Service data processing method and device, electronic equipment, computer readable storage medium and computer program product

Abstract

The invention provides a service data processing method and device, electronic equipment, a computer readable storage medium and a computer program product. The service data processing method and device are applied to various service aggregation scenes such as cloud technology, artificial intelligence, traffic and vehicle-mounted application. The service data processing method comprises the following steps: extracting K historical service features of services provided by S service providers in a region division result; determining K service area features and area integration weights by combining the to-be-served time, the K historical service features and the area position relationship; integrating the K service area features based on the area integration weight to obtain J area integration features; on the basis of the J region integration features and the K service region features, predicting S target probabilities that the S service providers provide services in the target service region; and in combination with the S target probabilities, presenting the recommended at least one service provider and the corresponding to-be-served probabilities. According to the invention, the recommendation accuracy of the service provider can be improved.

Description

Technical Field

[0001] This application relates to information processing technology in the field of computer applications, and in particular, to an information processing method, apparatus, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] A service aggregation scenario refers to a scenario where multiple service providers are aggregated and a service provider is selected from the multiple service providers to provide a service. In a service aggregation scenario, when responding to a service provision request, the recommended service provider is often a service provider whose selection frequency is greater than a frequency threshold, and this service provider may have a waiting time for providing a service that is greater than a waiting time threshold. Therefore, the recommendation accuracy of the service provider is affected. Summary of the Invention

[0003] Embodiments of this application provide a service data processing method, apparatus, electronic device, computer-readable storage medium, and computer program product, which can improve the recommendation accuracy of service providers.

[0004] The technical solution of the embodiments of this application is implemented as follows:

[0005] Embodiments of this application provide a service data processing method, the method comprising:

[0006] In response to a service provision request, with respect to the service time to be served, extract the historical service features of S service providers providing services in each service area in the area division result, to obtain K of the historical service features, where both S and K are integers greater than 1;

[0007] Combining the service time to be served, the K historical service features, and the area position relationship of the area division result, determine K service area features and area integration weights;

[0008] Based on the area integration weights, integrate the K service area features to obtain J area integration features, where J is a positive integer less than K;

[0009] Based on the J area integration features and the K service area features, predict the S target probabilities of the S service providers providing services in the target service area;

[0010] Combining the S target probabilities, present at least one of the recommended service providers, and the to-be-served probabilities corresponding to at least one of the service providers.

[0011] Embodiments of this application provide a service data processing apparatus, the service data processing apparatus comprising:

[0012] A feature extraction module, configured to, in response to a service provision request, extract historical service features of each service area provided by S service providers in the region division result relative to the time to be served, so as to obtain K historical service features, where both S and K are integers greater than 1;

[0013] The feature extraction module is further configured to determine K service area features and area integration weights by combining the time to be served, the K historical service features, and the regional position relationship of the region division result;

[0014] A feature integration module, configured to integrate the K service area features based on the area integration weights to obtain J area integration features, where J is a positive integer less than K;

[0015] A probability prediction module, configured to predict S target probabilities of the S service providers providing services in the target service area based on the J area integration features and the K service area features;

[0016] A service recommendation module, configured to present at least one of the recommended service providers and the corresponding service probabilities to be served by at least one of the service providers in combination with the S target probabilities.

[0017] In an embodiment of the present application, the feature extraction module is further configured to splice the time feature of the time to be served and the K historical service features to obtain an initial splicing feature; perform a linear transformation on the initial splicing feature to obtain K initial area features; enhance the K initial area features based on the regional position relationship of the region division result to obtain the K service area features; and obtain the area integration weights by combining the K initial area features and the regional position relationship.

[0018] In an embodiment of the present application, the feature extraction module is further configured to obtain a first service cycle where the time to be served is located to obtain a target first service cycle; obtain a first time period number of the time to be served within the target first service cycle; obtain a second time period number of the target first service cycle within a second service cycle, where the second service cycle includes a plurality of the first service cycles; and integrate the encoding results of the first time period number and the encoding results of the second time period number into the time feature.

[0019] In an embodiment of the present application, the feature extraction module is further configured to perform first attention processing on the K initial region features based on the region position relationship of the region division result to obtain K enhanced region features; perform second attention processing on the K enhanced region features based on the region position relationship to obtain K region features to be connected; and connect the K enhanced region features and the K region features to be connected to obtain the K service region features.

[0020] In an embodiment of the present application, the feature integration module is further configured to iterate through a specified number of sequences by a, and perform the following processing on each traversed specified number, where a is a positive integer variable: allocate the specified number of a-th to-be-allocated features based on the (a - 1)-th allocation weight to obtain the specified number of a-th to-be-allocated features. When a is 1, the (a - 1)-th allocation weight is the region integration weight, and the specified number of (a - 1)-th to-be-allocated features are the K service region features; when the traversed specified number is different from the last specified number, determine the a-th position relationship by combining the (a - 1)-th allocation weight and the (a - 1)-th position relationship, and determine the a-th allocation weight by combining the a-th position relationship and the specified number of a-th to-be-allocated features, and continue to iterate a. When a is 1, the (a - 1)-th position relationship is the region position relationship; when the traversed specified number is the last specified number, end the iteration of a, and determine the specified number of a-th to-be-allocated features at the end of the iteration of a as the J region integration features.

[0021] In an embodiment of the present application, the probability prediction module is further configured to restore the J region integration features to the region division result based on the region integration weight to obtain K to-be-enhanced features; combine the K to-be-enhanced features and the K service region features to obtain K to-be-predicted features; predict S service providing probabilities of the S service providers providing services in each service region based on the K to-be-predicted features; determine the target service region where the to-be-served position is located from the region division result; and determine S target probabilities corresponding to the target service region based on the S service providing probabilities of the S service providers providing services in each service region.

[0022] In an embodiment of the present application, the service data processing device further includes a probability multiplexing module, configured to obtain a requested service time and a requested service location in response to a next service provision request; when a time difference between the requested service time and the to-be-served time is less than a time difference threshold, determine the service area where the requested service location is located from the area division result to obtain a requested service area; determine S service provision probabilities corresponding to the requested service area based on S service provision probabilities of the S service providers providing services in each service area; and present service provision information based on the S service provision probabilities corresponding to the requested service area.

[0023] In an embodiment of the present application, the feature extraction module is further configured to, in response to the service provision request, obtain a time period service information sequence of the S service providers providing services in the service area and a specified sequence length relative to the to-be-served time; obtain a cycle service information sequence of the S service providers providing services in the service area and a specified cycle sequence relative to the to-be-served time; and determine the encoding result of the time period service information sequence and the encoding result of the cycle service information sequence as the historical service features to obtain K historical service features.

[0024] In an embodiment of the present application, the feature extraction module is further configured to, in response to the service provision request, use the to-be-served time as the time series end time and determine a time series start time based on the specified sequence length; slide from the time series start time to the time series end time with a specified sliding window and a specified sliding step to obtain a to-be-extracted time period sequence; for each to-be-extracted time period in the to-be-extracted time period sequence, combine the first service distribution volume and the first service provision volume of the S service providers in the service area respectively as the time period service information to obtain the time period service information sequence corresponding to the specified sequence length.

[0025] In an embodiment of the present application, the feature extraction module is further configured to determine a to-be-processed time corresponding to the to-be-served time in each specified cycle of the specified cycle sequence; obtain a historical service time period centered on the to-be-processed time based on a specified range length; and within the historical service time period, combine the second service distribution volume and the second service provision volume of the S service providers in the service area as the cycle service information to obtain the cycle service information sequence corresponding to the specified cycle sequence.

[0026] In an embodiment of the present application, the service data processing device further includes a region division module, configured to obtain a set of service locations within a specified historical duration before the current division moment for the region to be divided; cluster the set of service locations to obtain K service location clusters; determine the region corresponding to each service location cluster as the service region, and obtain K service regions corresponding to the K service location clusters, where the region division result includes the K service regions.

[0027] In an embodiment of the present application, the region division module is further configured to determine the regional center position of each service region in the region division result; determine the regional distance between any two service regions in the region division result based on the regional center position; determine two service regions with a regional distance less than the adjacent distance threshold as adjacent service regions; and determine the regional position relationship of the region division result based on the adjacent service regions.

[0028] In an embodiment of the present application, the prediction of the S target probabilities is implemented through a service prediction model. The service data processing device further includes a model training module, configured to extract historical service sample features of providing services in each sample region in the region division sample result relative to the time of the sample to be served; use the model to be trained to predict the time of the sample to be served, the historical service sample features, and the sample region position relationship, and obtain N prediction probabilities of N service providing sample parties providing services in each sample region, where the model to be trained is a neural network model to be trained for predicting the probability of providing services, and N is an integer greater than 1; calculate the predicted service dispatch quantity based on the service dispatch sample quantity of each service providing sample party in each sample region and the prediction probability; calculate the loss function value based on the difference between the predicted service dispatch quantity and the service dispatch sample quantity of each service providing sample party in each sample region; and train the model to be trained based on the loss function value to obtain the service prediction model.

[0029] In an embodiment of the present application, the service recommendation module is further configured to select at least one of the target probabilities with the largest target probability from the S target probabilities; determine at least one service providing party corresponding to the at least one target probability from the S service providing parties; calculate the service probability to be served corresponding to the at least one target probability; present the service probability to be served, and present the recommended at least one service providing party in a specified style.

[0030] In an embodiment of the present application, the service recommendation module is further configured to, in response to a selection and adjustment operation for the to-be-served probability, obtain at least one adjusted service provider; determine at least one adjusted target probability corresponding to the at least one adjusted service provider from the S target probabilities; calculate the current service probability corresponding to the at least one adjusted target probability; present the current service probability, and present the at least one adjusted service provider in the specified style.

[0031] In an embodiment of the present application, the service recommendation module is further configured to present the regional integration weight in a specified graph style, where the specified graph style includes a first graph style, a second graph style, and a third graph style. The first graph style is used to present the regional integration weight from the regional division result and the regional cluster dimension, the second graph style is used to present the regional integration weight based on the regional cluster corresponding to the highest weight, and the third graph style is used to present the regional integration weight from the longitude and latitude dimensions.

[0032] An embodiment of the present application provides an electronic device for service data processing. The electronic device includes:

[0033] A memory for storing computer-executable instructions or computer programs;

[0034] A processor, when executing the computer-executable instructions or computer programs stored in the memory, implements the service data processing method provided by the embodiment of the present application.

[0035] An embodiment of the present application provides a computer-readable storage medium storing computer-executable instructions or computer programs, where the computer-executable instructions or computer programs, when executed by a processor, implement the service data processing method provided by the embodiment of the present application.

[0036] An embodiment of the present application provides a computer program product including computer-executable instructions or computer programs, where the computer-executable instructions or computer programs, when executed by a processor, implement the service data processing method provided by the embodiment of the present application.

[0037] The embodiments of the present application have at least the following beneficial effects: When responding to a service provision request, first combine the K historical service characteristics of the service time to be served, the regional division result, and the regional location relationship to determine the K service area characteristics and regional integration weights of the regional division result, and then aggregate the relevant service area characteristics based on the regional integration weights to learn the overall characteristics of the regional division result. Finally, combine the J aggregated regional characteristics and service area characteristics to predict the S target probabilities of S service providers providing services in the target service area, so that at least one service provider can be recommended based on the S target probabilities in the target service area; thus, the service provider with the highest probability of providing services can be accurately determined through the at least one recommended service provider, and further, the recommendation accuracy of the service provider can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a schematic structural diagram of a service data processing system provided by an embodiment of the present application;

[0039] Figure 2 is a kind of Figure 1 terminal in the present application;

[0040] Figure 3 is a flowchart of a service data processing method provided by an embodiment of the present application Figure 1 ;

[0041] Figure 4 is a flowchart of a service data processing method provided by an embodiment of the present application Figure 2 ;

[0042] Figure 5 is an exemplary service provider recommendation schematic diagram provided by an embodiment of the present application;

[0043] Figure 6 is a flowchart of a service data processing method provided by an embodiment of the present application Figure 3 ;

[0044] Figure 7 is a flowchart of a model training method provided by an embodiment of the present application;

[0045] Figure 8 is an exemplary order acceptance rate prediction flowchart provided by an embodiment of the present application;

[0046] Figure 9 is an exemplary information extraction schematic diagram provided by an embodiment of the present application;

[0047] Figure 10 is an exemplary model structure schematic diagram provided by an embodiment of the present application;

[0048] Figure 11 It is a schematic structural diagram of an exemplary node characterization module provided by an embodiment of the present application;

[0049] Figure 12 It is an exemplary logic diagram provided by an embodiment of the present application;

[0050] Figure 13 It is an exemplary order receiving rate presentation diagram provided by an embodiment of the present application;

[0051] Figure 14 It is an exemplary visualization result diagram of an allocation matrix provided by an embodiment of the present application;

[0052] Figure 15 It is an exemplary structured diagram of an allocation matrix provided by an embodiment of the present application;

[0053] Figure 16 It is an exemplary two-dimensional diagram of an allocation matrix provided by an embodiment of the present application;

[0054] Figure 17 It is another exemplary visualization result diagram of an allocation matrix provided by an embodiment of the present application;

[0055] Figure 18 It is another exemplary structured diagram of an allocation matrix provided by an embodiment of the present application;

[0056] Figure 19 It is another exemplary two-dimensional diagram of an allocation matrix provided by an embodiment of the present application. Detailed implementation manners

[0057] In order to make the objectives, technical solutions, and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limitations on the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0058] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0059] In the following description, the terms "first / second", etc. are used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first / second", etc. can be interchanged in a specific order or sequence when allowed, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0060] In the embodiments of the present application, the term "module" or "unit" refers to a computer program with a predetermined function or a part of a computer program, which works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as a processing circuit or a memory), or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of an overall module or unit that includes the functions of that module or unit.

[0061] Unless otherwise defined, all technical and scientific terms used in the embodiments of the present application have the same meaning as commonly understood by those skilled in the technical field to which the present application belongs. The terms used in the embodiments of the present application are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0062] Before further elaborating on the embodiments of the present application, the nouns and terms involved in the embodiments of the present application are described. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0063] 1) Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to obtain the best results. That is to say, artificial intelligence is a comprehensive technology in computer science, used to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. Artificial intelligence enables machines to have the functions of perception, reasoning, and decision-making by studying the design principles and implementation methods of various intelligent machines.

[0064] It should be noted that artificial intelligence technology is an interdisciplinary subject, involving a wide range of fields, including both hardware-level technologies and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, pre-trained models are also known as large models and basic models; pre-trained models can be widely applied to downstream tasks in various directions of artificial intelligence after fine-tuning. Artificial intelligence software technologies include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning. In the embodiments of the present application, the prediction of probabilities such as the order acceptance rate, service provision probability, and target probability can be achieved through artificial intelligence technology.

[0065] 2) Machine Learning (ML) is an interdisciplinary subject that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It is used to study the computer simulation or implementation of human learning behaviors to acquire new knowledge or skills, and reorganize the existing knowledge structure to continuously improve its own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Machine learning applications cover all fields of artificial intelligence. Machine learning / deep learning usually includes technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, and inductive learning. Large models are the latest development results of machine learning / deep learning, integrating the above technologies. In the embodiments of the present application, the prediction of probabilities such as the order acceptance rate, service provision probability, and target probability can be achieved by combining machine learning / deep learning, and of course, it can also be achieved by large models.

[0066] 3) An artificial neural network is a mathematical model that imitates the structure and function of a biological neural network. The exemplary structures of the artificial neural network in the embodiments of the present application include a graph convolutional network (Graph Convolutional Network, GCN, a neural network for processing graph-structured data), a deep neural network (Deep Neural Networks, DNN), a convolutional neural network (Convolutional Neural Network, CNN), a recurrent neural network (Recurrent Neural Network, RNN), a neural state machine (Neural State Machine, NSM), and a phase-functioned neural network (Phase-Functioned Neural Network, PFNN), etc. In the embodiments of the present application, the prediction of probabilities such as the order acceptance rate, service provision probability, and target probability can be achieved through a model constructed based on an artificial neural network (abbreviated as a neural network model).

[0067] It should be noted that with the research and progress of artificial intelligence technology, artificial intelligence technology has been studied and applied in multiple fields. For example, smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, driverless, autonomous driving, drones, digital twins, virtual humans, robots, artificial intelligence-generated content (AIGC), conversational interactions, smart healthcare, smart customer service, and virtual scenarios, etc. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role. The service data processing method provided in the embodiments of the present application includes the processing of predicting probabilities such as the order acceptance rate, service provision probability, and target probability, and involves technologies such as machine learning / deep learning of artificial intelligence.

[0068] 4) The online car-hailing capacity provider, also known as the online car-hailing capacity organizer, refers to the provider of online car-hailing services, which is used to dispatch the most suitable online car-hailing in response to a car-hailing request for service. In the embodiments of the present application, the service provider may be an online car-hailing capacity provider, etc.

[0069] 5) The online car-hailing aggregation service platform aggregates multiple online car-hailing capacity providers through a unified platform to enable the processing of requests for online car-hailing services from multiple online car-hailing capacity providers simultaneously. Among them, the unified platform is the online car-hailing aggregation service platform.

[0070] 6) Order dispatching refers to the processing of sending a service provision request to the service provider. For example, after receiving a car-hailing request, the online car-hailing aggregation service platform dispatches orders to each online car-hailing capacity provider to provide online car-hailing services.

[0071] 7) Order receiving refers to the processing of the service provider providing services based on the order dispatching request. For example, after receiving the order dispatching information from the aggregation service platform, the online car-hailing capacity provider determines an online car-hailing and provides online car-hailing services through the determined online car-hailing. In addition, the probability of an online car-hailing capacity provider receiving an order is called the order receiving rate.

[0072] It should be noted that in the service aggregation scenario, when responding to a service provision request, the recommended service providers are often those with a selection frequency greater than the frequency threshold, and there may be a situation where the waiting time for these service providers to provide services is greater than the waiting time threshold. Therefore, the accuracy of recommending service providers is affected.

[0073] It should also be noted that in the service aggregation scenario, the service provider for providing services is often obtained by receiving selection operations for multiple service providers. Since the service providers selected by the selection operations may have a low service provision probability (lower than the specified probability), the service provision efficiency is affected. In addition, when artificial intelligence is used in the service aggregation scenario, due to the single features extracted, the prediction accuracy is affected, which in turn affects the service provision efficiency.

[0074] In addition, when estimating the service provision probability using the ratio of the service provision volume to the service dispatch volume, in areas where the service dispatch volume is 0, the service provision probability cannot be estimated; when the service dispatch volume is less than the dispatch volume threshold, the estimated service provision probability is accidental and volatile (for example, when the service dispatch volume is 1, the estimated value of the service provision probability is 0% or 100%); therefore, estimating the service provision probability based on the ratio of the service provision volume to the service dispatch volume affects the prediction accuracy and thus the service provision efficiency.

[0075] Based on this, the embodiments of the present application provide a service data processing method, apparatus, electronic device, computer-readable storage medium, and computer program product, which can improve the recommendation accuracy and service provision efficiency of service providers. The following describes an exemplary application of the service data processing device provided by the embodiments of the present application. The service data processing device provided by the embodiments of the present application can be implemented as various types of terminals such as robots, smartphones, smart watches, laptops, tablets, desktop computers, smart home appliances, set-top boxes, in-vehicle intelligent devices, portable music players, personal digital assistants, dedicated messaging devices, intelligent voice interaction devices, portable game devices, and smart speakers, or can be implemented as a server. Hereinafter, an exemplary application will be described when the service data processing device is implemented as a terminal.

[0076] Refer to Figure 1 , Figure 1 is a schematic architecture diagram of the service data processing system provided by the embodiments of the present application; as Figure 1 shown, to support a service data processing application, in the service data processing system 100, the terminal 400 is connected to the server 200 and multiple servers 600 (referred to as service providers) through the network 300; wherein, the network 300 can be a wide area network or a local area network, or a combination of the two; the server 200 is used to provide service computing to the terminal 400 through the network 300; the terminal 400 is used to send a service dispatch request to multiple servers 600 through the network 300. In addition, the service data processing system 100 further includes a database 500 for providing data support to the server 200; and, Figure 1 shows a case where the database 500 is independent of the server 200. In addition, the database 500 can also be integrated in the server 200, and the embodiments of the present application do not limit this; and, the service provider can also be implemented as a terminal. Here, an example where the service provider is implemented as a server is used for illustration.

[0077] Terminal 400 is configured to, in response to a service provision request, extract historical service features of services provided by S service providers in each service area of the regional division result relative to the time to be served, thereby obtaining K historical service features; determine K service area features and regional integration weights based on the time to be served, the K historical service features, and the regional location relationship of the regional division result; integrate the K service area features based on the regional integration weights to obtain J regional integration features; predict S target probabilities of S service providers providing services in the target service area based on the J regional integration features and the K service area features; and present at least one recommended service provider and the corresponding probability of being served for at least one service provider, based on the S target probabilities (graphical interface 410-1 is shown as an example, presenting at least one recommended online ride-hailing capacity provider and order acceptance rate). Terminal 400 is further configured to send a service dispatch request to a corresponding server 600 via network 300 based on the at least one recommended service provider.

[0078] In some embodiments, server 200 and server 600 may be independent physical servers, or a server cluster or distributed system composed of multiple physical servers, or cloud servers that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal and the server may be connected directly or indirectly via wired or wireless communication, which is not limited in the embodiments of the present application.

[0079] See also Figure 2 , Figure 2 This embodiment of the present application provides a Figure 1 The schematic diagram of the terminal structure in Figure 2 As shown, the terminal 400 includes: at least one processor 410, a memory 450, at least one network interface 420 and a user interface 430. The various components in the terminal 400 are coupled together via a bus system 440. It is understood that the bus system 440 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 440 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 2 Various buses are labeled as bus system 440 .

[0080] The processor 410 may be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or any conventional processor, etc.

[0081] The user interface 430 includes one or more output devices 431 that enable the presentation of media content, including one or more speakers and / or one or more visual display screens. The user interface 430 also includes one or more input devices 432, including user interface components that facilitate user input, such as a keyboard, a mouse, a microphone, a touch screen display, a camera, other input buttons, and controls.

[0082] The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard disk drives, optical disc drives, etc. The memory 450 optionally includes one or more storage devices that are physically remote from the processor 410.

[0083] The memory 450 includes volatile memory or non-volatile memory, and may also include both volatile and non-volatile memory. The non-volatile memory may be a read-only memory (ROM), and the volatile memory may be a random access memory (RAM). The memory 450 described in the embodiments of the present application is intended to include any suitable type of memory.

[0084] In some embodiments, the memory 450 is capable of storing data to support various operations. Examples of such data include programs, modules, and data structures, or subsets or supersets thereof, which are described below by way of example.

[0085] The operating system 451 includes system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks;

[0086] The network communication module 452 is used to reach other electronic devices via one or more (wired or wireless) network interfaces 420. Exemplary network interfaces 420 include: Bluetooth, Wi-Fi (Wireless Fidelity), and USB (Universal Serial Bus), etc.;

[0087] A presentation module 453 for enabling the presentation of information (e.g., a user interface for operating a peripheral device and displaying content and information) via one or more output devices 431 associated with the user interface 430 (e.g., a display screen, a speaker, etc.);

[0088] An input processing module 454 for detecting one or more user inputs or interactions from one of the one or more input devices 432 and translating the detected inputs or interactions.

[0089] In some embodiments, the service data processing device provided by the embodiments of the present application may be implemented in software. Figure 2 Shown is a service data processing device 455 stored in the memory 450, which may be software in the form of a program and a plug-in, etc., including the following software modules: a feature extraction module 4551, a feature integration module 4552, a probability prediction module 4553, a service recommendation module 4554, a probability reuse module 4555, a region division module 4556, and a model training module 4557. These modules are logical, and thus can be combined arbitrarily or further split according to the implemented functions. The functions of each module will be described below.

[0090] In some embodiments, the service data processing device provided by the embodiments of the present application may be implemented in hardware. As an example, the service data processing device provided by the embodiments of the present application may be a processor in the form of a hardware decoding processor, which is programmed to execute the service data processing method provided by the embodiments of the present application. For example, a processor in the form of a hardware decoding processor may employ one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs), or other electronic components.

[0091] In some embodiments, a terminal or a server may implement the service data processing method provided in the embodiments of the present application by running various computer-executable instructions or computer programs. For example, the computer-executable instructions may be microprogram-level commands, machine instructions, or software instructions. The computer program may be a native program or a software module in an operating system; it may be a local (Native) application (APPlication, APP), that is, a program that needs to be installed in an operating system to run, such as a car-hailing APP or an express delivery APP; it may also be a small program that can be embedded in any APP, that is, a program that only needs to be downloaded to a browser environment to run. In short, the above computer-executable instructions may be instructions in any form, and the above computer programs may be application programs, modules, or plugins in any form.

[0092] Next, the service data processing method provided in the embodiments of the present application will be described in conjunction with the exemplary applications and implementations of the service data processing device provided in the embodiments of the present application. In addition, the service data processing method provided in the embodiments of the present application is applied to various service aggregation scenarios that determine service providers, such as cloud technology, artificial intelligence, transportation, and vehicle-mounted.

[0093] See Figure 3 , Figure 3 is a flowchart showing the service data processing method provided in the embodiments of the present application Figure 1 , Figure 3 The execution subject of each step in Figure 3 is the service data processing device; the following will be described in conjunction with

[0094] Step 101: In response to a service provision request, extract the historical service features of S service providers providing services in each service area in the area division result relative to the service time to be served, and obtain K historical service features.

[0095] In an embodiment of the present application, when a service data processing device requests a service provider to provide a service, the service data processing device also receives a service provision request; at this time, in response to the service provision request, the service data processing device starts to execute the service provision process. Since the service provision request is a request for the service provider to provide a service, the service data processing device can, in response to the service provision request, obtain the time to be served from the service provision request, or use the time when the service provision request is received as the time to be served; during the execution of the service provision process, in order to recommend a service provider for providing the service, the service data processing device uses the time before the time to be served as the historical time, extracts features from the information on the service provided in each service area in the regional division result by the S service providers at the historical time, and thus obtains the historical service features of the service area; since the regional division result includes K service areas, the service data processing device can obtain K historical service features. Here, both S and K are integers greater than 1.

[0096] It should be noted that the service requested by the service provision request can be provided by any one of multiple service providers. The service requested by the service provision request can be, for example, a car-hailing service, an inter-regional express delivery service, an intra-regional express delivery service, etc. The time to be served refers to the time when the service is to be provided. For example, it is the time when the car-hailing service is to be provided. The regional division result refers to the division result of the area to be divided, which can be divided by means of clustering or uniform division, etc. The embodiments of the present application do not make any limitations in this regard; among them, the area to be divided refers to the entire area where the service is to be provided, which can be an urban area, a national area, a district or county area, a street area, etc. The embodiments of the present application do not make any limitations in this regard; here, the service area is the smallest area unit for predicting the probability of providing a service. The historical service feature represents the feature representation corresponding to the historical service information before the time to be served, and characterizes the service supply and demand characteristics corresponding to the historical time, while the historical service information refers to the service dispatch volume and service provision volume of the service area providing the service before the time to be served; among them, the service dispatch volume is the number of order dispatches, and the service provision volume is the number of order acceptances. In addition, the service provider is used to provide the service requested by the service provision request. Any one of the S service providers can independently provide the service requested by the service provision request, and the S service providers are different.

[0097] It can be understood that by extracting the historical service features of each service area in the regional division result, the prediction of probability based on the historical service features in units of service areas can reduce data calculation and resource consumption compared with the prediction based on the service provision request dimension.

[0098] See Figure 4, 4 is a flowchart of the service data processing method provided by an embodiment of the present application Figure 2 , Figure 4 The execution subject of each step in Figure 4 As shown, in an embodiment of the present application, step 101 can be implemented through steps 1011 to 1013; that is, in response to a service provision request, the service data processing device extracts, relative to the service time to be served, the historical service characteristics of S service providers providing services in each service area in the area division result, obtaining K historical service characteristics, including steps 1011 to 1013, which will be described separately below.

[0099] Step 1011: In response to a service provision request, relative to the service time to be served, obtain the time period service information sequence of S service providers providing services in the service area and within the specified sequence length.

[0100] In an embodiment of the present application, when the service data processing device extracts historical service characteristics, for the corresponding service area, first, with the service time to be served as the time cut-off point, it extracts the service dispatch volume and service provision volume of each time period sequence within the time period corresponding to the specified sequence length, thereby obtaining the time period service information sequence of providing services in the service area and within the specified sequence length. Thus, each time period service information in the time period service information sequence represents the service provision situation of S service providers in each time period sequence and service area before the service time to be served, and the service provision situation is measured by the service dispatch volume and service provision volume.

[0101] In an embodiment of the present application, the service data processing device, in response to a service provision request, relative to the service time to be served, obtains the time period service information sequence of providing services in the service area and within the specified sequence length, including: the service data processing device first responds to the service provision request, uses the service time to be served as the end time of the time series, and determines the start time of the time series based on the specified sequence length; then, from the start time of the time series to the end time of the time series, it slides with the specified sliding window and specified sliding step to obtain the time period sequence to be extracted; finally, for each time period to be extracted in the time period sequence to be extracted, it combines the first service dispatch volume and the first service provision volume of S service providers in the service area respectively into time period service information, obtaining the time period service information sequence corresponding to the specified sequence length.

[0102] It should be noted that the sequence of time periods to be extracted is a sequence composed of the time periods to be extracted determined by the sliding of a sliding window. Thus, each time period to be extracted is the time period corresponding to each sliding window between the start time and the end time of the time series. Among them, the sequence of time periods to be extracted is the above-mentioned various time period sequences. The first service dispatch volume is the service dispatch volume of each service provider in the service area and the time period to be extracted, and the first service provision volume is the service provision volume of each service provider in the service area and the time period to be extracted. Here, the time period service information includes the S first service dispatch volumes and the S first service provision volumes of the S service providers in the service area and the time period to be extracted.

[0103] It can be understood that since the sequence of time period service information is the service provision situation in the time period closest to the time to be served, when obtaining the historical service characteristics based on the sequence of time period service information, the historical service characteristics can include real-time change information.

[0104] Step 1012: Obtain the sequence of periodic service information of S service providers providing services in the service area and the specified cycle sequence relative to the time to be served.

[0105] It should be noted that when the service data processing device extracts historical service characteristics, it determines, in each specified cycle in the specified cycle sequence, the time period that is the same as the time period corresponding to the time to be served in the corresponding cycle, and extracts the service dispatch volume and service provision volume of S service providers in the service area within each determined identical time period, thus obtaining the sequence of periodic service information of providing services in the service area and the specified cycle sequence. Thus, each periodic service information in the sequence of periodic service information represents the service provision situation of S service providers in each historical service time period that is the same as the time period corresponding to the time to be served.

[0106] In the embodiment of the present application, for the service data processing device to obtain the sequence of periodic service information of providing services in the service area and the specified cycle sequence relative to the time to be served, it includes: the service data processing device first determines, in each specified cycle of the specified cycle sequence, the time to be processed corresponding to the time to be served; then, based on the specified range length, obtains the historical service time period centered on the time to be processed; finally, within the historical service time period, combines the second service dispatch volume and the second service provision volume of S service providers in the service area into periodic service information to obtain the sequence of periodic service information corresponding to the specified cycle sequence.

[0107] It should be noted that the historical service period is each period corresponding to the time to be served. The second service dispatch volume is the service dispatch volume of each service provider in the service area and the historical service period, and the second service provision volume is the service provision volume of each service provider in the service area and the historical service period. Here, the periodic service information includes the S second service dispatch volumes and the S second service provision volumes of the S service providers in the service area and the historical service period.

[0108] Step 1013: Determine the encoding results of the time period service information sequence and the periodic service information sequence as historical service features, and obtain K historical service features.

[0109] In the embodiment of the present application, the service data processing device may determine at least one of the encoding results of the time period service information sequence and the periodic service information sequence as historical service features; Figure 4 The example shows determining the encoding results of the time period service information sequence and the periodic service information sequence as historical service features.

[0110] Step 102: Combine the time to be served, the K historical service features, and the regional position relationship of the regional division result to determine K service area features and the regional integration weight.

[0111] In the embodiment of the present application, the service data processing device extracts features from the time to be served, combines the extracted features with the K historical service features, and then enhances the combined features using the regional position relationship, thereby obtaining K service area features. In addition, the service data processing device analyzes the correlation degree between the K service areas based on the combined features, and thus obtains the regional integration weight of the regional division result.

[0112] It should be noted that the regional integration weight represents the probability that the K service area features belong to the regional clusters of the specified number of regional clusters. Therefore, the regional integration weight includes K * the specified number of regional clusters of weights, and each weight represents the probability that the service area feature belongs to the regional cluster; where the specified number of regional clusters can be the number of specified regional clusters, and the regional cluster corresponding to the maximum weight of the service area belonging to the regional clusters of the specified number of regional clusters is the regional cluster to which the service area belongs. In addition, the regional position relationship represents the adjacency relationship of the K service areas.

[0113] Continue to refer to Figure 4 , in the embodiment of the present application, Step 102 can be implemented through Steps 1021 to 1024; that is, the service data processing device combines the time to be served, the K historical service features, and the regional position relationship of the regional division result to determine K service area features and the regional integration weight, including Steps 1021 to 1024, and the following will explain each step separately.

[0114] Step 1021: Concatenate the time feature of the time to be served and K historical service features to obtain an initial concatenated feature.

[0115] It should be noted that the time feature refers to the feature of the information corresponding to the time to be served; here, the service data processing device realizes the combination of the time feature and K historical service features by concatenating the time feature of the time to be served and K historical service features; among them, the concatenation result of the obtained time feature and K historical service features is the initial concatenated feature.

[0116] In the embodiment of the present application, the information corresponding to the time to be served includes at least one of the following: the number of the time to be served in various time periods (such as days, weeks, months, etc.), the change rate of service parameters (at least one of the service delivery volume and the service provision volume), and environmental information (such as traffic smoothness, weather information, etc.).

[0117] Step 1022: Perform a linear transformation on the initial concatenated feature to obtain K initial regional features.

[0118] It should be noted that the service data processing device performs a linear transformation on the initial concatenated feature, and the obtained linear transformation result is K initial regional features; among them, each initial regional feature represents the initial feature of the service area.

[0119] Step 1023: Based on the regional position relationship of the regional division result, enhance the K initial regional features to obtain K service area features.

[0120] It should be noted that since there is similarity in features between two service areas with an adjacency relationship, the service data processing device performs message passing on the K initial regional features based on the regional position relationship of the regional division result, so as to aggregate the initial features of adjacent service areas to the initial feature of this service area, realizing the enhancement of the initial regional features; here, the enhanced initial regional feature is the service area feature.

[0121] In the embodiment of the present application, the service data processing device enhances the K initial regional features based on the regional position relationship of the regional division result to obtain K service area features, including: the service data processing device first performs a first attention process on the K initial regional features based on the regional position relationship of the regional division result to obtain K enhanced regional features; then performs a second attention process on the K enhanced regional features based on the regional position relationship to obtain K regions to be connected features; finally, connects the K enhanced regional features and the K regions to be connected features to obtain K service area features.

[0122] It should be noted that the first attention processing is to perform message passing on the K initial region features, and the obtained result is K enhanced region features corresponding one-to-one to the K initial region features. The second attention processing is to perform message passing (which can be once or multiple times) on the K enhanced region features, and the obtained result is K to-be-connected region features corresponding one-to-one to the K enhanced region features. Here, the service data processing device uses a dense connection method to perform corresponding connections on the K enhanced region features and the K to-be-connected region features, thereby obtaining K service region features; thus, each service region feature is the connection result of the enhanced region feature and the corresponding to-be-connected region feature. In addition, the connection is, for example, processing such as fusion, splicing, and averaging.

[0123] It can be understood that by using the dense connection method to process the regional position relationship and the K initial region features, the flexibility of information transmission can be improved, and further the accuracy of the service region features can be improved.

[0124] Step 1024: Combine the K initial region features and the regional position relationship to obtain the regional integration weight.

[0125] In the embodiment of the present application, the service data processing device combines the K initial region features and the regional position relationship to obtain the regional integration weight, including: the service data processing device first performs third attention processing on the K initial region features based on the regional position relationship of the regional division result to obtain K first features; then performs fourth attention processing on the K first features based on the regional position relationship to obtain K second features; finally, connects the K first features and the K second features to obtain K to-be-normalized features; normalizes the K to-be-normalized features based on the specified number of region clusters of the region cluster, and thus obtains the regional integration weight. Among them, the parameters involved in the first attention processing to the fourth attention processing can be the same or different, and the embodiment of the present application does not limit this.

[0126] Step 103: Integrate the K service region features based on the regional integration weight to obtain J region integration features.

[0127] It should be noted that since the regional integration weight represents the probability that the K service area features belong to the regional clusters of the specified number of regional clusters, the service data processing device multiplies the transpose of the regional integration weight, the K service area features, and the regional integration weight, thereby obtaining the regional cluster features of the specified number of regional clusters; here, the service data processing device can directly use the product of the transpose of the regional integration weight, the K service area features, and the regional integration weight as the J regional integration features, or can iteratively integrate the product of the transpose of the domain integration weight, the K service area features, and the regional integration weight to obtain the J regional integration features. This application embodiment does not make a limitation in this regard. Among them, J is a positive integer.

[0128] In step 103 of the embodiment of the present application, the service data processing device integrates the K service area features based on the regional integration weight to obtain J regional integration features, including: the service data processing device traverses the specified number sequence through iteration a, and performs the following processing on each traversed specified number: allocating a plurality of (a - 1)-th to-be-allocated features based on the (a - 1)-th allocation weight to obtain the specified number of a-th to-be-allocated features; when the traversed specified number is different from the last specified number, combining the (a - 1)-th allocation weight and the (a - 1)-th position relationship to determine the a-th position relationship, and combining the a-th position relationship and the specified number of a-th to-be-allocated features to determine the a-th allocation weight, and continue to iterate a; when the traversed specified number is the last specified number, end the iteration a, and determine the specified number of a-th to-be-allocated features at the end of the iteration a as the J regional integration features.

[0129] It should be noted that the specified number sequence includes at least one specified number. When the specified number sequence includes multiple specified numbers, the sizes of the specified numbers in the specified number sequence decrease in turn; that is to say, the specified number sequence is the reverse order arrangement of the specified numbers. a is a positive integer variable, and the minimum value of a is 1, and the maximum value is the number of specified numbers in the specified number sequence. In addition, when a is 1, the (a - 1)-th allocation weight is the regional integration weight, the plurality of (a - 1)-th to-be-allocated features are the K service area features, and the (a - 1)-th position relationship is the regional position relationship. Here, the service data processing device can use the product of the transpose of the (a - 1)-th allocation weight, the plurality of (a - 1)-th to-be-allocated features, and the (a - 1)-th allocation weight as the specified number of a-th to-be-allocated features. When the traversed specified number is different from the last specified number, it indicates that the traversed specified number is the specified number before the last specified number in the specified number sequence. At this time, the allocation of the to-be-allocated features needs to be continued; when the traversed specified number is the last specified number in the specified number sequence, it indicates that the iterative allocation of the to-be-allocated features has been completed, so the iteration is ended at this time, and the specified number of a-th to-be-allocated features obtained is determined as the J regional integration features.

[0130] It can be understood that by integrating the K service area features through the area integration weights, the overall features of the K service areas can be integrated, and then the final features of each service area can be determined based on the area integration features, which can improve the accuracy of the final features.

[0131] Step 104: Based on the J area integration features and the K service area features, predict the S target probabilities of the S service providers providing services in the target service area.

[0132] In the embodiment of the present application, the service data processing device determines the final features of each service area based on the J area integration features and the K service area features; and predicts the S target probabilities of the S service providers providing services in the target service area based on the final features of each service area.

[0133] It should be noted that the target service area is the service area where the to-be-served location requested by the service provision request is located; here, the service data processing device can predict the S target probabilities based on the to-be-predicted features (also known as the final features) corresponding to the target service area, or can also predict the corresponding S service provision probabilities based on the to-be-predicted features of each service area, and then determine the S target probabilities of the target service area providing services based on the S service provision probabilities of each service area. The embodiment of the present application does not make any limitations in this regard.

[0134] When the service data processing device predicts the corresponding S service provision probabilities based on the to-be-predicted features of each service area, and then determines the S target probabilities of the target service area providing services based on the S service provision probabilities of each service area, continue to refer to Figure 4 , in the embodiment of the present application, Step 104 can be implemented through Steps 1041 to 1045; that is, the service data processing device predicts the S target probabilities of the S service providers providing services in the target service area based on the J area integration features and the K service area features, including Steps 1041 to 1045, and the following will explain each step separately.

[0135] Step 1041: Based on the area integration weights, restore the J area integration features to the area division result to obtain K to-be-enhanced features.

[0136] It should be noted that the service data processing device restoring the J area integration features to the area division result means restoring the J area integration features to the K service areas; here, the service data processing device can multiply the area integration weights by the J area integration features to achieve the restoration, and at this time, the obtained product is the K to-be-enhanced features.

[0137] Step 1042: Combine the K to-be-enhanced features and the K service area features to obtain K to-be-predicted features.

[0138] In the embodiments of the present application, the service data processing device may concatenate the K to-be-enhanced features and the K service area features correspondingly to obtain the corresponding K to-be-predicted features, or may also concatenate the K to-be-enhanced features, the K service area features, and the K initial area features correspondingly to obtain the corresponding K to-be-predicted features. The embodiments of the present application do not limit this.

[0139] Step 1043: Based on the K to-be-predicted features, predict the S service-providing probabilities of the S service providers for providing services in each service area.

[0140] It should be noted that the service data processing device takes the K to-be-predicted features as the final features corresponding to the K service areas respectively, and predicts the S service-providing probabilities of the S service providers for providing services in the corresponding service areas based on each to-be-predicted feature; among them, the S service providers correspond to the S service-providing probabilities one by one, and each service area corresponds to the S service-providing probabilities, indicating the probabilities of the S service providers providing services for the service-providing requests in this service area.

[0141] Step 1044: Determine the target service area where the to-be-served location is located from the area division result.

[0142] It should be noted that the service data processing device calculates the distances between the to-be-served location and each service area in the area division result, and determines the service area with the shortest distance as the target service area where the to-be-served location is located.

[0143] Step 1045: Based on the S service-providing probabilities of the S service providers for providing services in each service area, determine the S target probabilities corresponding to the target service area.

[0144] It should be noted that the service data processing device determines the S service-providing probabilities of the target service area based on the S service-providing probabilities of the S service providers for providing services in each service area, and calls the S service-providing probabilities of the target service area the S target probabilities corresponding to the target service area.

[0145] Step 105: Combine the S target probabilities to present at least one recommended service provider and the to-be-served probabilities corresponding to the at least one service provider.

[0146] In an embodiment of the present application, the service data processing device determines at least one service provider with the maximum target probability recommended from S target probabilities as the at least one recommended service provider, and calculates the probability of service to be provided based on the at least one target probability corresponding to the at least one recommended service provider. Here, when presenting the S service providers, the service data processing device presents at least one recommended service provider among the S service providers in a specified form, and presents the probability of service to be provided.

[0147] That is to say, the service data processing device presents at least one recommended service provider and the probability of service to be provided corresponding to the at least one service provider in combination with the S target probabilities, including: the service data processing device selects at least one target probability with the maximum target probability from the S target probabilities; and determines at least one service provider corresponding to the at least one target probability from the S service providers; then calculates the probability of service to be provided corresponding to the at least one target probability; finally, presents the probability of service to be provided, and presents at least one recommended service provider in a specified style.

[0148] It should be noted that the specified sample is used to highlight at least one recommended service provider among the S service providers. In addition, the service data processing device may calculate at least one failure probability of providing services corresponding to the at least one recommended service provider, and then calculate the probability of service to be provided based on the at least one failure probability of not providing services.

[0149] Exemplarily, refer to Figure 5 , Figure 5 which is an exemplary schematic diagram of service provider recommendation provided by an embodiment of the present application; as Figure 5 shown, in interface 5-1, S service providers 5-11 are presented; and, among the presented S service providers 5-11, there are two recommended service providers 5-12; in addition, the probability of service to be provided 5-13 (94%) and button 5-14 (OK) are also presented in interface 5-1, which is the probability of being provided with services when selecting the two service providers 5-12 to provide services.

[0150] It can be understood that when responding to a service provision request, first, in combination with the time to be served, the K historical service characteristics of the regional division result, and the regional location relationship, the K service area characteristics of the regional division result and the regional integration weights are determined. Then, based on the regional integration weights, the relevant service area characteristics are aggregated to learn the holistic characteristics of the regional division result. Finally, in combination with the J aggregated regional characteristics and the service area characteristics, the S target probabilities of the S service providers providing services in the target service area are predicted, so that at least one service provider can be recommended based on the S target probabilities in the target service area. Thus, through the at least one recommended service provider, the service provider with the highest service provision probability can be accurately determined, and further, the recommendation accuracy of the service provider can be improved.

[0151] In the embodiment of the present application, before the service data processing device splices the time characteristics of the time to be served and the K historical service characteristics to obtain the initial spliced characteristics, the service data processing method further includes a process of extracting time characteristics. Here, an example is given where the information corresponding to the time to be served includes the numbers of the time to be served in various time periods (for example, the first service period and the second service period). The service data processing first obtains the first service period where the time to be served is located to obtain the target first service period. Then, it obtains the first time period number of the time to be served within the target first service period. Next, it obtains the second time period number of the target first service period within the second service period. Finally, the encoding results of the first time period number and the encoding results of the second time period number are integrated into time characteristics.

[0152] It should be noted that the second service period includes multiple first service periods, that is, the cycle duration of the second service period is greater than the cycle duration of the first service period. For example, the second service period is a week, and the first service period is a day. Here, the target first service period is the first service period where the time to be served is located. And, the first service period includes multiple first cycle time periods (for example, a day includes 1340 minutes), and the second service period includes multiple second cycle time periods (for example, a week includes Monday to Sunday).

[0153] In the embodiment of the present application, the service data processing device can use any one of the one-hot encoding and the time series modeling method based on a neural network to obtain the encoding result of the first time period number (or obtain the encoding result of the second time period number). Among them, the time series modeling method based on a neural network is, for example, Long Short-Term Memory Recurrent Neural Networks (LSTM), Dilated Causal Convolution (DCC), Gated Recurrent Unit (GRU).

[0154] See Figure 6 , Figure 6 which is a flowchart of the service data processing method provided by an embodiment of the present application Figure 3 , Figure 6 and the execution subject of each step in Figure 6 is a service data processing device; as

[0155] Step 1046: In response to the next service provision request, obtain the requested service time and the requested service location.

[0156] It should be noted that when the service data processing device receives a request for the next service provision, it also receives the next service provision request; at this time, in response to the next service provision request, the service data processing device obtains the location of the service requested by the next service provision request, thereby obtaining the requested service location, and obtains the time of the service requested by the next service provision request, thereby obtaining the requested service time.

[0157] Step 1047: When the time difference between the requested service time and the time to be served is less than the time difference threshold, determine the service area where the requested service location is located from the area division result to obtain the requested service area.

[0158] It should be noted that the time difference threshold refers to the maximum time difference for reusing the S service provision probabilities of providing services in each service area at the time to be served; thus, the service data processing device obtains the time difference between the requested service time and the time to be served; if the time difference between the requested service time and the time to be served is greater than the time difference threshold, it indicates that the S service provision probabilities of providing services in each service area at the time to be served cannot be reused at this time, so the S probabilities of providing services in each service area at the service request time are obtained by the acquisition method of the S service provision probabilities of providing services in each service area at the time to be served, and the response to the next service provision request is completed based on the S probabilities. If the time difference between the requested service time and the time to be served is less than or equal to the time difference threshold, it indicates that the S service provision probabilities of providing services in each service area at the time to be served can be reused at this time; thus, the service data processing device determines the service area with the smallest distance based on the distances between the requested service location and the K service areas respectively, and the determined service area with the smallest distance is called the requested service area.

[0159] Step 1048: Determine S service providing probabilities corresponding to the requested service area based on the S service providing probabilities of S service providers in each service area.

[0160] It should be noted that the service data processing device first determines the service area corresponding to the requested service area among the K service areas, and then determines the S service providing probabilities corresponding to the requested service area based on the S service providing probabilities of S service providers in each service area.

[0161] Step 1049: Present service providing information based on the S service providing probabilities corresponding to the requested service area.

[0162] It should be noted that the process of the service data processing device presenting service providing information based on the S service providing probabilities corresponding to the requested service area is similar to the process described in step 105, and will not be elaborated herein in this embodiment of the present application.

[0163] In this embodiment of the present application, before step 101, there is also a process of obtaining the area division result; that is, before the service data processing device responds to the service providing request, this service data processing method further includes: the service data processing device first obtains the service location set within a specified historical duration before the current division moment for the area to be divided; then clusters the service location set to obtain K service location clusters; finally, determines the area corresponding to each service location cluster as a service area, and obtains K service areas corresponding to the K service location clusters.

[0164] It should be noted that each service location in the service location set is a location where a service is provided; each clustering cluster in the clustering result is a service location cluster. In addition, the area division result can be periodically divided; that is, the service data processing device periodically divides the area to be divided based on each service location within the specified historical duration.

[0165] In this embodiment of the present application, before the service data processing device determines the K service area characteristics and area integration weights by combining the to-be-served time, the K historical service characteristics, and the area location relationship of the area division result, this service data processing method further includes a process of obtaining the area location relationship: the service data processing device first determines the area center location of each service area in the area division result; then determines the area distance between any two service areas in the area division result based on the area center location; then determines two service areas with an area distance less than the adjacent distance threshold as adjacent service areas; finally, determines the area location relationship of the area division result based on the adjacent service areas. That is, the area location relationship indicates whether there is an adjacent relationship between any two service areas among the K service areas, and there is an adjacent relationship between adjacent service areas.

[0166] In an embodiment of the present application, the prediction of S target probabilities can be achieved through a service prediction model, which is used to predict the probability that each service provider provides services in each service area.

[0167] See Figure 7 , Figure 7 which is a schematic flowchart of the model training method provided by an embodiment of the present application. Figure 7 The execution subject of each step in Figure 7 is a service data processing device; as

[0168] Step 106: Extract historical service sample features of N service provider sample parties providing services in each sample area in the area division sample result relative to the time of the to-be-served sample.

[0169] It should be noted that the time of the to-be-served sample refers to the time when the probability of providing services is to be predicted during model training. Thus, the time of the to-be-served sample is a historical time. The area division sample result is the division result of the to-be-divided sample area. The to-be-divided sample area may be the same as or different from the to-be-divided area. The division process of the area division sample result is similar to that of the area division result. The acquisition of historical service sample features is similar to the acquisition of historical service features, which will not be elaborated in this embodiment of the present application.

[0170] Step 107: Use the to-be-trained model to predict the time of the to-be-served sample, the historical service sample features, and the sample area location relationship, and obtain N prediction probabilities of N service provider sample parties providing services in each sample area.

[0171] It should be noted that the to-be-trained model is a neural network model to be trained for predicting the probability of providing services. N is an integer greater than 1, and N may be equal to or different from S. Here, the acquisition of N prediction probabilities is similar to the acquisition process of S service providing probabilities, which will not be elaborated in this embodiment of the present application.

[0172] Step 108: Calculate the predicted service dispatch quantity based on the service dispatch sample quantity and the prediction probability of each service provider sample party in each sample area.

[0173] It should be noted that the service data processing device can use the product of the service dispatch sample quantity and the prediction probability as the predicted service dispatch quantity.

[0174] Step 109: Calculate the loss function value based on the difference between the predicted service dispatch quantity and the service dispatch sample quantity of each service provider sample party in each sample area.

[0175] In the embodiments of the present application, the service dispatch sample size represents the actual order dispatch quantity; the service data processing device can obtain the difference between the service dispatch prediction quantity and the service dispatch sample size by acquiring the difference therebetween, and can also obtain the difference between the two by acquiring the ratio between the service dispatch prediction quantity and the service dispatch sample size, etc. The embodiments of the present application do not limit this.

[0176] Step 110: Train the model to be trained based on the loss function value to obtain a service prediction model.

[0177] It should be noted that the service data processing device performs backpropagation in the model to be trained based on the loss function value to adjust the model parameters of the model to be trained; here, the training of the model to be trained can be carried out iteratively, and the training ends when the training end condition is met, and the model trained at this time is used as the service prediction model. Among them, the training end condition can be reaching the accuracy index threshold, or reaching the iteration number threshold, or reaching the iteration duration threshold, or a combination of the above, etc. The embodiments of the present application do not limit this.

[0178] In the embodiments of the present application, after the service data processing device presents at least one recommended service provider and the service probability to be served corresponding to at least one service provider in combination with S target probabilities, the service data processing method further includes: the service data processing device obtains at least one adjusted service provider in response to a selection adjustment operation for the service probability to be served; then, determines at least one adjusted target probability corresponding to at least one adjusted service provider from the S target probabilities; calculates the current service probability corresponding to at least one adjusted target probability; and finally, presents the current service probability and presents at least one adjusted service provider in a selected style.

[0179] It should be noted that since the service data processing device presents the service probability to be served and the S target probabilities, an operation for adjusting the selected service provider can be received for the presented information, which can be an increase or decrease operation based on at least one recommended service provider. The current service probability represents the probability that at least one adjusted service provider provides a service.

[0180] In an embodiment of the present application, after the service data processing device determines K service area features and area integration weights in combination with the service time to be served, K historical service features, and the area location relationship of the area division result, the service data processing method further includes: The service data processing device presents the area integration weights in a specified graph style, and the specified graph style includes a first graph style, a second graph style, and a third graph style. The first graph style is used to present the area integration weights from the dimensions of the area division result and the area cluster. The second graph style is used to present the area integration weights based on the area cluster corresponding to the highest weight. The third graph style is used to present the area integration weights from the longitude and latitude dimensions.

[0181] It should be noted that the first graph style can be implemented by using the area cluster and the area division result as the coordinate axes respectively; for the second graph style, it is possible to first determine the area cluster to which each service area belongs, and while presenting the structure of the area division result, the same area cluster adopts the same presentation form, for example, the same shape, the same color, etc.; for the third graph style, it is possible to first determine the area cluster to which each service area belongs, and while presenting the structure of the area division result based on the geographical location, the same area cluster adopts the same presentation form; here, by visually presenting the area integration weights, the intuitiveness of the association between regions can be improved.

[0182] Next, an exemplary application of the embodiment of the present application in an actual application scenario will be described. This exemplary application describes the process of predicting the order acceptance rates corresponding to each online car-hailing capacity provider in the online car-hailing service scenario and determining the online car-hailing capacity provider to which an order is to be assigned in combination with each order acceptance rate. It is easy to know that the service data processing method provided by the embodiment of the present application can be applied to any scenario of selecting a service provider from multiple service providers to provide services. Here, the online car-hailing service scenario is taken as an example for description.

[0183] It should be noted that this exemplary application predicts the order acceptance rate in an artificial intelligence manner; that is, the order acceptance rate prediction is realized by training an order acceptance rate prediction model. Refer to Figure 8 , Figure 8 is an exemplary order acceptance rate prediction flowchart provided by the embodiment of the present application; as Figure 8 shown, this exemplary order acceptance rate prediction process includes steps 201 to 204, and each step will be described separately below.

[0184] Step 201: Based on the spatial area division result of the sample area to be divided (referred to as the area division sample result), construct an undirected graph.

[0185] It should be noted that the to-be-partitioned sample area is partitioned into K (for example, 50) non-overlapping spatial areas (referred to as sample areas). Thus, the spatial area partitioning result is K non-overlapping spatial areas. Here, an undirected graph is constructed based on the adjacency relationship between the K non-overlapping spatial areas.

[0186] When partitioning the to-be-partitioned sample area, the boarding positions within the to-be-partitioned sample area in the past month are extracted. The boarding positions can be represented by longitude and latitude. The extracted boarding positions are clustered (for example, K-Means clustering), and all the extracted boarding positions are partitioned into K clusters (referred to as service location clusters in application). The area corresponding to each cluster is a spatial area. Thus, K spatial areas are partitioned. The clustering center of each cluster is obtained. When any to-be-served position on the given map is given, the cluster where the clustering center with the shortest straight-line distance to the to-be-served position among the K clusters is located is the classification result of the to-be-served position. Here, if the straight-line distance between two clustering centers is less than ε (referred to as the adjacent distance threshold, for example, 10 kilometers (km)), it is determined that there is an adjacency relationship between the two clusters corresponding to the two clustering centers. When constructing the undirected graph, each cluster is used as a node in the undirected graph structure, and the edges between the nodes in the undirected graph structure are determined according to the determined adjacency relationship of the clusters. Thus, the construction of the undirected graph is completed. Here, the to-be-partitioned area can also be partitioned into geometric grids such as regular hexagons to achieve the partitioning of the to-be-partitioned area.

[0187] Step 202: Extract information based on the spatial area partitioning result.

[0188] It should be noted that extracting information based on the spatial area partitioning result means extracting information related to the order receiving of each online car-hailing capacity provider for each to-be-predicted time and each spatial area.

[0189] It should be noted that in the embodiments of the present application, the order receiving rate is obtained in an aggregated manner. For example, with one minute as the prediction interval, the aggregated time period corresponding to the first 15 minutes and the last 15 minutes before the to-be-predicted time is used as the to-be-predicted time period, and the order receiving rate of this to-be-predicted time period is predicted. In addition, since the aggregated service platform corresponds to multiple online car-hailing capacity providers, the prediction target is multi-dimensional, and the dimension is the number of online car-hailing capacity providers.

[0190] It can be understood that when predicting the order receiving rate using a prediction time interval smaller than the prediction interval threshold, one order receiving rate is predicted per minute, which can be sensitive to the real-time change situation of the input features; while using an aggregated time period larger than the aggregation duration threshold can reduce the influence of contingency.

[0191] In addition, in terms of information extraction, periodic information (referred to as the periodic service information sequence during application) and short-time time series information (referred to as the time period service information sequence during application) are combined. In terms of periodic information extraction, starting from the date where the time to be predicted is located, it is deduced forward on a daily basis, and the order dispatch volume and order acceptance volume of each online car-hailing capacity provider within the first 15 minutes and the last 15 minutes corresponding to the time to be predicted on different historical dates in the same spatio-temporal region are extracted. In terms of short-time time series information extraction, the sliding window method is adopted, with a 30-minute sliding window and a 5-minute sliding step size, and a 60-minute sequence length before the time to be predicted (referred to as the specified sequence length during application) is used to extract the order dispatch volume and order acceptance volume of each online car-hailing capacity provider in the same spatio-temporal region.

[0192] Exemplarily, refer to Figure 9 , Figure 9 which is an exemplary information extraction schematic diagram provided by an embodiment of the present application; as Figure 9 shown, the prediction target 9-1 is the order acceptance rate of each online car-hailing capacity provider within the first 15 minutes and the last 15 minutes of the time to be predicted; the periodic information 9-2 is the order dispatch volume and order acceptance volume of each online car-hailing capacity provider within the first 15 minutes and the last 15 minutes corresponding to the time to be predicted on the day before, two days before, and seven days before the date where the time to be predicted is located; the short-time time series information 9-3 is the order dispatch volume and order acceptance volume of each online car-hailing capacity provider within multiple 30-minute periods before the time to be predicted in a day.

[0193] In addition, in terms of information extraction, the time period number is also combined; in terms of time period number extraction, the minute number of the time to be predicted in a day (ranging from 0 to 1339, referred to as the first time period number during application) and the week number (ranging from 0 to 6, referred to as the second time period number during application) are extracted.

[0194] Step 203, train the model based on the extracted information.

[0195] It should be noted that by extracting information, in an undirected graph including K nodes, each node includes its respective time period number, periodic information, and short-time time series information. Here, the undirected graph G is represented as (A, X, P, Q); where A ∈ {0, 1} K×K is the adjacency matrix of the undirected graph (referred to as the sample area position relationship); is the short-time time series information of K nodes, the length of the short-time time series information of each node is C, and the dimension of each short-time time series information is d1; is the periodic information of K nodes, and the dimension of the periodic information of each node is d2; is the time period number of K nodes, and the dimension of the time period number of each node is d3. Correspondingly, the order acceptance rate of the undirected graph G is denoted as π ∈ (0, 1) K×S, each node corresponds to an S-dimensional vector, where S is the number of online car-hailing capacity providers, and π ks represents the order acceptance rate of the s-th (s ∈ [1, S]) online car-hailing capacity provider of the k-th (k ∈ [1, K]) node, and π ks ∈ (0, 1). Based on the extracted information, a Temporal Hierarchical Graph Neural Network (THGNN, referred to as the model to be trained) is trained to learn the mapping information from the undirected graph to the order acceptance rate.

[0196] See Figure 10 , Figure 10 is an exemplary model structure diagram provided by an embodiment of the present application; as Figure 10 shown, in the temporal hierarchical graph neural network model 10-1, the time period number 10-21 is embedded and represented (denoted as f embedding , which is called the time feature in application) through the embedding module 10-111. Here, the week number in the time period number 10-21 is represented in the form of one-hot encoding, and the minute number in the time period number 10-21 is embedded and represented by using an embedding layer to map the discrete minute number into a continuous vector space; the short-term time series information 10-22 is transformed (denoted as f GRU ) through a Gated Recurrent Unit (GRU) 10-112 to obtain the corresponding temporal feature (which is called the encoded result of the time period service information sequence in application); here, the embedding representation of the time period number 10-21, the periodic information 10-23, and the temporal feature of the short-term time series information 10-22 are concatenated through a fully connected layer 10-12 (including a linear transformation and a ReLU activation function), and the node representations 10-3 (denoted as H, which is called the initial regional feature in application) of K nodes are determined (denoted as f c ) as shown in Equation (1).

[0197] H = f c (f GRU (X) ⊕ P ⊕ f embedding (Q)) (1);

[0198] It should be noted that processing the short-term time series information 10-22 based on a neural network can capture the data relationships and patterns in the short-term time series information. Here, long short-term memory network dilated causal convolution, gated recurrent unit, etc. can be used to process the short-term time series information 10-22. Exemplarily, the gated recurrent unit is taken as an example for illustration. Additionally, by using the gated recurrent unit, for the short-term time series information 10-22, the flow and update of information can be controlled.

[0199] Continue to refer to Figure 10 , the node representation module 10-131 (Embedding DiffBlock1) is used to perform feature representation (denoted as f embed1 ) on the node representation 10-3 and the adjacency matrix 10-24 (denoted as A), obtaining the node representation 10-41 (denoted as Z); the assignment module 10-132 (Pooling DiffBlock + Softmax) is used to process the node representation 10-3 and the adjacency matrix 10-24 (denoted as softmax(f pool )) to obtain the assignment matrix 10-42 (denoted as M). Then, through the graph structure pooling module 10-14, graph structure pooling operations are performed on the node representation 10-41, the assignment matrix 10-42, and the adjacency matrix 10-24, obtaining the pooling operation results, including the node cluster representation 10-51 (denoted as H′, called the region integration feature in application) and the node cluster adjacency matrix 10-52 (denoted as A′), as shown in equations (2) and (3).

[0200]

[0201]

[0202] Among them, f embed1 and f pool respectively include two graph attention network (GAT) structures. The two GAT structures adopt the dense connections method. Here, f embed1 is taken as an example for illustration, as shown in equation (3).

[0203] f embed1 (A, H) = H1 ⊕ H2, H1 = f GAT1 (A, H), H2 = f GAT2 (A, H1) (3);

[0204] Among them, f GAT1 and f GAT2 are for f embed1The two GAT structures included. H1 is called the enhanced region feature when applied, and H2 is called the region feature to be connected when applied.

[0205] It can be understood that by adopting a dense connection method for the two GAT structures, the linear transformation layer can be directly connected to the outputs of all previous graph convolutional layers, enabling information to propagate flexibly in the network, reducing information loss, and reducing the probability of gradient disappearance.

[0206] See Figure 11 , Figure 11 is a schematic structural diagram of an exemplary node representation module provided by an embodiment of the present application; as Figure 11 shown, the node representation module 10-131 is used to perform feature representation on the node representation 10-3 and the adjacency matrix 10-24 to obtain the node representation 10-41. In the node representation module 10-131, the node representation 10-3 and the adjacency matrix 10-24 sequentially pass through the graph neural network 11-1 (Graph Neural Network, GNN) and the activation function layer 11-2 (ReLU) to obtain the feature 11-31 (denoted as H1); the feature 11-31 is combined with the adjacency matrix 10-24 and sequentially passes through the graph neural network 11-3 and the activation function layer 11-4 to obtain the feature 11-32 (denoted as H2); finally, the fully connected layer 11-5 (Linear Transformation) is used to integrate the feature 11-31 and the feature 11-32 to obtain the node representation

[0207] It should be noted that the graph neural network can model the nodes and edges in the graph to capture the structure and relationships of the graph. There are various types of graph neural networks, and here the graph convolutional network (Graph Convolutional Network, GCN) can be used to obtain the neighbor relationships between the red nodes in the graph. The graph attention network can also be used to learn the relationships between nodes and their neighbors to obtain reasonable local representations.

[0208] In addition, the pooling operation of the graph structure refers to allocating the node representations Z of each layer to at least one node cluster H (l) (called the feature to be allocated when applied) to extract the overall features of the graph structure. The pooling operation of the graph structure is shown in Equation (4). (lJ1) (In application, it is called the (a - 1)-th allocation weight) for each node in layer l; f

[0209]

[0210] where A (l) is the adjacency matrix of each node in layer l; f l，embed and f embed1The processing process is similar; K l represents the number of nodes in layer l, d l is the dimension of each node representation; K l+1 is the number of upper-level clusters (referred to as the specified number traversed). At this time, the adjacency matrix A of the l-th layer (l) also changes to A (l+1) (referred to as the a-th allocation weight in application), and the corresponding shape changes from K l ×K l to K l+1 ×K l+1 , as shown in Equation (5).

[0211]

[0212] The allocation matrix M (l) is as shown in Equation (6).

[0213] M (l) = softmax(f l，pool (A (l) , H (l) )) (6);

[0214] where the processing processes of softmax(f l，pool ) and softmax(f pool ) are similar.

[0215] It should be noted that the implementation structures of f l，embed and f l，pool are the same (for example, both are DiffBlock, including two GAT structures and a fully connected layer); f l，pool uses the same input tensor and network structure as f l，embed to learn the allocation matrix. The difference is that the data dimension in the linear transformation is the number of node clusters, and the activation function (such as the Softmax function) is applied row by row to obtain the probability that each node is assigned to different node clusters. The number of node clusters is a hyperparameter of f pool , which can be decreased proportionally according to the initial number of nodes, or the value of each f l，pool operation can be specified.

[0216] Continue to refer to Figure 10 , and the result of the pooling operation is transformed through the linear transformation module 10 - 15 (Embedding DiffBlock2) to obtain the abstract representation of each node cluster (denoted as f embed(A′, H′), which is called the integrated feature of J regions in application), and based on the assignment matrix 10-42, the abstract representation is restored to each node to obtain the restored node representation 10-6 (which is called the K features to be enhanced in application), denoted as Z′, as shown in Equation (7).

[0217] Z′ = M × f embed2 (A′, H′) (7);

[0218] where f embed2 includes two GAT structures, and the two GAT structures are densely connected.

[0219] Finally, the probability prediction module 10-16 (including Linear Transformation and Sigmoid activation function) is used to predict the concatenation result of the restored node representation 10-6, node representation 10-3, and node representation 10-41, to obtain the order acceptance rate 10-7 of each online car-hailing capacity provider at each node in the time period to be predicted, as shown in Equation (8).

[0220] π = f0(Z′ ⊕ H ⊕ Z) (8);

[0221] where Z′ ⊕ H ⊕ Z represents what is called the K features to be predicted in application.

[0222] Here, the loss function Loss is as shown in Equation (9).

[0223]

[0224] where O k，s is the order dispatch volume of the s-th online car-hailing capacity provider at the k-th node; R k，s is the order acceptance volume of the s-th online car-hailing capacity provider at the k-th node.

[0225] It can be understood that the loss function Loss takes nodes and online car-hailing capacity providers as units, and can improve the accuracy of order acceptance rate prediction.

[0226] Exemplarily, refer to Figure 12 , Figure 12 which is an exemplary logical schematic diagram provided by an embodiment of the present application; as Figure 12 shown, first, an undirected Figure 12-1 is constructed based on step S201, and information of each node in the undirected Figure 12-1 is extracted based on step S202 (as shown by the three connected cubes in Figure 12 ). Then, based on the information of the nodes and the adjacency matrix of the undirected Figure 12-1 , the node representation and the assignment matrix are obtained. Different node clusters represented by different filling methods are shown in Information 12-2, and the undirectedFigure 12-1 The nodes are assigned to the corresponding node clusters, and this assignment method is a soft partitioning method. Subsequently, based on the assignment matrix, the obtained node representations are assigned to obtain the node cluster representations and node cluster adjacency matrices as shown in Information 12-3. Then, based on the node cluster representations and node cluster adjacency matrices, the abstract representations of each node cluster as shown in Information 12-4 are obtained. Finally, the abstract representations of each node cluster are restored to each node of the undirected Figure 12-1 to obtain the restored node representations as shown in Information 12-5, and combined with the restored node representations and the node representations of the undirected Figure 12-1 to predict the order acceptance rates of different freight capacity providers for each node through a fully connected network.

[0227] Step 204: Based on the trained model, for each order acceptance rate of each online car-hailing freight capacity provider within a specified spatio-temporal range.

[0228] It should be noted that after obtaining the order acceptance rates of each online car-hailing freight capacity provider within a specified spatio-temporal range, refer to Figure 13 , Figure 13 is an exemplary order acceptance rate presentation schematic diagram provided by an embodiment of the present application; as Figure 13 shown, in Interface 13-1, there are S online car-hailing freight capacity providers 13-11 presented; and, among the presented S online car-hailing freight capacity providers 13-11, there are two recommended online car-hailing freight capacity providers 13-12; in addition, in Interface 13-1, there are also presented order acceptance rate prompt information 13-13 (current estimated response rate 94%, adding a vehicle type (referred to as an online car-hailing freight capacity provider) will respond faster, where the current estimated response rate 94% is the order acceptance rate corresponding to dispatching orders to the two online car-hailing freight capacity providers 13-12) and a button 13-14 (call simultaneously). In addition, in Interface 13-1, there are also presented a tab 13-15, consumption estimation information 13-16, and prompt information 13-17.

[0229] It should be noted that the online car-hailing aggregation service platform recommends online car-hailing freight capacity providers based on the order acceptance rate (for example, Figure 13 the two online car-hailing freight capacity providers 13-12 selected in). For the current estimated response rate calculated based on the order acceptance rate, if the current estimated response rate is lower than the response rate threshold, the current estimated response rate can be increased by receiving the tick operation of adding an online car-hailing freight capacity provider, thereby improving the service provision efficiency.

[0230] The following explains the spatial dynamic soft partitioning.

[0231] It should be noted that the value of M tkm indicates the probability that the spatial region k at time t belongs to the node cluster m, and the Manhattan distance and between Explicitly measure the similarity of the supply and demand characteristics of the online car-hailing service in spatial region k1 at time t1 and spatial region k2 at time t2, where the smaller the Manhattan distance, the greater the similarity; thus indicating that the allocation matrix describes the interpretable holistic representation existing between the representations of each node.

[0232] Here, the allocation matrix can be visualized. Refer to Figure 14 , Figure 14 which is an exemplary visualization result diagram of the allocation matrix provided by an embodiment of the present application; as Figure 14 shown, the visualization from Figure 14-1 to visualization Figure 14-16 sequentially describe the probabilities that 50 nodes corresponding to 16 minutes (from the 0th minute to the 15th minute) in the early morning period of the 34th day belong to 10 node clusters. That is to say, in the visualization from Figure 14-1 to visualization Figure 14-16 , each visualization diagram corresponds to a moment to be predicted, and in each visualization diagram, the horizontal axis number is the number of each node cluster, the vertical axis number is the number of each node in the undirected graph, and the color of the grid at the cross position of the horizontal and vertical axis coordinates represents the probability that the node belongs to the node cluster, the darker the color, the greater the probability, and the lighter the color, the smaller the probability.

[0233] In addition, based on the visualization result of the allocation matrix shown in Figure 14 , refer to Figure 15 , Figure 15 which is an exemplary structural schematic diagram of the allocation matrix provided by an embodiment of the present application; as Figure 15 shown, for each node, select the node cluster with the highest probability, and based on the color of the node cluster, determine the node coloring of the color map structure of the node in the undirected graph, obtaining the structure corresponding one-to-one with the visualization from Figure 14 in Figure 14-1 to visualization Figure 14-16 ; among them, in the structure from Figure 15-1 to structure Figure 15-16 , the node cluster corresponding to the node filled in the large square includes the largest number of nodes, which is called the optimal node cluster. Figure 15-1 to structure Figure 15-16 ; among them, in the structure from

[0234] In addition, based on the structural schematic diagram of the allocation matrix shown in Figure 15 , refer to Figure 16 , Figure 16 which is an exemplary two-dimensional schematic diagram of the allocation matrix provided by an embodiment of the present application; as Figure 16 shown, by restoring the structure from Figure 15 in Figure 15-1 to structure Figure 15-16 to a two-dimensional plane with longitude and latitude as the horizontal and vertical coordinates, the corresponding two-dimensional schematic diagrams from Figure 16-1 to two-dimensional schematicFigure 16-16 ; In each two-dimensional schematic diagram Figure 16-1 to Figure 16-16 of the two-dimensional schematic diagrams, each point represents the central position of each spatial region, and different colors correspond to different node clusters.

[0235] Similarly, referring to Figures 17 to 19 , Figures 17 to 19 describes a related schematic diagram of the peak vehicle usage period of a day. Referring to Figure 17 , Figure 17 is another exemplary visualization result graph of the allocation matrix provided by the embodiment of the present application; as Figure 17 shown, the visualization Figure 17-1 to Figure 17-16 successively describe the probabilities that 50 nodes corresponding to 16 minutes of the peak vehicle usage period of a day belong to 10 node clusters.

[0236] Based on Figure 17 , referring to Figure 18 , Figure 18 is another exemplary structured schematic diagram of the allocation matrix provided by the embodiment of the present application; as Figure 18 shown, for each node, the node cluster with the highest probability is selected, and based on the color of the node cluster, the node coloring of the node color graph structure in the undirected graph is determined, obtaining a structure Figure 17 in Figure 17-1 to Figure 17-16 that corresponds one-to-one with the visualization Figure 18-1 to Figure 18-16 ; among them, in the structure Figure 18-1 to Figure 18-16 , the nodes corresponding to the nodes filled with slashes and small squares have the largest number of nodes included in the node cluster, which is the optimal node cluster.

[0237] Based on Figure 18 , referring to Figure 19 , Figure 19 is another exemplary two-dimensional schematic diagram of the allocation matrix provided by the embodiment of the present application; as Figure 19 shown, by restoring the structure Figure 18 in Figure 18-1 to Figure 18-16 to a two-dimensional plane with longitude and latitude as the horizontal and vertical coordinates, the corresponding two-dimensional schematic diagrams Figure 19-1 to Figure 19-16 are obtained.

[0238] It should be noted that the allocation matrix varies at different times, indicating that the allocation matrix can improve the accuracy of the final node representation of the nodes. In addition, by visualizing the allocation matrix, the detection of each spatial region can be achieved.

[0239] The beneficial effects of the present application will be described below.

[0240] Exemplarily, based on the model training process described in step 203, the data of 20 working days is used as the training set and the validation set, and the data of the subsequent nearest 3 working days is used as the test set.

[0241] It should be noted that three evaluation indicators are used in the embodiments of the present application to evaluate the prediction effect; among them, the first evaluation indicator is formula (9), the second evaluation indicator is the MAE of the order receiving rate of the test set (as shown in formula (10)), and the third evaluation indicator is the logarithmic probability measure of the test set (as shown in formula (11)).

[0242]

[0243]

[0244] Among them, C represents the permutation and combination calculation. Formula (11) describes that by introducing the binomial distribution, the order receiving situation of the same online car-hailing capacity provider within the same spatio-temporal range is regarded as a Bernoulli trial, and the actual order dispatch volume and the predicted order receiving rate are used as the parameters of the binomial distribution to calculate the logarithmic probability of the actual order receiving volume in this discrete distribution.

[0245] It should be noted that in the embodiments of the present application, based on the above three evaluation indicators, the order receiving rate prediction method provided in the embodiments of the present application (referred to as the method of the present application) is compared with the baseline prediction methods 1 to 4. Among them, the baseline prediction methods 1 to 4 are the prediction methods described as follows.

[0246] For the baseline prediction method 1, taking the time to be predicted as the benchmark, calculate the order receiving frequency of each online car-hailing capacity provider within the time range [35 minutes before, 5 minutes before] to obtain the predicted value of the order receiving rate. If the order dispatch volume is 0, the predicted value of the order receiving rate is taken as 0.

[0247] For the baseline prediction method 2, taking the time to be predicted as the benchmark, obtain the order dispatch volume and the order receiving volume of each online car-hailing capacity provider within the current range [35 minutes before, 5 minutes before]. If the total order dispatch volume is less than the threshold (for example, 30), aggregate the order dispatch volumes of the nearest adjacent areas; if the total order dispatch volume is still less than the threshold, aggregate the order dispatch volumes of the second-order adjacent areas, and so on. Calculate the order receiving frequency of the aggregated area to obtain the predicted value of the order receiving rate. If the order dispatch volume is 0, the predicted value of the order receiving rate is taken as 0.

[0248] For the baseline prediction method 3, taking the time to be predicted as the benchmark, extract the period number, cycle information, and short-time time series information, and splice them into an input of several dimensions to predict the predicted value of the order receiving rate.

[0249] The baseline prediction method 4 takes the time to be predicted as the benchmark, extracts the period number, cycle information, and short-term time series information, inputs the time series information into the GRU, and then concatenates the GRU output with the cycle information and period number to form multi-dimensional features. Then, an undirected graph is constructed, and the multi-dimensional features are processed based on the adjacency matrix of the undirected graph to obtain the feature representation of each node. Finally, the predicted order acceptance rate value is predicted based on the feature representation of the node.

[0250] The comparison results of baseline prediction methods 1 to 4 are shown in Table 1.

[0251] Table 1

[0252]

[0253]

[0254] As can be seen from Table 1, in terms of the evaluation index values of the three evaluation indexes, the method of the present application is lower than baseline prediction methods 1 to 4, indicating that the method of the present application can improve the prediction accuracy of the order acceptance rate.

[0255] It can be understood that by learning the spatio-temporal correlation of service supply and demand from historical service data and using the allocation matrix as an intermediate result, the embodiments of the present application can explicitly display the correlation relationship of the supply and demand of online car-hailing in different spatial regions during the same period, and realize the prediction of the order acceptance rate of multiple online car-hailing capacity providers. In addition, when the order dispatch volume in the spatio-temporal region is 0, the embodiments of the present application can still construct the model; and constructing the same model for multiple online car-hailing capacity providers can improve the model training efficiency. In addition, by extracting information such as the period number, short-term time series information, and cycle information, and introducing time series modeling, graph neural network, and hierarchical representation structure based on the neural network framework, the neural network model of the embodiments of the present application can automatically learn time correlation, spatial local relationship, and overall representation, improve the prediction accuracy of the order acceptance rate, thereby improving the accuracy of recommending online car-hailing capacity providers by the online car-hailing aggregation service platform, and thus improving the online car-hailing service efficiency.

[0256] Next, the exemplary structure of the service data processing device 455 provided by the embodiments of the present application implemented as software modules will be continued. In some embodiments, as Figure 2 shown, the software modules stored in the service data processing device 455 in the memory 450 may include:

[0257] The feature extraction module 4551 is configured to, in response to a service provision request, extract historical service features of S service providers providing services in each service area in the area division result with respect to the time to be served, and obtain K historical service features, where both S and K are integers greater than 1;

[0258] The feature extraction module 4551 is further configured to determine K service area features and area integration weights by combining the to-be-served time, the K historical service features, and the area position relationship of the area division result.

[0259] The feature integration module 4552 is configured to integrate the K service area features based on the area integration weights to obtain J area integration features, where J is a positive integer less than K.

[0260] The probability prediction module 4553 is configured to predict S target probabilities that the S service providers provide services in the target service area based on the J area integration features and the K service area features.

[0261] The service recommendation module 4554 is configured to present at least one of the recommended service providers and the to-be-served probabilities corresponding to at least one of the service providers in combination with the S target probabilities.

[0262] In an embodiment of the present application, the feature extraction module 4551 is further configured to splice the time feature of the to-be-served time and the K historical service features to obtain an initial spliced feature; perform a linear transformation on the initial spliced feature to obtain K initial area features; enhance the K initial area features based on the area position relationship of the area division result to obtain the K service area features; combine the K initial area features and the area position relationship to obtain the area integration weights.

[0263] In an embodiment of the present application, the feature extraction module 4551 is further configured to obtain a target first service cycle where the to-be-served time is located to obtain a target first service cycle; obtain a first time period number of the to-be-served time within the target first service cycle; obtain a second time period number of the target first service cycle within a second service cycle, where the second service cycle includes a plurality of the first service cycles; integrate the encoding results of the first time period number and the encoding result of the second time period number into the time feature.

[0264] In an embodiment of the present application, the feature extraction module 4551 is further configured to perform a first attention process on the K initial area features based on the area position relationship of the area division result to obtain K enhanced area features; perform a second attention process on the K enhanced area features based on the area position relationship to obtain K area features to be connected; connect the K enhanced area features and the K area features to be connected to obtain the K service area features.

[0265] In the embodiment of the present application, the feature integration module 4552 is further configured to traverse a specified number of sequences through iteration a, and perform the following processing on each traversed specified number, where a is a positive integer variable: allocate multiple (a - 1)-th features to be allocated based on the (a - 1)-th allocation weight to obtain the a-th features to be allocated of the specified number. When a is 1, the (a - 1)-th allocation weight is the region integration weight, and the multiple (a - 1)-th features to be allocated are K service region features; when the traversed specified number is different from the last specified number, determine the a-th positional relationship by combining the (a - 1)-th allocation weight and the (a - 1)-th positional relationship, and determine the a-th allocation weight by combining the a-th positional relationship and the a-th features to be allocated of the specified number, and continue to iterate a. When a is 1, the (a - 1)-th positional relationship is the region positional relationship; when the traversed specified number is the last specified number, end the iteration of a, and determine the a-th features to be allocated of the specified number when ending the iteration of a as J region integration features.

[0266] In the embodiment of the present application, the probability prediction module 4553 is further configured to restore J region integration features to the region division result based on the region integration weight to obtain K features to be enhanced; combine the K features to be enhanced and the K service region features to obtain K features to be predicted; predict S service providing probabilities of S service providers providing services in each service region based on the K features to be predicted; determine the target service region where the location to be served is located from the region division result; and determine S target probabilities corresponding to the target service region based on the S service providing probabilities of S service providers providing services in each service region.

[0267] In the embodiment of the present application, the service data processing device 455 further includes a probability multiplexing module 4555, configured to obtain a requested service time and a requested service location in response to a next service providing request; when the time difference between the requested service time and the time to be served is less than a time difference threshold, determine the service region where the requested service location is located from the region division result to obtain a requested service region; determine S service providing probabilities corresponding to the requested service region based on the S service providing probabilities of S service providers providing services in each service region; and present service providing information based on the S service providing probabilities corresponding to the requested service region.

[0268] In an embodiment of the present application, the feature extraction module 4551 is further configured to, in response to the service provision request, obtain, relative to the time to be served, a time period service information sequence of S service providers providing services in the service area and a specified sequence length; obtain, relative to the time to be served, a periodic service information sequence of S service providers providing services in the service area and a specified cycle sequence; and determine the encoded results of the time period service information sequence and the encoded results of the periodic service information sequence as the historical service features, thereby obtaining K historical service features.

[0269] In an embodiment of the present application, the feature extraction module 4551 is further configured to, in response to the service provision request, use the time to be served as the end time of the time series, and determine the start time of the time series based on the specified sequence length; slide from the start time of the time series to the end time of the time series with a specified sliding window and a specified sliding step length to obtain a time period sequence to be extracted; for each time period to be extracted in the time period sequence to be extracted, combine the first service dispatch volume and the first service provision volume of S service providers in the service area respectively as the time period service information, thereby obtaining the time period service information sequence corresponding to the specified sequence length.

[0270] In an embodiment of the present application, the feature extraction module 4551 is further configured to, in each specified cycle of the specified cycle sequence, determine the time to be processed corresponding to the time to be served; obtain a historical service time period centered on the time to be processed based on a specified range length; and within the historical service time period, combine the second service dispatch volume and the second service provision volume of S service providers in the service area as the periodic service information, thereby obtaining the periodic service information sequence corresponding to the specified cycle sequence.

[0271] In an embodiment of the present application, the service data processing device 455 further includes a region division module 4556, configured to, for the region to be divided, obtain a set of service locations within a specified historical duration before the current division moment; cluster the set of service locations to obtain K service location clusters; and determine the region corresponding to each service location cluster as the service area, thereby obtaining K service areas corresponding to the K service location clusters, where the region division result includes K service areas.

[0272] In an embodiment of the present application, the region division module 4556 is further configured to determine the regional center position of each service area in the region division result; determine the regional distance between any two service areas in the region division result based on the regional center position; determine two service areas with a regional distance less than the adjacent distance threshold as adjacent service areas; and determine the regional position relationship of the region division result based on the adjacent service areas.

[0273] In the embodiment of the present application, the prediction of the S target probabilities is implemented by a service prediction model. The service data processing device 455 further includes a model training module 4557, which is used to extract historical service sample features for providing services in each sample area in the area division sample result relative to the time of the sample to be served; use the model to be trained to predict the time of the sample to be served, the historical service sample features, and the sample area position relationship, and obtain N prediction probabilities for N service-providing sample parties to provide services in each sample area, where the model to be trained is a neural network model to be trained for predicting the probability of providing services, and N is an integer greater than 1; calculate the predicted service dispatch volume based on the service dispatch sample volume of each service-providing sample party in each sample area and the prediction probability; calculate the loss function value based on the difference between the predicted service dispatch volume and the service dispatch sample volume of each service-providing sample party in each sample area; train the model to be trained based on the loss function value to obtain the service prediction model.

[0274] In the embodiment of the present application, the service recommendation module 4554 is further used to select at least one of the target probabilities with the largest target probability from the S target probabilities; determine at least one of the service-providing parties corresponding to at least one of the target probabilities from the S service-providing parties; calculate the probability of the sample to be served corresponding to at least one of the target probabilities; present the probability of the sample to be served, and present at least one of the recommended service-providing parties in a specified style.

[0275] In the embodiment of the present application, the service recommendation module 4554 is further used to obtain at least one adjusted service-providing party in response to a selection adjustment operation for the probability of the sample to be served; determine at least one adjusted target probability corresponding to at least one adjusted service-providing party from the S target probabilities; calculate the current service probability corresponding to at least one of the adjusted target probabilities; present the current service probability, and present at least one of the adjusted service-providing parties in the specified style.

[0276] In the embodiment of the present application, the service recommendation module 4554 is further used to present the area integration weight in a specified graph style, where the specified graph style includes a first graph style, a second graph style, and a third graph style. The first graph style is used to present the area integration weight from the area division result and the area cluster dimension, the second graph style is used to present the area integration weight based on the area cluster corresponding to the highest weight, and the third graph style is used to present the area integration weight from the longitude and latitude dimensions.

[0277] An embodiment of the present application provides a computer program product, which includes computer-executable instructions or a computer program. The computer-executable instructions or the computer program are stored in a computer-readable storage medium. A processor of an electronic device reads the computer-executable instructions or the computer program from the computer-readable storage medium, and the processor executes the computer-executable instructions or the computer program, so that the electronic device executes the service data processing method described above in the embodiments of the present application.

[0278] An embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions or a computer program are stored. When the computer-executable instructions or the computer program are executed by a processor, the processor will be caused to execute the service data processing method provided by the embodiment of the present application. For example, as Figure 3 the service data processing method shown.

[0279] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, flash memory, magnetic surface memory, optical disc, or CD-ROM; or may be various devices including one or any combination of the above memories.

[0280] In some embodiments, the computer-executable instructions may be in the form of a program, software, software module, script, or code, and may be written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including being deployed as an independent program or being deployed as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0281] As an example, the computer-executable instructions may or may not correspond to a file in the file system, and may be stored as part of a file that stores other programs or data. For example, they may be stored in one or more scripts in a Hyper Text Markup Language (HTML) document, stored in a single file dedicated to the program being discussed, or stored in multiple cooperating files (for example, files that store one or more modules, subroutines, or code portions).

[0282] As an example, the computer-executable instructions may be deployed to be executed on one electronic device (in this case, this one electronic device is the service data processing device), or on multiple electronic devices located at one location (in this case, the multiple electronic devices located at one location are the service data processing devices), or on multiple electronic devices distributed at multiple locations and interconnected by a communication network (in this case, the multiple electronic devices distributed at multiple locations and interconnected by a communication network are the service data processing devices).

[0283] It is understandable that in the embodiments of the present application, relevant data such as service location sets, period service information, and periodic service information are involved. When the embodiments of the present application are applied to specific products or technologies, user permission or consent needs to be obtained, and the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards in relevant countries and regions. In addition, the collection and processing of relevant data in this application should strictly comply with the requirements of relevant national laws and regulations in actual applications, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing behaviors within the scope authorized by laws, regulations, and the personal information subject. In addition, in this application, when the data scraping technical solution is implemented and the above embodiments of this application are applied to specific products or technologies, the process of collecting, using, and processing relevant data should comply with the requirements of national laws and regulations, conform to the principles of legality, legitimacy, and necessity, do not involve obtaining data types prohibited or restricted by laws and regulations, and will not interfere with the normal operation of the target website.

[0284] In summary, when the embodiments of the present application respond to a service provision request, they first combine the K historical service features of the area division result, the area location relationship, and the service time to be served to determine the K service area features and area integration weights of the area division result. Then, based on the area integration weights, the relevant service area features are aggregated to learn the overall features of the area division result. Finally, the S target probabilities of S service providers providing services in the target service area are predicted by combining the J area aggregation features and service area features aggregated, so that at least one service provider can be recommended based on the S target probabilities in the target service area. Thus, the service provider with the highest service provision possibility can be accurately determined through the at least one recommended service provider, and furthermore, the recommendation accuracy of the service provider can be improved.

[0285] The above description is only for the embodiments of the present application and is not intended to limit the protection scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A service data processing method, characterized in that, The method includes: In response to a service provision request, with respect to the service time to be served, extract the historical service characteristics of each service area provided by S service providers in the area division result, obtaining K historical service characteristics, where both S and K are integers greater than 1; Combining the service time to be served, the K historical service characteristics, and the regional position relationship of the area division result, determine K service area characteristics and regional integration weights; Based on the regional integration weights, integrate the K service area characteristics to obtain J regional integration characteristics, where J is a positive integer less than K; Based on the J regional integration characteristics and the K service area characteristics, predict the S target probabilities of the S service providers providing services in the target service area; Combining the S target probabilities, present at least one of the recommended service providers and the service probability to be served corresponding to at least one of the service providers.

2. The method according to claim 1, characterized in that, The combining the service time to be served, the K historical service characteristics, and the regional position relationship of the area division result to determine K service area characteristics and regional integration weights includes: Concatenate the time characteristics of the service time to be served and the K historical service characteristics to obtain an initial concatenated characteristic; Perform a linear transformation on the initial concatenated characteristic to obtain K initial area characteristics; Based on the regional position relationship of the area division result, enhance the K initial area characteristics to obtain K service area characteristics; Combine the K initial area characteristics and the regional position relationship to obtain the regional integration weights.

3. The method according to claim 2, characterized in that, Before the concatenating the time characteristics of the service time to be served and the K historical service characteristics to obtain an initial concatenated characteristic, the method further includes: Obtain the first service cycle in which the service time to be served is located to obtain a target first service cycle; Obtain the first time period number of the service time to be served within the target first service cycle; Obtain the second time period number of the target first service cycle within a second service cycle, where the second service cycle includes a plurality of the first service cycles; Integrate the encoding results of the first time period number and the encoding results of the second time period number into the time characteristics.

4. The method according to claim 2, wherein The enhancing the K initial area characteristics based on the regional position relationship of the area division result to obtain K service area characteristics includes: Based on the regional position relationship of the area division result, perform a first attention process on the K initial area characteristics to obtain K enhanced area characteristics; Based on the regional position relationship, perform a second attention process on the K enhanced area characteristics to obtain K area characteristics to be connected; Connect the K enhanced area characteristics and the K area characteristics to be connected to obtain K service area characteristics.

5. The method according to claim 1, characterized in that The integrating the K service area characteristics based on the regional integration weights to obtain J regional integration characteristics includes: Iterate through a specified number of sequences by a, and perform the following processing for each traversed specified number, where a is a positive integer variable: Based on the (a - 1)-th allocation weight, allocate multiple (a - 1)-th features to be allocated to obtain the specified number of a-th features to be allocated. When a is 1, the (a - 1)-th allocation weight is the region integration weight, and the multiple (a - 1)-th features to be allocated are K service region features; When the traversed specified number is different from the last specified number, determine the a-th positional relationship by combining the (a - 1)-th allocation weight and the (a - 1)-th positional relationship, and determine the a-th allocation weight by combining the a-th positional relationship and the specified number of a-th features to be allocated. Continue to iterate a. When a is 1, the (a - 1)-th positional relationship is the region positional relationship; When the traversed specified number is the last specified number, end the iteration of a, and determine the specified number of a-th features to be allocated at the end of the iteration of a as J region integration features.

6. The method according to any one of claims 1 to 5, characterized in that Predicting S target probabilities that S service providers provide services in the target service region based on the J region integration features and the K service region features includes: Based on the region integration weight, restore the J region integration features to the region division result to obtain K features to be enhanced; Combine the K features to be enhanced and the K service region features to obtain K features to be predicted; Predict S service providing probabilities that S service providers provide services in each service region based on the K features to be predicted; Determine the target service region where the location to be served is located from the region division result; Determine S target probabilities corresponding to the target service region based on the S service providing probabilities that S service providers provide services in each service region.

7. The method according to claim 6, wherein After predicting S service providing probabilities that S service providers provide services in each service region based on the K features to be predicted, the method further includes: In response to the next service providing request, obtain the requested service time and the requested service location; When the time difference between the requested service time and the to-be-served time is less than the time difference threshold, determine the service region where the requested service location is located from the region division result to obtain the requested service region; Determine S service providing probabilities corresponding to the requested service region based on the S service providing probabilities that S service providers provide services in each service region; Present service providing information based on the S service providing probabilities corresponding to the requested service region.

8. The method according to any one of claims 1 to 5, characterized in that Extracting K historical service features of S service providers providing services in each service region in the region division result in response to the service providing request, relative to the to-be-served time, includes: In response to the service providing request, obtain a time period service information sequence of S service providers providing services in the service region and the specified sequence length relative to the to-be-served time; Obtain a sequence of periodic service information of S service providers in the service area and a specified period sequence with respect to the to-be-served time; Determine the encoding results of the time-period service information sequence and the periodic service information sequence as the historical service features, and obtain K historical service features.

9. The method according to claim 8, wherein The obtaining, in response to the service provision request, of a sequence of time-period service information of S service providers in the service area and with a specified sequence length with respect to the to-be-served time includes: In response to the service provision request, with the to-be-served time as the end time of the time sequence, determine the start time of the time sequence based on the specified sequence length; From the start time of the time sequence to the end time of the time sequence, slide with a specified sliding window and a specified sliding step to obtain a sequence of time periods to be extracted; For each time period to be extracted in the sequence of time periods to be extracted, combine the first service dispatch volume and the first service provision volume of S service providers in the service area respectively as time-period service information, and obtain the sequence of time-period service information corresponding to the specified sequence length.

10. The method according to claim 8, wherein The obtaining, with respect to the to-be-served time, of a sequence of periodic service information of S service providers in the service area and a specified period sequence includes: In each specified period of the specified period sequence, determine the to-be-processed time corresponding to the to-be-served time; Based on the specified range length, obtain a historical service time period centered on the to-be-processed time; Within the historical service time period, combine the second service dispatch volume and the second service provision volume of S service providers in the service area as periodic service information, and obtain the sequence of periodic service information corresponding to the specified period sequence.

11. The method according to any one of claims 1 to 5, characterized in that, Before the responding to the service provision request, the method further includes: For the area to be divided, obtain a set of service locations within a specified historical duration before the current division moment; Cluster the set of service locations to obtain K service location clusters; Determine the area corresponding to each service location cluster as the service area, and obtain K service areas corresponding to the K service location clusters, where the area division result includes K service areas.

12. The method according to any one of claims 1 to 5, characterized in that, Before determining K service area features and area integration weights by combining the area position relationship of the to-be-served time, K historical service features, and the area division result, the method further includes: Determine the central position of each service area in the area division result; Based on the central position of the area, determine the area distance between any two service areas in the area division result; Determine two service areas with the area distance less than the adjacent distance threshold as adjacent service areas; Based on the adjacent service areas, determine the area position relationship of the area division result.

13. The method according to claim 1, characterized in that The prediction of S target probabilities is achieved through a service prediction model, and the service prediction model is obtained through training by the following steps: With respect to the to-be-served sample time, extract historical service sample features for providing services in each sample area in the area division sample result; Using the model to be trained, predict the time of the sample to be served, the historical service sample features, and the relationship between the sample area locations, to obtain N predicted probabilities of N service-providing sample parties providing services in each of the sample areas, where the model to be trained is a neural network model to be trained for predicting the probability of providing services, and N is an integer greater than 1; Based on the service dispatch sample quantity of each service-providing sample party in each sample area and the predicted probability, calculate the predicted service dispatch quantity; Based on the difference between the predicted service dispatch quantity and the service dispatch sample quantity of each service-providing sample party in each sample area, calculate the loss function value; Based on the loss function value, train the model to be trained to obtain the service prediction model.

14. The method according to any one of claims 1 to 5 and 13, characterized in that, The combining the S target probabilities to present at least one of the recommended service providers and the corresponding to-be-served probabilities of at least one service provider includes: Select at least one of the target probabilities with the largest target probability from the S target probabilities; Determine at least one of the service providers corresponding to at least one of the target probabilities from the S service providers; Calculate the to-be-served probabilities corresponding to at least one of the target probabilities; Present the to-be-served probabilities and present at least one of the recommended service providers in a specified style.

15. The method according to claim 14, characterized in that, After combining the S target probabilities to present at least one of the recommended service providers and the corresponding to-be-served probabilities of at least one service provider, the method further includes: In response to a selection adjustment operation for the to-be-served probability, obtain at least one adjusted service provider; Determine at least one adjusted target probability corresponding to at least one adjusted service provider from the S target probabilities; Calculate the current service probabilities corresponding to at least one of the adjusted target probabilities; Present the current service probabilities and present at least one of the adjusted service providers in the specified style.

16. The method according to any one of claims 1 to 5 and 13, characterized in that, After combining the to-be-served time, K historical service features, and the regional location relationship of the regional division result to determine K service area features and regional integration weights, the method further includes: Present the regional integration weights in a specified graph style, where the specified graph style includes a first graph style, a second graph style, and a third graph style. The first graph style is used to present the regional integration weights from the dimensions of the regional division result and the regional cluster. The second graph style is used to present the regional integration weights based on the regional cluster corresponding to the highest weight. The third graph style is used to present the regional integration weights from the longitude and latitude dimensions.

17. A service data processing device, characterized in that, The service data processing device includes: A feature extraction module, configured to, in response to a service provision request, extract historical service features of S service providers providing services in each service area in the regional division result with respect to the time of the sample to be served, to obtain K historical service features, where both S and K are integers greater than 1; The feature extraction module is further configured to determine K service area features and area integration weights in combination with the to-be-served time, the K historical service features, and the area location relationship of the area division result; The feature integration module is configured to integrate the K service area features based on the area integration weights to obtain J area integration features, where J is a positive integer less than K; The probability prediction module is configured to predict S target probabilities that the S service providers provide services in the target service area based on the J area integration features and the K service area features; The service recommendation module is configured to present at least one of the recommended service providers and the to-be-served probabilities corresponding to at least one of the service providers in combination with the S target probabilities.

18. An electronic device for service data processing, characterized in that, The electronic device includes: A memory for storing computer-executable instructions or computer programs; A processor for implementing the service data processing method according to any one of claims 1 to 16 when executing the computer-executable instructions or computer programs stored in the memory.

19. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer programs are executed by the processor, the service data processing method according to any one of claims 1 to 16 is implemented.

20. A computer program product, comprising computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer programs are executed by the processor, the service data processing method according to any one of claims 1 to 16 is implemented.

Images