A data reporting scheduling method, apparatus, device, medium and product

Abstract

The application provides a data reporting scheduling method, device, equipment, medium and product, and belongs to the technical field of data processing. The method comprises the following steps: extracting a multi-dimensional task feature vector of each task in a task set to be processed, and acquiring current resource state information; inputting the multi-dimensional task feature vector and the resource state information into a strategy network to obtain game decision parameters output by the strategy network; performing balanced solving on the multi-task game model based on the game decision parameters, the multi-dimensional task feature vector and the resource state information to obtain a scheduling strategy distribution of each task; determining a target scheduling scheme of the task set to be processed based on the resource state information, the scheduling strategy distribution of all tasks and the multi-dimensional task feature vector; and performing data reporting of each task based on the target scheduling scheme. The application is an intelligent scheduling method with multi-dimensional value perception, dynamic resource game and autonomous strategy evolution capability, and solves the multi-objective optimization problem in the data reporting scene.

Description

Technical Field

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

[0002] In complex business scenarios such as insurance industry regulation, data reporting typically involves multi-target platform collaboration, processing of massive amounts of heterogeneous data, and stringent timeliness requirements. Therefore, how to efficiently and rationally coordinate and distribute a large number of concurrent reporting tasks while ensuring the stable operation of the underlying system is the overall technical requirement facing this field.

[0003] To meet these requirements, existing mainstream technical solutions generally adopt a priority queue-based scheduling mechanism. Specifically, this solution mainly determines the queuing and execution order of tasks directly through manually preset static rules (such as preset single priority labels, supervision levels, or deadlines). During execution, the system usually combines a monitoring module to collect the occupancy of underlying physical resources or the concurrent number of external interfaces. When resource pressure is detected, the scheduling system will trigger fixed passive adjustment strategies, such as directly suspending low-priority tasks at the back of the queue, or performing basic rate limiting and degradation actions to maintain the basic operation of the system.

[0004] However, the existing technical solutions mentioned above have the following systemic defects: First, the evaluation and measurement system of the existing solutions is too rigid, relying only on single-dimensional static labels for task sorting. It cannot comprehensively measure the diverse value attributes of each task in complex business scenarios. Under dynamically changing system loads, it is easy to generate structural imbalance problems such as high-priority tasks excessively seizing resources while other key tasks cannot be executed due to lag, or resource idleness and task backlog coexisting. Second, the existing scheduling mechanism tends to treat each task in concurrent scenarios as an isolated decision-making unit. Its essence is a greedy one-sided resource seizure mode. When faced with sudden resource constraints and complex resource competition environments, this isolated execution mode cannot coordinate and resolve the mutual squeezing and conflict between concurrent tasks. As a result, the system is prone to falling into local optimum allocation or even resource deadlock during peak reporting periods, which seriously reduces the overall task reporting efficiency and execution success rate. Summary of the Invention

[0005] This application provides a data reporting and scheduling method, apparatus, equipment, medium, and product, aiming to solve the above-mentioned problems.

[0006] Firstly, this application provides a data reporting scheduling method, including: Extract the multidimensional task feature vectors of each task in the set of tasks to be processed, and obtain the current resource status information; The multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network. Based on the game decision parameters, the multi-dimensional task feature vector, and the resource status information, the equilibrium solution is obtained by using a multi-task game model to obtain the scheduling strategy distribution of each task. The target scheduling scheme for the set of tasks to be processed is determined based on the resource status information, the scheduling strategy distribution of all the tasks, and the multidimensional task feature vector. Data reporting for each task is executed based on the target scheduling scheme.

[0007] As an example, the game decision parameters include the weight vector of each dimension of the task features in the multi-dimensional task feature vector and the game temperature parameter; Accordingly, the step of obtaining the scheduling strategy distribution of each task by using a multi-task game model to solve for equilibrium based on the game decision parameters, the multi-dimensional task feature vector, and the resource state information includes: The weight vector is used to weight the task features of each dimension of the multidimensional task feature vector to determine the dynamic value score of each task. The multi-task game model is constructed based on the game temperature parameters. The dynamic value score is used as the basic data of the multi-task game model's revenue. The resource constraints of the multi-task game model are constructed using the resource status information. The solution is performed with the goal of maximizing the global utility of all tasks, and the distribution of scheduling strategies for each task output by the multi-task game model is obtained.

[0008] As one embodiment, determining the target scheduling scheme for the set of tasks to be processed based on the resource status information, the scheduling policy distribution of all the tasks, and the multi-dimensional task feature vector includes: Based on the scheduling strategy distribution, a target strategy action is determined for each of the tasks, and the target strategy action is one of execution, waiting, and degraded execution. Tasks that are selected as either executed or downgraded according to the target strategy action are designated as tasks to be executed, and their execution priority is determined from high to low based on the dynamic value score of each task to be executed. Based on the execution priority order, the system resources are allocated sequentially to each task according to the resource status information and the predicted resource consumption values ​​in the multi-dimensional task feature vectors of each task to be executed, so as to generate the target scheduling scheme for the set of tasks to be processed.

[0009] As one embodiment, after the data reporting of each task is performed based on the target scheduling scheme, the method further includes: Extract real-time feedback data, delay feedback data, and efficiency feedback data during the execution of each of the tasks; The corresponding multi-dimensional feedback reward value is calculated by combining the real-time feedback data, the delayed feedback data, and the efficiency feedback data. The network parameters of the policy network are updated using the multi-dimensional feedback reward value to obtain the updated policy network.

[0010] As one embodiment, the resource status information includes computing resource utilization, network resource utilization, and external interface resource utilization; Before inputting the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network, the method further includes: When at least one of the computing resource utilization rate, the network resource utilization rate, and the external interface resource utilization rate exceeds the corresponding resource bottleneck threshold, the resource game scheduling mode is triggered. Accordingly, in the resource game scheduling mode, the multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network.

[0011] As one embodiment, updating the network parameters of the policy network using the multi-dimensional feedback reward value to obtain an updated policy network includes: Obtain the time context and system load statistics, and input the multidimensional task feature vector, the resource status information, the time context and the system load statistics into the value assessment network to obtain the long-range value estimate of the current state output by the value assessment network; Determine the evaluation error between the multi-dimensional feedback reward value and the long-range value estimate; Based on the evaluation error results, the distribution tendency of the game decision parameters output by the policy network is adjusted to obtain an updated policy network.

[0012] Secondly, this application provides a data reporting and scheduling device, comprising: The system status determination module is used to extract the multi-dimensional task feature vectors of each task in the set of tasks to be processed, and to obtain the current resource status information. The game decision parameter acquisition module is used to input the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network. The scheduling strategy distribution acquisition module is used to obtain the scheduling strategy distribution of each task by using a multi-task game model to solve the equilibrium based on the game decision parameters, the multi-dimensional task feature vector and the resource status information. The target scheduling scheme determination module is used to determine the target scheduling scheme of the set of tasks to be processed based on the resource status information, the scheduling strategy distribution of all the tasks, and the multi-dimensional task feature vector. The execution module is used to perform data reporting for each of the tasks based on the target scheduling scheme.

[0013] Thirdly, this application also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned data reporting scheduling methods.

[0014] Fourthly, this application also provides a non-transitory computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the above-described data reporting scheduling methods.

[0015] Fifthly, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements any of the above-described data reporting scheduling methods. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is one of the flowcharts illustrating the data reporting and scheduling method provided in this application; Figure 2 This is the second flowchart illustrating the data reporting and scheduling method provided in this application; Figure 3 This is a schematic diagram of the data reporting and scheduling device provided in this application; Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0020] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class, without limiting the number of objects; for example, a first object can be one or more. Furthermore, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects have an "or" relationship.

[0021] Insurance industry regulatory data reporting involves multi-objective platform collaboration, processing massive amounts of heterogeneous data, and stringent timeliness requirements. Current mainstream technical solutions mainly revolve around the following three aspects: At the task scheduling level: Existing solutions generally adopt a priority queue-based scheduling mechanism, which determines the task execution order through manually preset rules (such as supervision level and deadline), and starts the data extraction, transformation and reporting process by combining timed triggering or event-driven methods.

[0022] At the resource management level: Some solutions introduce a resource monitoring module to collect real-time data on the usage of computing resources (such as central processing unit (CPU) and memory), network resources (bandwidth and latency), and external interface resources (queries per second (QPS) limit and concurrent connections of the target platform). When resource shortage is detected, passive adjustments are made through preset rules (such as pausing low-priority tasks and rate limiting and degradation).

[0023] At the strategy optimization level: A few solutions attempt to optimize scheduling parameters based on historical statistical data (such as dynamically adjusting priority weights), but the optimization objectives are singular, usually focusing only on single indicators such as task completion rate or average waiting time, lacking comprehensive consideration of multi-dimensional business value and adaptive learning capabilities.

[0024] Current technical solutions have the following systemic flaws when dealing with the complex scenarios of regulatory data reporting in the insurance industry: Defect 1: A single value metric cannot support multi-objective trade-off decisions. Existing solutions simplify task value to a single priority label or deadline, neglecting the diverse value attributes of insurance regulatory reporting. The same data may simultaneously serve different objectives such as solvency supervision, fiscal subsidy settlement, and agricultural risk monitoring, each with fundamentally different requirements for timeliness, completeness, and accuracy. Single-dimensional ranking leads to structural imbalances such as "high-priority tasks excessively consuming resources while low-priority critical tasks are neglected" or "resource idleness coexists with task backlog."

[0025] Defect 2: Static rules are difficult to adapt to dynamic resource competition environments. Manually preset rules are built based on historical experience and cannot respond in real time to sudden changes in resource status (such as sudden rate limiting on the target platform, network link jitter, or a surge in load on core business systems). Conflicts and cumulative effects between rules are difficult to predict manually, leading to resource scheduling falling into a local optimum or deadlock state during peak reporting periods such as the end of the month, quarter, and year, resulting in a sharp decline in overall reporting efficiency.

[0026] Defect 3: Lack of characterization of resource game relationships: In multi-task concurrent scenarios, complex resource competition and coordination relationships exist between tasks (such as input / output (I / O) contention caused by shared database connection pools, and preemption of interface quotas on the same target platform). Existing solutions treat task scheduling as an isolated decision, failing to establish a game-theoretic interaction model between tasks. This makes it impossible to predict the global consequences of strategies such as "parallel execution vs. serial execution" and "resource preemption vs. proactive yielding," resulting in a difficulty in balancing resource utilization and task satisfaction.

[0027] Defect 4: Lack of the ability to continuously evolve from execution feedback: Existing solutions treat scheduling strategies as fixed configurations, and feedback information such as success, failure, and delay after task execution is only used for alarms or statistical reports, failing to form a closed loop of "decision-execution-feedback-optimization". When regulatory policies are adjusted (new reporting platforms are added, data standards are changed) or business scale expands, it is necessary to rely on manual re-analysis, redesign, and verification of rules, which results in long response cycles and is prone to human error.

[0028] Overall, existing technologies are characterized by "rule-driven" approaches, treating reporting tasks as independent and homogeneous execution units. They fail to fully capture the differences in regulatory requirements across multiple platforms, the hierarchical nature of data sensitivity, and the dynamic complexity of resource competition in insurance business scenarios.

[0029] The data reporting and scheduling method, apparatus, equipment, medium, and product provided in this application are described below with reference to the accompanying drawings. They are applicable to data reporting services in the banking, insurance, and other fields.

[0030] It should be noted that the data reporting and scheduling method provided in this application embodiment is implemented based on a data reporting and scheduling device. The data reporting and scheduling method constructs an intelligent scheduling method with multi-dimensional value perception, dynamic resource game and autonomous strategy evolution capabilities, and solves the multi-objective optimization problem in the insurance regulatory data reporting scenario.

[0031] This application describes the data reporting and scheduling method using a data reporting and scheduling device as the execution subject as an example.

[0032] Figure 1 This is one of the flowcharts illustrating the data reporting and scheduling method provided in this application. Figure 2 This is the second flowchart of the data reporting and scheduling method provided in this application.

[0033] like Figure 1 As shown, the data reporting scheduling method provided in this application includes: S110: Extract the multi-dimensional task feature vectors of each task in the set of tasks to be processed, and obtain the current resource status information.

[0034] Combination Figure 2 As shown, the data reporting and scheduling system provided in this application includes an input layer and a perception layer.

[0035] The input layer receives raw reporting tasks, forming a set of tasks to be processed. In the context of insurance industry regulatory data reporting, the set of tasks to be processed may simultaneously include reporting tasks serving different target platforms such as solvency supervision, fiscal subsidy settlement, and agricultural risk monitoring.

[0036] The perception layer parses the metadata of the task set to be processed, quantifies and evaluates it from multiple dimensions of business and technical attributes, and extracts multi-dimensional task feature vectors as a digital profile that comprehensively depicts the diverse value attributes of the tasks. In practice, discrete labels can be converted into numerical values ​​through a preset mapping function, or multi-dimensional task feature vectors can be extracted based on methods such as calculating influence scores using knowledge graphs.

[0037] In one possible implementation, the multidimensional task feature vector includes five dimensions: time urgency, data sensitivity, business impact, predicted resource consumption, and historical reliability. Time urgency primarily reflects the remaining time a task has until a legally mandated deadline or the promised delivery time. This feature directly drives the scheduling system to identify urgent tasks and prioritize them. Data sensitivity primarily reflects the security level and privacy requirements of the data carried by the task. This feature is mainly used by the system during scheduling to assess whether a dedicated encrypted transmission channel, a high-isolation memory sandbox, or additional data auditing and verification processes should be allocated to the task. Business impact primarily reflects the degree of substantial impact on upper-level macro-business operations or external entities caused by the success or delay of the task. Through this feature, the system can accurately discard low-impact tasks to protect core business operations when facing extreme resource depletion. Predicted resource consumption primarily reflects the expected load pressure on the underlying physical and network infrastructure throughout the task's lifecycle. Resource consumption prediction features are used in resource binning algorithms for precise capacity deduction and conflict detection, preventing large tasks from instantly exceeding the system's physical capacity limit. Historical reliability primarily reflects the stability of this type of task or its corresponding target receiving platform in historical execution records. Historical reliability features are used to assist in predicting high-risk tasks and intervening in advance (such as triggering pre-verification, degradation sampling, etc.), transforming post-failure retries into pre-failure risk mitigation, and improving the overall reporting success rate and stability.

[0038] Table 1 provides examples of data sources and quantification methods for the five dimensions of task characteristics.

[0039] Table 1. Data sources and quantification methods for task characteristics

[0040] Resource status information is used to characterize the available resources or occupancy levels of various key resources in the current physical and network environment. This data can be collected in real time through system monitoring probes deployed in operating systems or application containers, network traffic analysis components, and external interface call statistics modules.

[0041] S120: Input the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network.

[0042] Combination Figure 2 As shown, the data reporting and scheduling system provided in this application also includes a decision layer, which is used to determine the target scheduling scheme, namely, to execute S120, S130 and S140.

[0043] The policy network is an artificial intelligence model based on a deep learning architecture. Multi-dimensional task feature vectors and resource state information are encoded and input into the policy network for forward propagation computation. The hidden layers of the policy network perform multi-level implicit feature extraction and representation learning on the multi-dimensional task feature vectors and resource state information through non-linear activation functions. Finally, the output layer outputs the game decision parameters.

[0044] Game decision parameters are not low-level hard instructions that directly guide whether tasks should be executed or suspended, but rather macro-level control variables used to regulate the operating environment of subsequent multi-task game models. When faced with different multi-dimensional task feature vectors and resource state information, such as during peak reporting periods at the end of the month and off-peak periods in the middle of the month, the policy network automatically outputs different values ​​of game decision parameters, thereby dynamically changing the system's emphasis on different feature dimensions and the intensity of competition between tasks, achieving an end-to-end mapping from environmental perception to policy generation.

[0045] In one possible implementation, the game decision parameters include the weight vectors of each dimension of the task features in the multidimensional task feature vector and the game temperature parameter. The game temperature parameter is used to control the balance between "exploration and exploitation" during game solving, or to control the smoothness of Nash equilibrium solution. It determines whether the task tends to be "absolutely rational and selfish" or has a certain "tolerance for random concessions" when faced with resource conflicts.

[0046] In one possible implementation, combining Figure 2 As shown, the policy network is a dual-network architecture based on reinforcement learning, consisting of a value evaluation network and a policy generation network, i.e., an actor-critic architecture. The policy generation network (Actor) is responsible for outputting specific game decision parameters (weight vector, game temperature parameters, etc.) based on the current state. The value evaluation network (Critic) is responsible for evaluating the "goodness" (expected long-term payoff) of the current state. The evaluation results of the value evaluation network guide the policy generation network in updating its parameters, thereby reducing the variance during policy updates, accelerating model convergence, and avoiding getting trapped in local optima.

[0047] The input data for both the value assessment network and the policy generation network is the system state vector. The system state vector is obtained by encoding multi-dimensional task feature vectors, resource state information, temporal context (such as relative position to the end of the month / quarter / year), and system load statistics. The output parameters of the policy generation network are game decision parameters, used to assist in decision-making. The output data of the value assessment network is a long-range value estimate of the current state, used to update the network parameters of the policy network, as explained below.

[0048] In one possible implementation, the input layers of the value evaluation network and the policy generation network are 10-dimensional, and the hidden layers employ multi-layer fully connected neural networks, such as a multi-layer fully connected neural network with 256-128 units compressed layer by layer. The output layer of the policy generation network is 6-dimensional, and the activation policy is weight-normalized activation. The output layer of the value evaluation network is 1-dimensional, and the activation policy is a linear output policy.

[0049] The reinforcement learning-driven strategy evolution mechanism can automatically identify regulatory reporting time patterns (such as end-of-quarter task peaks and concentrated reporting during agricultural seasons) and adjust resource reservation and scheduling strategies in advance. The strategy network can quickly adapt to new regulatory platforms, policy and standard changes, or business expansion through online learning, significantly reducing the cost of manual rule maintenance.

[0050] S130: Based on the game decision parameters, the multi-dimensional task feature vector, and the resource status information, the multi-task game model is used to solve for equilibrium and obtain the scheduling strategy distribution of each task.

[0051] A multi-task game theory model is a mathematical framework used to simulate resource competition and cooperation among concurrent tasks. First, the game environment and boundary conditions of the multi-task game theory model are defined based on game decision parameters, the multi-dimensional task feature vectors, and the resource state information. Then, an iterative optimization algorithm is used to solve the equilibrium of the multi-task game theory model. The solution process essentially evaluates the mutual influence of each task taking different actions within a limited resource pool, seeking a set of strategies that relatively balances the interests of all parties within the system. After multiple rounds of calculation, the multi-task game theory model finally outputs the scheduling strategy distribution for each task.

[0052] The scheduling strategy distribution is the set of probabilities that each task will take different candidate actions, representing the probability that each task will take different strategies under the current multi-party game equilibrium state.

[0053] This step replaces greedy priority preemption with game equilibrium solution, achieving Pareto improvement in task satisfaction and resource utilization in resource bottleneck scenarios. The strategy probability distribution output by the multi-task game model provides a complete quantitative basis for scheduling decisions, meeting the requirements for decision transparency (such as the requirements for decision transparency in financial industry regulatory audits), and supporting accurate responses to complex inquiries such as "why is a task delayed?"

[0054] S140: Determine the target scheduling scheme for the set of tasks to be processed based on the resource status information, the scheduling policy distribution of all the tasks, and the multi-dimensional task feature vector.

[0055] The scheduling strategy distribution is the macro-probability distribution of each task. Based on this, the resource consumption of each task and its priority order in the set of tasks to be processed are determined by combining each candidate action and multi-dimensional task feature vector. Based on the priority order, the resource consumption of each task is deducted from the overall resources reflected by the resource status information, and finally the target scheduling scheme is obtained.

[0056] S150: Execute data reporting for each of the tasks based on the target scheduling scheme.

[0057] Combination Figure 2 As shown, the data reporting and scheduling system provided in this application also includes an execution and evolution layer, which is used to execute the target scheduling scheme and evolve the strategy network based on the task execution feedback data.

[0058] After generating the target scheduling plan, strictly follow the order and resource allocation quota arranged in the target scheduling plan to initiate the reporting process, accurately push the relevant business data to the target supervision platform, and thus complete the overall data reporting work.

[0059] This application's embodiments characterize task attributes through multi-dimensional task feature vectors, breaking the limitations of single priority labels and enabling accurate identification of the diverse values ​​of tasks. Furthermore, it deeply integrates the dynamic game decision parameters output by the policy network with a multi-task game model, reconstructing data reporting scheduling from "static decision-making based on manual rules" to "an intelligent decision-making method with cognitive capabilities." In resource-constrained or concurrent scheduling scenarios, it establishes strategic interaction relationships between tasks through a multi-task game model, replacing traditional greedy rule-based preemption with game equilibrium solutions. This achieves a decision-making leap from local optima to global equilibrium, effectively avoiding system deadlock caused by excessive preemption of high-value tasks and lag in low-value tasks. It significantly improves resource utilization and the global success rate of data reporting in complex dynamic environments. In addition, this application's embodiments are the first to systematically integrate game theory and deep reinforcement learning into the field of financial data reporting, solving the resource optimization problem under multi-objective regulatory requirements and possessing significant methodological innovation and industry application value.

[0060] In one possible implementation, the resource status information includes computing resource utilization, network resource utilization, and external interface resource utilization. Before inputting the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network (S120), the method further includes: When at least one of the computing resource utilization rate, the network resource utilization rate, and the external interface resource utilization rate exceeds the corresponding resource bottleneck threshold, the resource game scheduling mode is triggered. Accordingly, in the resource game scheduling mode, S120 is executed to input the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network.

[0061] In this embodiment, the resource status information includes the occupancy status of three core resources: computing resources, network resources, and external constraints, thereby establishing a multi-level resource monitoring mechanism. Table 2 shows the monitoring indicators and bottleneck determination logic for the three core resources.

[0062] Table 2 Monitoring Indicators and Bottleneck Judgment Logic for Three Types of Core Resources

[0063] Combination Figure 2 As shown, when any of the three core resource types—computing resources, network resources, and external constraints—reaches a bottleneck, the system switches from the conventional parallel scheduling mode to a resource game-theoretic scheduling mode. Only in the resource game-theoretic scheduling mode does a complex intelligent decision-making chain emerge. This involves inputting multi-dimensional task feature vectors and the resource state information into the policy network for forward propagation, thereby taking over the underlying task distribution through a game-theoretic mechanism.

[0064] This application embodiment monitors the resource usage of computing, network, and external interfaces in real time and sets trigger thresholds to dynamically switch scheduling modes. This ensures both the system's rapid response and low computing power consumption under low load conditions, and enables the system to respond in real time to sudden resource shortages such as rate limiting on the target platform or network jitter. By introducing a game-theoretic scheduling mechanism in a timely manner, system congestion is effectively prevented, and the adaptive stress resistance of the scheduling architecture is greatly enhanced.

[0065] In one possible implementation, the game decision parameters include the weight vector of each dimension of the task features in the multidimensional task feature vector and the game temperature parameter. Accordingly, in S130, the step of using a multi-task game model to solve for equilibrium based on the game decision parameters, the multi-dimensional task feature vector, and the resource state information to obtain the scheduling strategy distribution of each task includes: S1310: The weight vector is used to weight the task features of each dimension of the multidimensional task feature vector to determine the dynamic value score of each task.

[0066] This application's embodiments construct a comprehensive value assessment function that maps multi-dimensional features into comparable value scores. The comprehensive value assessment function employs a weighted aggregation approach, normalizing the task features of each dimension and then fusing them together using weight vectors to obtain a dynamic value score. Taking a five-dimensional task feature vector as an example: ; in, For the first Standardized transformation functions for task characteristics of each dimension (Example: Logarithmic transformation is used to compress the numerical range of business impact and avoid over-dominance). For the first The weight vectors in each dimension are dynamically output by the reinforcement learning agent (policy network) based on the current system state, rather than being manually fixed and preset.

[0067] Dynamic value scores determine the magnitude of the "payoff function" in multi-task game models (the higher the dynamic value score of a task, the greater the payoff from executing that task, and the greater the penalty from downgrading / waiting). In the financial industry, dynamic value scores provide a complete quantitative basis for scheduling decisions, meeting the transparency requirements of financial industry regulatory audits and supporting accurate responses to complex inquiries such as "why is a task delayed?"

[0068] S1320: Construct the multi-task game model based on the game temperature parameters, use the dynamic value score as the basic data of the multi-task game model's revenue, construct the resource constraints of the multi-task game model using the resource state information, solve the problem with the goal of maximizing the global utility of all tasks, and obtain the distribution of scheduling strategies for each task output by the multi-task game model.

[0069] The game elements of a multi-task game model include participants, policy space, payoff function, and resource constraints. Participants are the task instances in the current set of tasks to be processed. The policy space includes three policies: execution policy, waiting policy, and degraded execution policy. The execution policy represents submitting the report with all data and the highest precision. The waiting policy represents temporarily not occupying any processing resources and residing in the buffer queue. The degraded execution policy, while ensuring basic requirements are met, simplifies processing by using partially sampled data instead of full data or removing some non-core business fields, significantly reducing the task's memory and network bandwidth consumption. This allows the task to "voluntarily concede" in resource contention, sacrificing some resources to achieve the bottom-line goal of timely task completion.

[0070] During the determination of the target scheduling scheme, each task can choose one of these three strategies. In the reward function, if the task is successfully executed, it receives the dynamic value score obtained above; if the task fails or is delayed, it is penalized, and the severity of the penalty is positively correlated with the business impact. The total resource consumption of tasks that choose to execute or degrade the execution strategy must not exceed the current available resource capacity (determined by resource status information).

[0071] Specifically, a multi-task game model is constructed using the game decision parameters output by the policy network. Based on the dynamic value scores of each task, the expected payoffs and penalties for each participant (i.e., each task) are set for the multi-task game model, i.e., the payoff baseline data. At the same time, the current resource state information is transformed into the overall resource constraint upper limit of the multi-task game model in the solution process.

[0072] Subsequently, with the objective of maximizing the global utility of all tasks, an iterative optimization algorithm (such as virtual regret minimization and its variants) is used to solve for the approximate Nash equilibrium, outputting the scheduling policy distribution for each task. Table 3 provides examples of the scheduling policy distributions for the three tasks: Table 3. Distribution of scheduling strategies for the three tasks

[0073] In the scheduling strategy distribution in Table 3, reporting tasks with high business impact (such as fiscal subsidy settlement) receive more reliable timeliness guarantees, while low-value tasks are prevented from excessively consuming scarce resources. Reporting tasks with higher value but higher resource consumption (such as solvency monitoring) are proactively deferred when bottlenecks occur to avoid affecting transmission efficiency. Reporting tasks with relatively lower value (such as agricultural statistics) are proactively downgraded to release resources.

[0074] This application embodiment uses the dynamic weight vector output by the policy network for weighted calculation, which realizes a refined characterization of the multidimensional value of the reporting task. This avoids high-priority non-urgent tasks from excessively occupying resources while low-priority but critical tasks with a wide impact are starved. At the same time, a game temperature parameter is introduced to construct a multi-task game model, which transforms unilateral resource grabbing into a non-cooperative game between multiple tasks. By using equilibrium solution, tasks can achieve organized global optimal compromise when resources are limited, which significantly improves the overall utility output ratio of data reporting in complex business scenarios.

[0075] In one possible implementation, S140, determining the target scheduling scheme for the set of tasks to be processed based on the resource status information, the scheduling policy distribution of all the tasks, and the multi-dimensional task feature vector, includes: S1410: Based on the scheduling policy distribution, determine the target policy action for each of the tasks, wherein the target policy action is one of execution, waiting, and degraded execution.

[0076] First, analyze the macroscopic distribution of scheduling strategies. Based on the principle of maximizing probability or the roulette wheel sampling algorithm, uniquely determine a target strategy action for each task in the set of tasks to be processed.

[0077] S1420: Select tasks that are to be executed or downgraded according to the target strategy action as tasks to be executed, and determine the execution priority order according to the dynamic value score of each task to be executed from high to low.

[0078] After determining the target strategy action for each task, tasks with the target strategy action of "waiting" are separated, and tasks with the target strategy action of "execution" or "downgraded execution" are selected and merged into a group of tasks to be executed.

[0079] To ensure the smooth operation of high-value business, all tasks to be executed are arranged in strict descending order according to the dynamic value score calculated in the early stage, forming a definite execution priority order.

[0080] S1430: According to the execution priority order, based on the resource status information and the resource consumption prediction value in the multi-dimensional task feature vector of each task to be executed, system resources are allocated sequentially to each task to be executed in order to generate the target scheduling scheme of the set of tasks to be processed.

[0081] Following the ordered execution priority, each task to be executed is examined for its predicted resource consumption value, which is embedded in its multi-dimensional task feature vector. This value includes estimated CPU clock cycles and the number of interface calls. An equivalent amount is then deducted from the available resource pool indicated by the current resource status information. If the remaining resources in the pool can cover the predicted resource consumption value of the current task, the task is officially added to the execution list and its corresponding resource quota is locked. If the remaining resources are insufficient, even if the task is to be executed, it must be forcibly rolled back to a waiting state or further degraded scheduling must be triggered. The complete task sorting queue and resource locking mapping together constitute the final target scheduling scheme issued to the underlying processing engine.

[0082] This application's embodiments break the deadlock of traditional scheduling's "either execute or wait" by introducing a degraded execution action in the policy space, giving the system the ability to proactively yield tasks when facing severe resource constraints. Combined with an absolute sorting algorithm based on dynamic value scores and a precise resource binning algorithm based on predicted resource consumption values, it ensures absolute resource protection for core high-value reporting tasks while maximizing the utilization of the remaining space of underlying resources, effectively balancing resource utilization and the satisfaction rate of concurrent tasks, and mitigating the risk of resource deadlock.

[0083] In one possible implementation, such as Figure 2 As shown, after the data reporting of each task is executed based on the target scheduling scheme, the process further includes: S160: Update the network parameters of the policy network using feedback data during task execution.

[0084] like Figure 2 As shown, the network parameter updates are performed at the execution and evolution layer of the data reporting scheduling system.

[0085] Specifically, the network parameters of the policy network are updated using feedback data during task execution, including: S1610: Extract real-time feedback data, delay feedback data, and efficiency feedback data during the execution of each of the tasks.

[0086] After completing a round of data reporting according to the target scheduling plan, a distributed log collection and business tracking component is activated to retrieve multi-granularity, cross-temporal feedback information on the scheduling results. Specifically, it extracts immediate feedback data, delayed feedback data, and efficiency feedback data. Immediate feedback data includes task startup time, actual resource consumption at the physical machine level, and task-level success or failure status markers. This data is used to calculate single-step rewards, reflecting the effectiveness of immediate decisions. Delayed feedback data includes whether the reporting task was completed before the final regulatory deadline, whether the downstream fund disbursement system experienced a cascading disruption due to the reporting delay, and whether an inquiry response was received from the target regulatory platform. This data is used to assess long-term impacts and correct short-sighted decisions. Efficiency feedback data includes the actual resource utilization rate of the global egress, the change in the task waiting queue backlog length, and the deviation analysis values ​​between expected and actual resource consumption. This data is used to optimize the resource prediction model for extracting resource status information.

[0087] S1620: Calculate the corresponding multi-dimensional feedback reward value by combining the real-time feedback data, the delayed feedback data, and the efficiency feedback data.

[0088] Specifically, different scalarization transformation rules and fusion weights are assigned to real-time feedback data, delayed feedback data, and efficiency feedback data. Through mathematical transformations such as weighted accumulation or penalty deduction, these complex post-execution performances are aggregated into a pure numerical scalar that can be directly read by the neural network, namely, the multi-dimensional feedback reward value. The multi-dimensional feedback reward value intuitively quantifies the quality of the previous round of scheduling decisions.

[0089] S1630: Update the network parameters of the policy network using the multi-dimensional feedback reward value to obtain the updated policy network.

[0090] Specifically, by utilizing backpropagation algorithms such as policy gradient update mechanisms in reinforcement learning (e.g., proximal policy optimization and its variants), and using the state-action-reward sequence stored in the experience replay buffer, with the optimization objective of maximizing expected reward, the multi-dimensional feedback reward value is used as a correction signal to act on the connection weight matrix inside the policy network, thereby fine-tuning and correcting the network parameters of the policy network, and thus obtaining the updated policy network.

[0091] It should be noted that the update frequency of network parameters is configurable (e.g., an update is performed after each fixed batch of tasks is completed), enabling the system to quickly learn from recent execution experience and adapt to environmental changes.

[0092] This application establishes a multi-granular panoramic feedback mechanism covering underlying resource consumption, upper-layer business timeliness, and macro-system efficiency. It accurately transforms the execution results of the business layer into machine learning reward signals, driving the policy network to perform online closed-loop updates. This endows the scheduling system with the ability to evolve autonomously and learn continuously, significantly reducing the cost of manual rule reconstruction required when dealing with new regulatory platforms or policy standard changes.

[0093] In one possible implementation, in S1630, the network parameters of the policy network are updated using the multi-dimensional feedback reward value to obtain an updated policy network, including: S1631: Obtain the time context and system load statistics, and input the multidimensional task feature vector, the resource status information, the time context and the system load statistics into the value assessment network to obtain the long-range value estimate of the current state output by the value assessment network.

[0094] When updating network parameters, the system not only retrieves multi-dimensional task feature vectors and resource status information stored in the current or historical experience pool, but also collaboratively obtains their corresponding time context and system load statistics. Time context refers to the relative position of the absolute timestamp of a task's creation or execution within the business cycle, such as the number of days remaining until the end of the month or whether it's the end of a fiscal year's settlement quarter. System load statistics refer to macro-level trend characteristics such as the slope of system throughput changes or average request frequency within a specific time sliding window in the past.

[0095] The multidimensional task feature vector, resource status information, temporal context, and system load statistics are concatenated and fed into the value assessment network as an input vector. The value assessment network, through a nonlinear mapping of a multi-layer fully connected perceptron, ultimately outputs a one-dimensional scalar, namely the long-range value estimate. The long-range value estimate represents the sum of future long-term discounted returns that the system can expect to obtain by continuing to execute the current strategy under specific load conditions at a specific time point.

[0096] S1632: Determine the evaluation error result between the multi-dimensional feedback reward value and the long-range value estimate.

[0097] The evaluation error is calculated by subtracting the multi-dimensional feedback reward value actually calculated from the execution environment from the long-range value estimate predicted by the value assessment network. This evaluation error indicates the specific extent to which the actual execution performance exceeds or falls short of the model's historical expectations.

[0098] S1633: Adjust the distribution tendency of the game decision parameters output by the strategy network based on the evaluation error results to obtain an updated strategy network.

[0099] The evaluation error is used as a gradient guidance signal and fed back to the policy network. If the error is positive, it means that the previous policy exceeded expectations. In this case, the internal parameters of the policy network are adjusted to increase the probability distribution tendency of the network to output the set of game decision parameters under similar environmental conditions. Conversely, if the error is negative, the distribution tendency is suppressed. After repeated iterations, a more adaptive and updated policy network is finally output.

[0100] This application's embodiments introduce time context and system load trends into independent network branches with evaluation capabilities, and refine the policy network update based on the temporal difference error between predicted value and actual reward. This greatly reduces gradient variance in the policy optimization process and accelerates network convergence. This mechanism enables the system to automatically learn and extract the peak patterns of specific time nodes such as the end of the month and the end of the quarter from massive historical execution data. It can predict conflicts and proactively adjust the scheduling game parameters in advance, overcoming the limitations of manual prediction capabilities. In regulatory reporting scenarios with periodic fluctuations, it achieves proactive resource coordination and global efficiency optimization.

[0101] Based on the above, this application also provides a data reporting and scheduling apparatus. The data reporting and scheduling apparatus and the data reporting and scheduling method described above can be referred to in correspondence.

[0102] As an example, such as Figure 3 As shown, the data reporting and scheduling device includes: The system status determination module 310 is used to extract the multi-dimensional task feature vectors of each task in the set of tasks to be processed, and to obtain the current resource status information. The game decision parameter acquisition module 320 is used to input the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network. The scheduling strategy distribution acquisition module 330 is used to obtain the scheduling strategy distribution of each task by using a multi-task game model to solve the equilibrium based on the game decision parameters, the multi-dimensional task feature vector and the resource status information. The target scheduling scheme determination module 340 is used to determine the target scheduling scheme of the set of tasks to be processed based on the resource status information, the scheduling strategy distribution of all the tasks, and the multi-dimensional task feature vector. The execution module 350 is used to perform data reporting for each of the tasks based on the target scheduling scheme.

[0103] This application's embodiments characterize task attributes through multi-dimensional task feature vectors, breaking the limitations of single priority labels and enabling accurate identification of the diverse values ​​of tasks. Furthermore, it deeply integrates the dynamic game decision parameters output by the policy network with a multi-task game model, reconstructing data reporting scheduling from "static decision-making based on manual rules" to "an intelligent decision-making method with cognitive capabilities." In resource-constrained or concurrent scheduling scenarios, it establishes strategic interaction relationships between tasks through a multi-task game model, replacing traditional greedy rule-based preemption with game equilibrium solutions. This achieves a decision-making leap from local optima to global equilibrium, effectively avoiding system deadlock caused by excessive preemption of high-value tasks and lag in low-value tasks. It significantly improves resource utilization and the global success rate of data reporting in complex dynamic environments. In addition, this application's embodiments are the first to systematically integrate game theory and deep reinforcement learning into the field of financial data reporting, solving the resource optimization problem under multi-objective regulatory requirements and possessing significant methodological innovation and industry application value.

[0104] Figure 4 This is a schematic diagram of the structure of the electronic device provided in this application, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions in the memory 430 to execute a data delivery scheduling method, which includes: Extract the multidimensional task feature vectors of each task in the set of tasks to be processed, and obtain the current resource status information; The multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network. Based on the game decision parameters, the multi-dimensional task feature vector, and the resource status information, the equilibrium solution is obtained by using a multi-task game model to obtain the scheduling strategy distribution of each task. The target scheduling scheme for the set of tasks to be processed is determined based on the resource status information, the scheduling strategy distribution of all the tasks, and the multidimensional task feature vector. Data reporting for each task is executed based on the target scheduling scheme.

[0105] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0106] On the other hand, this application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer is able to execute the data reporting and scheduling method provided in the above embodiments. The method includes: Extract the multidimensional task feature vectors of each task in the set of tasks to be processed, and obtain the current resource status information; The multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network. Based on the game decision parameters, the multi-dimensional task feature vector, and the resource status information, the equilibrium solution is obtained by using a multi-task game model to obtain the scheduling strategy distribution of each task. The target scheduling scheme for the set of tasks to be processed is determined based on the resource status information, the scheduling strategy distribution of all the tasks, and the multidimensional task feature vector. Data reporting for each task is executed based on the target scheduling scheme.

[0107] In another aspect, this application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, is implemented to perform the data reporting scheduling method provided in the above embodiments, the method comprising: Extract the multidimensional task feature vectors of each task in the set of tasks to be processed, and obtain the current resource status information; The multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network. Based on the game decision parameters, the multi-dimensional task feature vector, and the resource status information, the equilibrium solution is obtained by using a multi-task game model to obtain the scheduling strategy distribution of each task. The target scheduling scheme for the set of tasks to be processed is determined based on the resource status information, the scheduling strategy distribution of all the tasks, and the multidimensional task feature vector. Data reporting for each task is executed based on the target scheduling scheme.

[0108] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0109] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A data reporting and scheduling method, characterized in that, include: Extract the multidimensional task feature vectors of each task in the set of tasks to be processed, and obtain the current resource status information; The multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network. Based on the game decision parameters, the multi-dimensional task feature vector, and the resource status information, the equilibrium solution is obtained by using a multi-task game model to obtain the scheduling strategy distribution of each task. The target scheduling scheme for the set of tasks to be processed is determined based on the resource status information, the scheduling strategy distribution of all the tasks, and the multidimensional task feature vector. Data reporting for each task is executed based on the target scheduling scheme.

2. The data reporting and scheduling method according to claim 1, characterized in that, The game decision parameters include the weight vector of each dimension of the task features in the multidimensional task feature vector and the game temperature parameter. Accordingly, the step of obtaining the scheduling strategy distribution of each task by using a multi-task game model to solve for equilibrium based on the game decision parameters, the multi-dimensional task feature vector, and the resource state information includes: The weight vector is used to weight the task features of each dimension of the multidimensional task feature vector to determine the dynamic value score of each task. The multi-task game model is constructed based on the game temperature parameters. The dynamic value score is used as the basic data of the multi-task game model's revenue. The resource constraints of the multi-task game model are constructed using the resource status information. The solution is performed with the goal of maximizing the global utility of all tasks, and the distribution of scheduling strategies for each task output by the multi-task game model is obtained.

3. The data reporting and scheduling method according to claim 2, characterized in that, The step of determining the target scheduling scheme for the set of tasks to be processed based on the resource status information, the scheduling policy distribution of all the tasks, and the multi-dimensional task feature vector includes: Based on the scheduling strategy distribution, a target strategy action is determined for each of the tasks, and the target strategy action is one of execution, waiting, and degraded execution. Tasks that are selected as either executed or downgraded according to the target strategy action are designated as tasks to be executed, and their execution priority is determined from high to low based on the dynamic value score of each task to be executed. Based on the execution priority order, the system resources are allocated sequentially to each task according to the resource status information and the predicted resource consumption values ​​in the multi-dimensional task feature vectors of each task to be executed, so as to generate the target scheduling scheme for the set of tasks to be processed.

4. The data reporting and scheduling method according to claim 1, characterized in that, After the data reporting for each task is executed based on the target scheduling scheme, the process further includes: Extract real-time feedback data, delay feedback data, and efficiency feedback data during the execution of each of the tasks; The corresponding multi-dimensional feedback reward value is calculated by combining the real-time feedback data, the delayed feedback data, and the efficiency feedback data. The network parameters of the policy network are updated using the multi-dimensional feedback reward value to obtain the updated policy network.

5. The data reporting and scheduling method according to claim 1, characterized in that, The resource status information includes computing resource utilization rate, network resource utilization rate, and external interface resource utilization rate. Before inputting the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network, the method further includes: When at least one of the computing resource utilization rate, the network resource utilization rate, and the external interface resource utilization rate exceeds the corresponding resource bottleneck threshold, the resource game scheduling mode is triggered. Accordingly, in the resource game scheduling mode, the multidimensional task feature vector and the resource state information are input into the policy network to obtain the game decision parameters output by the policy network.

6. The method according to claim 4, characterized in that, The network parameters of the policy network are updated using the multi-dimensional feedback reward value to obtain the updated policy network, including: Obtain the time context and system load statistics, and input the multidimensional task feature vector, the resource status information, the time context and the system load statistics into the value assessment network to obtain the long-range value estimate of the current state output by the value assessment network; Determine the evaluation error between the multi-dimensional feedback reward value and the long-range value estimate; Based on the evaluation error results, the distribution tendency of the game decision parameters output by the policy network is adjusted to obtain an updated policy network.

7. A data reporting and scheduling device, characterized in that, include: The system status determination module is used to extract the multi-dimensional task feature vectors of each task in the set of tasks to be processed, and to obtain the current resource status information. The game decision parameter acquisition module is used to input the multidimensional task feature vector and the resource state information into the policy network to obtain the game decision parameters output by the policy network. The scheduling strategy distribution acquisition module is used to obtain the scheduling strategy distribution of each task by using a multi-task game model to solve the equilibrium based on the game decision parameters, the multi-dimensional task feature vector and the resource status information. The target scheduling scheme determination module is used to determine the target scheduling scheme of the set of tasks to be processed based on the resource status information, the scheduling strategy distribution of all the tasks, and the multi-dimensional task feature vector. The execution module is used to perform data reporting for each of the tasks based on the target scheduling scheme.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the data reporting scheduling method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium, wherein a computer program is stored on the non-transitory computer-readable storage medium, characterized in that, When the computer program is executed by the processor, it implements the data reporting scheduling method as described in any one of claims 1 to 6.

10. A computer program product, the computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the data reporting scheduling method as described in any one of claims 1 to 6.

Images