is ​​used to represent the aggregation pattern of at least one target point. The time matrix is ​​used to represent the temporal information corresponding to at least one target point, and the spatial matrix is ​​used to represent the spatial information corresponding to at least one target point.

[0070] For example, refer to Figure 4 Mapping the two target points 403 yields a mapping matrix 404. This mapping method can be as follows: Based on the above embodiment, it can be... Spacetime mapping to (That is, the mapping matrix of the second observable object u at n target points, where t is the time information of the target points), and the specific mapping method can be as follows:

[0071]

[0072] Among them, spatial dimension mapping: Given a trainable d-dimensional vector, the positions of all target points in the spatial matrix can be obtained using... Let L represent the number of positions.

[0073] Time dimension mapping: Here, i represents the i-th dimension, and the time dimension and the space dimension are the same, both being d-dimensional vectors.

[0074] 2. For at least one object feature of the target second observable object, obtain the predicted target point corresponding to the target object feature based on the mapping matrix corresponding to at least one target point.

[0075] Object features can refer to the basic characteristics of an object. Taking an observable vehicle as an example, object features can refer to things like the vehicle's driving patterns, driving purpose, driving preferences, and driving environment. This application does not limit the scope of object features. Target object features can refer to any one of at least one object feature. Predicted target points are used to characterize possible target locations under a certain object feature.

[0076] In one example, the process of obtaining the predicted target point can be as follows: Obtain a set of historical target points, where each historical target point has a corresponding key and value; based on the mapping matrix corresponding to at least one target point and the keys corresponding to each historical target point, obtain a similarity measure between the mapping matrix and each historical target point; scale and standardize the similarity measure between the mapping matrix and each historical target point to obtain the weights corresponding to each historical target point; based on the weights corresponding to each historical target point, fuse the values ​​corresponding to each historical target point to obtain the predicted target point corresponding to the target object features.

[0077] The historical target point set can consist of historical target points corresponding to multiple historical target time periods. The key of each historical target point is used to uniquely identify it, and the value of each historical target point refers to its location coordinates and time. Optionally, the historical target point set can refer to at least one of the aforementioned target points, or it can include at least one of the aforementioned target points; this embodiment of the application does not limit this.

[0078] For example, a similarity measure can be represented as follows:

[0079]

[0080] Where h represents the object characteristic h, for and Similarity measure between For the mapping matrix, Used to represent the k-th historical target point, where p is the number of target points. and The weight parameters under the object feature h (can be obtained through training), Q is used to identify the corresponding mapping matrix, and K is used to identify the key of the corresponding historical target point.

[0081] The weights can be represented as follows:

[0082]

[0083] in, Let T be the weight of the k-th historical target point under object feature h, and T be the number of historical target points. Used to characterize the g-th historical target point.

[0084] Therefore, the predicted target point can be represented as follows:

[0085] in, For the predicted target point under object feature h, The weight parameters (which can be obtained through training) are given by the object feature h, and V is used to identify the value of the corresponding historical target point.

[0086] The essence of the above process is to use historical target points as key-value stores for training, then use the aggregated target points as the input of the query (question), determine the corresponding weights based on the similarity measure between the query and the key, and finally perform weighted fusion based on the weights to obtain the predicted target points.

[0087] 3. Overlay the predicted target points corresponding to at least one object feature to obtain the overlaid predicted target points.

[0088] Alternatively, the process can be represented as follows:

[0089]

[0090] in, H represents the number of object features. For any object feature, the above method can be used to obtain the predicted target point under that object feature.

[0091] 4. Based on the superimposed predicted target point and the first target point, the inferred target location within the target time period corresponding to the second observable object of the target is obtained; where the first target point refers to the last target point before the end time of the target time period.

[0092] Alternatively, the process can be represented as follows:

[0093]

[0094] in, Here, the inferred target location within the target time period corresponding to the second observable object u is... The first target point is represented by the trainable weight parameters.

[0095] Alternatively, other Transformer-type machine learning models can be used to infer the target location within the target time period based on at least one target point. This application does not limit this approach.

[0096] Thus, by utilizing the multi-head attention mechanism, the target location can be inferred within the target time period based on the aggregated target points, which can effectively avoid the interference of location points on the target points, thereby improving the accuracy of target location inference. At the same time, since the target location inference within the target time period only needs to be based on the aggregated target points, instead of inferring the target location within the target time period based on all location points, the computational load in the target location inference process is reduced, thereby improving the efficiency of target location inference.

[0097] Step 205: Based on the actual target location within the target time period corresponding to each of the first observable objects and the inferred target location within the target time period corresponding to each of the second observable objects, update the state of each object within the target area and the target time period to obtain the full set of objects after state update; wherein, the full set of objects after state update includes the observable object set.

[0098] The full object set can include all objects within the target region and target time period, encompassing both observable and unobservable objects. Unobservable objects refer to those not observed within the target region and target time period. For each object in the updated full object set, its state is the latest state within the target time period.

[0099] In one example, the agent model can be used to update the state of all objects based on observable objects. Specifically, this can be achieved as follows:

[0100] 1. Identify target observables from each first observable and each second observable. Each target observable represents a class of observables, and each class of observables has similar or identical object characteristics.

[0101] Based on the target data (including historical target data) of each observable object, observable objects can be classified into multiple categories. For example, observable vehicles can be classified according to driving patterns, driving trajectories, driving preferences, and driving environments. Optionally, the most representative observable object in each category (such as the most active observable object) can be identified as the target observable object.

[0102] 2. Construct an agent model based on the target observable object; the agent model includes multiple agents, and each agent corresponds one-to-one with the target observable object.

[0103] Each agent corresponds to a class of observables. For example, reference Figure 5 The agent model 501 includes n agents, each agent corresponding to a class of observable objects.

[0104] 3. Obtain input data corresponding to multiple agents respectively; wherein, the input data includes at least one of the following: production location, permanent residence within the threshold period, actual or predicted target location within the target period, historical target location set, first target point, and number of representative objects.

[0105] Optionally, the input data may include the actual or predicted target locations within the target time period. The input data may also include a set of actual or predicted target locations and historical target locations within the target time period. The input data may be increased or decreased as appropriate according to the richness of the target data or business needs (such as adding images of the object). This application embodiment does not limit this.

[0106] For vehicles, the place of origin can refer to the area where the vehicle was first activated, or the precise place of origin that can be returned to within the device. For animals, the place of origin can refer to the animal's birthplace.

[0107] The permanent residence can refer to the place where an object most frequently stays. For example, the permanent residence of a vehicle can be its garage, and the permanent residence of an animal can be its habitat. The threshold period can be set and adjusted according to actual usage needs, such as 3 months, 4 months, etc.

[0108] The set of historical destinations can include the historical destinations corresponding to each historical target time period before the target time period, which can be used for internal state updates of the agent model.

[0109] The number of objects represented refers to the number of objects each agent can represent. For example, an agent can represent 10 objects, and the number of objects represented can be set based on experience. In this way, by using virtual individuals to represent a group, the characteristics or states of all objects can be inferred based on observable objects.

[0110] For example, refer to Figure 6 The input data for agent 601 may include the production location, the permanent location within the threshold time period, the actual or predicted target location within the target time period, the set of historical target locations, the first target point, and the number of representative objects. If agent 601 has target data, the actual target location can be extracted from the target data. If agent 601 does not have target data, the predicted target location for agent 601 can be inferred through a multi-head attention mechanism.

[0111] 4. Based on the input data corresponding to each agent, update the status of each agent to obtain the agents with updated status.

[0112] Alternatively, a finite state machine can be used to update the state of multiple agents according to their respective input data, thus obtaining multiple agents with updated states.

[0113] For example, refer to Figure 6 Agent 601 can be modeled using a finite state machine, and a corresponding function model can be constructed to achieve the state update of agent 601.

[0114] 5. Based on the multiple proxies after the state update and the number of represented objects, obtain the full set of objects after the state update.

[0115] Optionally, the state corresponding to the agent can be determined as the state of each object under the number of objects represented by the agent, thereby realizing the state update of all objects.

[0116] In one example, the state of an object can include a new state, an active state, and a dead state. A new state is one in which the object has been continuously observed for a duration greater than a first threshold and less than or equal to a second threshold. An active state is one in which the object has been continuously observed for a duration greater than the second threshold. A dead state is one in which the object has not been observed for a duration greater than a third threshold.

[0117] The first, second, and third thresholds can be set and adjusted based on experience. For example, the first threshold can be set to 7 days, the second threshold to 30 days, and the third threshold to 6 months.

[0118] Optionally, the number of objects in the newly added state and the number of objects in the extinct state are used to update the total number of objects in the full set after the state update.

[0119] Since there is currently insufficient target data to determine the permanent location of newly added objects, they are not included in the classification statistics calculation to avoid affecting the accuracy of business statistics. For objects in the extinction state, they can be removed from the full object collection, and the number of representative objects corresponding to their respective agents will be updated simultaneously.

[0120] In some feasible embodiments, the state of an object may also include temporary state, permanent state, inflow state, outflow state, etc., and the state can be added or reduced according to actual usage requirements.

[0121] This application embodiment dynamically updates the total number of all objects by considering both new and dead states, thereby making the dynamic statistics of objects more accurate.

[0122] In one example, the target object set can be determined from the full set of objects after the status update, and each target object in the target object set is in an active state; based on the actual or inferred target location of each target object in the target area and target time period, the business statistics are updated to obtain the updated business statistics.

[0123] The updated business statistics can include changes in objects within the target area during the target period (e.g., daily) and during each historical target period prior to the target period, including resident objects, migrating objects, the origin and destination of migrating objects, and the total number of objects within the target area.

[0124] For example, refer to Figure 5 Taking a target time period of one day as an example, after obtaining the status of all objects within the target area through agent model 501, the target objects in the active state are obtained. These active target objects are then categorized and statistically analyzed to obtain the current predicted resident objects, predicted local objects, predicted expatriate objects, predicted inbound objects, predicted outbound objects, and predicted object changes for each sub-region within the target area. This enables dynamic and automatic updates of objects in each sub-region. The actual target location and the inferred target location can be used as underlying data to update the dynamics of all objects (such as resident location, migration dynamics, etc.).

[0125] In one example, after obtaining updated business statistics, the object can be controlled, managed, and serviced based on these statistics. For instance, in intelligent connected vehicle scenarios or unmanned vehicle intelligent logistics scenarios, vehicles can be controlled and managed based on the updated business statistics. This control and management can refer to scheduling vehicles within a target area, controlling vehicles to avoid congestion, and controlling vehicles to drive to or park at target locations. Services can also be provided to vehicles based on the updated business statistics, such as providing driving plans, congestion alerts, and stop recommendations; however, this embodiment does not limit this specific provision.

[0126] For example, the predicted number of migrating objects within the target area and in the next target time period can be determined based on updated business statistics. If the predicted number of migrating objects is greater than or equal to a fourth threshold, an object control instruction is generated. This instruction is used to control and manage the predicted migrating objects corresponding to the target area in the next target time period. The predicted number of migrating objects can be the sum of the predicted number of outgoing objects and the predicted number of embedded objects. The fourth threshold can be adaptively set and adjusted according to actual usage requirements.

[0127] For example, taking the target object as the target vehicle (i.e., the vehicle in an active state), for the target area A (such as a city, road, scenic spot, etc.), based on the updated business statistics data of the target area A for the day, we obtain the predicted resident vehicles, predicted inbound vehicles, and predicted outbound vehicles corresponding to the target area A for that day. By removing the predicted resident vehicles from the predicted inbound vehicles, we can predict the initial predicted outbound vehicles corresponding to the target area A for the next day. Then, by adjusting the weight parameters, we obtain the final predicted outbound vehicles corresponding to the target area A for the next day, as well as the number of the final predicted outbound vehicles.

[0128] By removing non-resident vehicles from the predicted outflow vehicles, we can predict the initial predicted inflow vehicles for target area A tomorrow. Then, by adjusting the weight parameters, we can obtain the final predicted inflow vehicles for target area A tomorrow, as well as the number of final predicted inflow vehicles. Finally, we sum the number of final predicted outflow vehicles and the number of final predicted inflow vehicles for tomorrow to obtain the number of predicted migration objects for target area A tomorrow.

[0129] If the number of predicted migrating vehicles corresponding to target area A for the next day is greater than or equal to the fourth threshold (such as traffic pressure, parking pressure, etc.), an object control instruction for target area A for the next day is generated. If target area A is congested, the object control instruction can control the predicted migrating vehicles to avoid target area A, control the predicted migrating vehicles to drive into newly opened roads, or control the predicted migrating vehicles to park in newly opened parking lots, etc.

[0130] Optionally, based on the updated business statistics, the predicted number of objects staying in the target area and the next target time period can be determined; if the predicted number of objects staying is greater than or equal to the fifth threshold, an object control strategy is generated, which is used to control and manage the stay of the predicted objects staying in the target area in the next target time period.

[0131] Optionally, the origin of each object within the target area and target time period can be determined based on the updated business statistics (such as business statistics corresponding to each historical target time period within and before the target time period).

[0132] This application embodiment supports the automatic updating of business statistics data within a target area and target time period, thereby shortening the update cycle of business statistics data to the daily or weekly level. This improves the update efficiency and granularity of object control instructions and object management strategies. Compared to the monthly or yearly level, the technical solution provided by this application embodiment can enhance the effectiveness and timeliness of object management, thereby expanding the applicable scenarios of the technical solution provided by this application embodiment and improving business capabilities.

[0133] In summary, the technical solution provided by this application, by inferring the target location within a target time period for observable objects without target data, and updating the status of all objects within the target area and target time period based on the inferred or actual target location corresponding to each observable object, achieves automatic status updates for all objects within the target area and target time period. This shortens the update cycle of the status of all objects to the daily or weekly level, thereby improving the efficiency and granularity of dynamic object updates, expanding the applicable scenarios for dynamic object statistics, and enhancing business statistical capabilities.

[0134] Furthermore, by utilizing a multi-head attention mechanism (i.e., multi-dimensional object features) to infer the destination within a target time period based on aggregated target points, the accuracy of destination inference can be improved. Additionally, since destination inference within a target time period is based on aggregated target points, interference from location points can be effectively avoided, further improving the accuracy of destination inference. Moreover, since destination inference within a target time period does not need to be based on all location points, the computational load in the destination inference process is reduced, thereby improving the efficiency of destination inference.

[0135] In addition, by using the agent model to update the state of all objects based on the state of observable objects, the rationality of state updates can be improved.

[0136] In an exemplary embodiment, taking observable objects as observable vehicles, target areas as region A, and target time periods as days as an example, the data processing method provided in this application embodiment is described, and its specific content can be as follows:

[0137] Retrieve the set of observable vehicles within region A for the current day. For example, retrieve all observable vehicles within each sub-region of region A for the current day, generating a set of observable vehicles.

[0138] A target observable vehicle set is determined from the observable vehicle set. Each target observable vehicle represents a class of observable vehicles, and each class of observable vehicles has similar or identical object characteristics. For example, vehicles with the same origin and destination can be classified into one class of observable vehicles, and vehicles with similar driving trajectories can be classified into another class of observable vehicles.

[0139] Based on the target observable vehicle set, an agent model is constructed. This agent model includes multiple agents, and each agent corresponds one-to-one with a target observable vehicle, that is, each agent represents a class of observable vehicles.

[0140] Obtain the input data for each agent. This input data may include the vehicle's (i.e., the agent's) place of manufacture, the vehicle's permanent residence (i.e., the garage) within the threshold time period, the vehicle's corresponding place of stay on the current day (i.e., the destination), the vehicle's corresponding set of historical places of stay, the vehicle's corresponding target place of stay (i.e., the target point), and the number of representative vehicles (i.e., the number of representative objects).

[0141] Optionally, if there is driving data (i.e. target data) for the agent on that day, the agent's place of stay for that day can be extracted from the driving data. If there is no driving data for the agent on that day, a multi-head attention mechanism can be used to infer the agent's predicted place of stay for that day.

[0142] For example, the target agent can be any agent who does not have driving data for that day. A set of historical trajectories for the target agent within a historical time period is obtained, which may include multiple historical trajectories from the previous 30 days. These historical trajectories are aggregated to obtain an aggregated trajectory. From the aggregated trajectory, at least one stop point is extracted.

[0143] For example, refer to Figure 7 , Figure 7 On the left is an aggregation of the vehicle's (agent's) historical trajectory over the past 30 days, with three distinct points of stoppage clearly visible. Figure 7 First, cluster the stop points on the right side to obtain stop point 701, stop point 702 and stop point 703, and then obtain the precise latitude and longitude coordinates and time corresponding to these three stop points respectively.

[0144] Obtain the mapping matrix corresponding to the three stops, and then use a multi-head attention mechanism to infer the agent's predicted stop location for the day based on the mapping matrix, multiple object features corresponding to the agent, and the historical stop point set corresponding to the agent.

[0145] Using a finite state machine, the state of each agent is updated based on the input data corresponding to each agent, resulting in the updated state of the agent. Then, based on the state of each agent and the data of the vehicles they represent, the state of all vehicles in region A for the day is updated.

[0146] Finally, based on the underlying data (such as the location of the vehicle on the day) corresponding to the active vehicle, business statistics are obtained, such as vehicle changes in region A on the day, resident and migrating vehicles in region A on the day, origin and destination of migrating vehicles on the day, and the total number of all vehicles.

[0147] For example, the daily vehicle migration platform launched based on the technical solution provided in the embodiments of this application can display the resident vehicles, moving-in vehicles, moving-out vehicles, and the origin of these three types of vehicles in the target area every day, which plays a very important data support role in vehicle management.

[0148] In one example, reference Figure 8 One month's worth of driving data from a specific region was used as the evaluation dataset. The DeepMove solution (a stop prediction method provided by related technologies) and the solution provided in the embodiments of this application were compared. Figure 8 As can be seen, the solution provided in this application significantly improves upon the computational load (only about 10% of the total load needs to be calculated) and update cycle (this application can achieve daily updates, while related technologies can only achieve monthly updates at the lowest level). Regarding the error distance for location estimation, the solution provided in this application also has good accuracy (the error distance of this application is about 472 meters, while the error distance of related technologies reaches about 1322 meters). Migrating vehicles are generally a group that actual business operations and customers pay more attention to, thus facilitating business expansion.

[0149] In summary, the technical solution provided in this application, by inferring the stopping locations of observable vehicles without driving data within a target time period, and updating the status of all vehicles within the target area and target time period based on the inferred or actual stopping locations corresponding to each observable vehicle, achieves automatic status updates for all vehicles within the target area and target time period. This shortens the update cycle for the status of all vehicles to the daily or weekly level, thereby improving the efficiency and granularity of dynamic vehicle updates, expanding the applicable scenarios for vehicle dynamic statistics, and enhancing business statistical capabilities.

[0150] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0151] refer to Figure 9 This diagram illustrates a block diagram of a data processing apparatus according to an embodiment of this application. The apparatus has the functionality to implement the method example described above; this functionality can be implemented in hardware or by hardware executing corresponding software. The apparatus can be the computer device described above, or it can be located within a computer device. For example... Figure 9 As shown, the device 900 includes: an observable object acquisition module 901, an observable object segmentation module 902, a real target location acquisition module 903, a predicted target location acquisition module 904, and a full state update module 905.

[0152] The observable object acquisition module 901 is used to acquire a set of observable objects within a target area and a target time period.

[0153] The observable object segmentation module 902 is used to determine a first observable object set and a second observable object set from the observable object set. Each first observable object in the first observable object set corresponds to target data within the target time period, and each second observable object in the second observable object set does not have target data within the target time period.

[0154] The real target location acquisition module 903 is used to acquire the real target location within the target time period corresponding to each of the first observable objects based on the target data within the target time period corresponding to each of the first observable objects.

[0155] The target location acquisition module 904 is used to infer the target location of each of the second observable objects within the target time period, and obtain the inferred target location of each of the second observable objects within the target time period.

[0156] The full state update module 905 is used to update the state of each object within the target area and the target time period based on the actual target location within the target time period corresponding to each of the first observable objects and the inferred target location within the target time period corresponding to each of the second observable objects, to obtain a full set of objects with updated state; wherein, the full set of objects with updated state includes the observable object set.

[0157] In one exemplary embodiment, the speculative target location acquisition module 904 is configured to:

[0158] For each of the target second observable objects, obtain the set of historical trajectories of the target second observable object within a historical time period; wherein, the historical time period refers to the time period before the target time period, and each historical trajectory in the set of historical trajectories is composed of location points;

[0159] The individual historical trajectories are aggregated to obtain the aggregated trajectory.

[0160] Extract at least one target point from the aggregated trajectory;

[0161] Based on the at least one target point, the inferred target location within the target time period corresponding to the second observable object of the target is obtained.

[0162] In one exemplary embodiment, the inferred target location acquisition module 904 is further configured to:

[0163] Mapping the at least one target point yields a mapping matrix corresponding to the at least one target point, the mapping matrix comprising a time matrix in the time dimension and a space matrix in the space dimension;

[0164] For at least one object feature of the target second observable object, the predicted target point corresponding to the target object feature is obtained based on the mapping matrix corresponding to the at least one target point.

[0165] The predicted target points corresponding to the at least one object feature are superimposed to obtain the superimposed predicted target points;

[0166] Based on the superimposed predicted target point and the first target point, the inferred target location within the target time period corresponding to the second observable object of the target is obtained; wherein, the first target point refers to the last target point before the end time of the target time period.

[0167] In one exemplary embodiment, the inferred target location acquisition module 904 is further configured to:

[0168] Obtain a set of historical target points, where each historical target point in the set corresponds to a key and a value;

[0169] Based on the mapping matrix corresponding to the at least one target point and the keys corresponding to each historical target point, a similarity measure between the mapping matrix and each historical target point is obtained.

[0170] The similarity measure between the mapping matrix and each historical target point is scaled and standardized to obtain the weights corresponding to each historical target point.

[0171] Based on the weights corresponding to each historical target point, the values ​​corresponding to each historical target point are fused to obtain the predicted target point corresponding to the target object feature.

[0172] In one exemplary embodiment, the full state update module 905 is configured to:

[0173] Target observable objects are determined from the first observable objects and the second observable objects, each target observable object representing a class of observable objects, and each class of observable objects having similar or identical object characteristics;

[0174] Based on the target observable object, an agent model is constructed; wherein, the agent model includes multiple agents, and each agent corresponds one-to-one with the target observable object;

[0175] Obtain input data corresponding to the plurality of agents respectively; wherein, the input data includes at least one of the following: production location, permanent residence location within the threshold period, actual or predicted target location within the target period, historical target location set, first target point, and number of representative objects;

[0176] Based on the input data corresponding to the multiple agents, the states of the multiple agents are updated respectively to obtain the multiple agents with updated states;

[0177] Based on the multiple agents after the state update and the number of representative objects, the full set of objects after the state update is obtained.

[0178] In an exemplary embodiment, the state includes a new state, an active state, and a dead state. The new state refers to a duration that is continuously observed that is greater than a first threshold and less than or equal to a second threshold. The active state refers to a duration that is continuously observed that is greater than the second threshold. The dead state refers to a duration that is not observed that is greater than the third threshold.

[0179] The number of objects in the newly added state and the number of objects in the extinct state are used to update the total number of objects in the updated state.

[0180] In one exemplary embodiment, such as Figure 10 As shown, the device 900 further includes a target object determination module 906 and a statistical data update module 907.

[0181] The target object determination module 906 is used to determine a target object set from the full set of objects after the state update, wherein each target object in the target object set is in the active state.

[0182] The statistical data update module 907 is used to update the business statistical data based on the actual or inferred target locations of each target object within the target area and the target time period, respectively, to obtain the updated business statistical data.

[0183] In summary, the technical solution provided by this application, by inferring the target location within a target time period for observable objects without target data, and updating the status of all objects within the target area and target time period based on the inferred or actual target location corresponding to each observable object, achieves automatic status updates for all objects within the target area and target time period. This shortens the update cycle of the status of all objects to the daily or weekly level, thereby improving the efficiency and granularity of dynamic object updates, expanding the applicable scenarios for dynamic object statistics, and enhancing business statistical capabilities.

[0184] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0185] Please refer to Figure 11 This diagram illustrates a structural block diagram of a computer device according to an embodiment of this application. This computer device can be used to implement the data processing methods provided in the above embodiments. Specifically, it may include the following:

[0186] The computer device 1100 includes a central processing unit (such as a CPU, GPU, or FPGA) 1101, a system memory 1104 including RAM (Random-Access Memory) 1102 and ROM (Read-Only Memory) 1103, and a system bus 1105 connecting the system memory 1104 and the central processing unit 1101. The computer device 1100 also includes a basic input / output system (I / O system) 1106 to facilitate information transfer between various devices within the server, and a large-capacity storage device 1107 for storing the operating system 1113, application programs 1114, and other program modules 1115.

[0187] The basic input / output system 1106 includes a display 1108 for displaying information and an input device 1109 for user input, such as a mouse or keyboard. Both the display 1108 and the input device 1109 are connected to the central processing unit 1101 via an input / output controller 1110 connected to the system bus 1105. The basic input / output system 1106 may also include the input / output controller 1110 for receiving and processing input from multiple other devices such as a keyboard, mouse, or electronic stylus. Similarly, the input / output controller 1110 also provides output to a display screen, printer, or other types of output devices.

[0188] The mass storage device 1107 is connected to the central processing unit 1101 via a mass storage controller (not shown) connected to the system bus 1105. The mass storage device 1107 and its associated computer-readable media provide non-volatile storage for the computer device 1100. That is, the mass storage device 1107 may include computer-readable media (not shown) such as a hard disk or a CD-ROM (Compact Disc Read-Only Memory) drive.

[0189] Without loss of generality, the computer-readable medium may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include RAM, ROM, EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory or other solid-state storage technologies, CD-ROM, DVD (Digital Video Disc) or other optical storage, magnetic tape cassettes, magnetic tape, disk storage, or other magnetic storage devices. Of course, those skilled in the art will recognize that the computer storage medium is not limited to the above-mentioned types. The system memory 1104 and mass storage device 1107 described above can be collectively referred to as memory.

[0190] According to an embodiment of this application, the computer device 1100 can also be connected to a remote computer on a network, such as the Internet. That is, the computer device 1100 can be connected to the network 1112 via the network interface unit 1111 connected to the system bus 1105, or the network interface unit 1111 can be used to connect to other types of networks or remote computer systems (not shown).

[0191] The memory also includes a computer program stored in the memory and configured to be executed by one or more processors to implement the above-described data processing method.

[0192] In one exemplary embodiment, a computer-readable storage medium is also provided, wherein a computer program is stored therein, which, when executed by a processor, implements the above-described data processing method.

[0193] Optionally, the computer-readable storage medium may include: ROM (Read-Only Memory), RAM (Random-Access Memory), SSD (Solid State Drives), or optical disc, etc. The random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0194] In one exemplary embodiment, a computer program product or computer program is also provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, causing the computer device to perform the data processing method described above.

[0195] It should be noted that the information (including but not limited to object device information, object personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the object or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0196] It should be understood that "multiple" as used herein refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. Furthermore, the step numbers described herein are merely illustrative of one possible execution order. In some other embodiments, the steps may not be executed in numerical order, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the illustration. This application does not limit this.

[0197] The above description is merely an exemplary embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A data processing method, characterized by, The method comprises: acquiring a set of observable objects in a target region and in a target period; determining a first set of observable objects and a second set of observable objects from the set of observable objects, each first observable object in the first set of observable objects corresponding to target data in the target period, each second observable object in the second set of observable objects not having target data in the target period; acquiring a real target location in the target period corresponding to each first observable object according to the target data in the target period corresponding to each first observable object; speculating a target location in the target period corresponding to each second observable object to obtain a speculated target location in the target period corresponding to each second observable object; updating the state of each object in the target region and in the target period based on the real target location in the target period corresponding to each first observable object and the speculated target location in the target period corresponding to each second observable object to obtain a set of full objects after state updating; wherein the set of full objects after state updating comprises the set of observable objects.

2. The method of claim 1, wherein, The method of speculating a target location in the target period corresponding to each second observable object to obtain a speculated target location in the target period corresponding to each second observable object comprises: for a target second observable object in the second set of observable objects, acquiring a set of historical trajectories of the target second observable object in a historical period; wherein the historical period refers to a period before the target period, and each historical trajectory in the set of historical trajectories is composed of position points; aggregating each historical trajectory to obtain an aggregated trajectory; extracting at least one target point from the aggregated trajectory; obtaining a speculated target location in the target period corresponding to the target second observable object according to the at least one target point.

3. The method of claim 2, wherein, The method of obtaining a speculated target location in the target period corresponding to the target second observable object according to the at least one target point comprises: mapping the at least one target point to obtain a mapping matrix corresponding to the at least one target point, the mapping matrix comprising a time matrix in a time dimension and a space matrix in a space dimension; for a target object feature in at least one object feature corresponding to the target second observable object, acquiring a predicted target point corresponding to the target object feature based on the mapping matrix corresponding to the at least one target point; superimposing the predicted target points corresponding to the at least one object feature to obtain superimposed predicted target points; obtaining a speculated target location in the target period corresponding to the target second observable object based on the superimposed predicted target points and a first target point; wherein the first target point refers to a last target point before the end time of the target period.

4. The method of claim 3, wherein, The method of acquiring a predicted target point corresponding to the target object feature based on the mapping matrix corresponding to the at least one target point comprises: obtain a historical target point set, each historical target point in the historical target point set corresponding to a key and a value; obtain a similarity measure between the mapping matrix corresponding to the at least one target point and each historical target point corresponding to the key, respectively; scale and normalize the similarity measure between the mapping matrix and each historical target point to obtain a weight corresponding to each historical target point, respectively; fuse the values corresponding to each historical target point based on the weight corresponding to each historical target point to obtain a predicted target point corresponding to the target object feature.

5. The method of claim 1, wherein, update the state of each object in the target region and in the target period based on the real target location of each first observable object in the target period and the speculative target location of each second observable object in the target period to obtain a full quantity object set after state update, including: determine target observable objects from the first observable objects and the second observable objects, each target observable object representing a type of observable object, each type of observable object having similar or identical object features; construct an agent model based on the target observable objects, wherein the agent model includes a plurality of agents, and each agent corresponds to a target observable object; obtain input data corresponding to each agent, wherein the input data includes at least one of a production location, a resident location in a threshold period, a real target location or a predicted target location in the target period, a historical target location set, a first target point, and a representative object number; update the state of each agent based on the input data corresponding to each agent to obtain a plurality of agents after state update; obtain the full quantity object set after state update based on the plurality of agents after state update and the representative object number.

6. The method of claim 5, wherein, The state includes an added state, an active state, and a dead state, the added state refers to a duration of continuous observation greater than a first threshold and less than or equal to a second threshold, the active state refers to a duration of continuous observation greater than the second threshold, and the dead state refers to a duration of non-observation greater than a third threshold. The number of objects in the added state and the number of objects in the dead state are used to update the number corresponding to the full quantity object set after state update.

7. The method of claim 6, wherein, The method further includes: determine a target object set from the full quantity object set after state update, each target object in the target object set being in the active state; update business statistics data based on the real target location or the speculative target location of each target object in the target region and in the target period to obtain updated business statistics data.

8. A data processing apparatus, characterized by, The device includes: an observable object acquisition module configured to obtain a set of observable objects in a target region and in a target period; An observable object division module is configured to determine a first observable object set and a second observable object set from the observable object set, each first observable object in the first observable object set corresponds to target data in the target period, and each second observable object in the second observable object set does not have target data in the target period. A real target location acquisition module is configured to acquire a real target location in the target period corresponding to each first observable object according to the target data in the target period corresponding to each first observable object. A speculative target location acquisition module is configured to perform target location speculation in the target period on each second observable object to obtain a speculative target location in the target period corresponding to each second observable object. A full-quantity state update module is configured to perform state update on each object in the target region and in the target period based on the real target location in the target period corresponding to each first observable object and the speculative target location in the target period corresponding to each second observable object, to obtain a full-quantity object set after state update, wherein the full-quantity object set after state update includes the observable object set.

9. A computer device, comprising: The computer device includes a processor and a memory, the memory stores a computer program, the computer program is loaded and executed by the processor to implement the data processing method in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, the computer program is loaded and executed by the processor to implement the data processing method in any one of claims 1 to 7.

11. A computer program product, characterised in that, The computer program product includes computer instructions, the computer instructions are executed by the processor to implement the data processing method in any one of claims 1 to 7.

Images