Multi-dimensional data acquisition dynamic scheduling system and method

Abstract

The invention discloses a data acquisition optimization method, system and equipment, and belongs to the technical field of data processing. The method comprises the following steps: predicting expected utility of unacquired data based on analysis of acquired data; and dynamically distributing and collecting resources according to the utility. The analysis is based only on intrinsic characteristics of the data. The system correspondingly comprises a utility prediction device and a resource allocation device. According to the method, passive collection is changed into active screening through a core normal form of value pre-judgment driving dynamic scheduling, and resources can be preferentially allocated to high-value data in a prospective manner, so that resource waste is remarkably reduced, and the overall collection efficiency and the data quality are improved. The scheme does not depend on a specific implementation mode, is generally suitable for wide data acquisition scenes such as artificial intelligence and the Internet of Things, and is widely applied to scenes in which high-value data needs to be efficiently acquired from massive and heterogeneous data sources.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent data acquisition optimization method and system based on value prediction. This invention predicts the potential utility of data by analyzing its inherent characteristics and dynamically optimizes the allocation of acquisition resources. It can be widely applied to scenarios requiring efficient acquisition of high-value data from massive, heterogeneous data sources. Background Technology

[0002] With the rapid development of technologies such as artificial intelligence and the Internet of Things, data has become a core production factor. However, traditional data acquisition systems mostly adopt a passive "collect first, evaluate later" model, that is, first consuming bandwidth, storage, and other resources to collect data, and then analyzing and filtering it for value. This model suffers from common defects such as blind resource consumption, low collection efficiency, and low data value density when facing large-scale, multi-source, and dynamically changing data environments. Existing technologies have begun to focus on optimizing data acquisition.

[0003] Existing technologies have begun to explore smarter data acquisition paths. For example, cutting-edge research indicates that integrating data content with contextual information is a key direction for optimizing data acquisition accuracy and latency in scenarios such as the Artificial Intelligence of Things (AIoT). However, such methods typically rely on specific layered architectures (such as edge-cloud collaboration) and awareness of explicit user context (such as settings preferences). Their optimization logic is deeply tied to specific application scenarios, and a core paradigm that does not depend on specific architectures and external contexts and makes general value predictions based on the intrinsic characteristics of the data has not yet been abstracted.

[0004] Furthermore, the publicly disclosed prior art patent CN202411787618.7 (“A Method and System for Broad Data Acquisition”) provides a rule-based multi-source data acquisition framework. While this patent's contribution lies in providing a standardized acquisition process, its core limitation is the lag in quality assessment—its resource allocation decisions primarily rely on pre-set static rules or simple scoring after acquisition, failing to achieve forward-looking dynamic resource scheduling based on the intrinsic value of the data. This results in the system still exhibiting blind resource allocation when faced with uncollected data of uncertain value, making it difficult to fundamentally solve the problem of resource waste.

[0005] Therefore, there is an urgent need in this field for a general technical solution that can intelligently predict the potential value of unknown data before data collection occurs, and thereby achieve dynamic and precise allocation of collection resources, in order to break through the bottleneck of existing technologies. Purpose of the invention

[0006] To address the aforementioned shortcomings of existing technologies, the present invention aims to provide a data acquisition optimization scheme, the core of which is to establish a general paradigm of "value prediction-driven dynamic scheduling." Specific objectives include: 1. Establish a value prediction mechanism based on the analysis of the inherent characteristics of data, and quantify the potential utility of data before it is actually collected, thus changing from passive reception to active screening; 2. Construct a dynamic resource allocation logic that is linked in real time with the value prediction results to ensure that system resources (such as bandwidth, computing power, and storage space) are prioritized for data with high expected value, thereby achieving optimal global collection efficiency; 3. Design a functional system architecture that is independent of specific application scenarios, data formats, or specific algorithms, ensuring that the technical solution is universally applicable to various heterogeneous data acquisition tasks, from Internet information and IoT sensor signals to scientific observation data. Summary of the Invention

[0007] To achieve the above objectives, this invention proposes a general data acquisition optimization method and system.

[0008] This solution does not rely on any specific algorithm implementation, hardware platform, or data format; its core logic lies in the closed loop of "analysis-prediction-allocation".

[0009] I. Core Method: A data acquisition optimization method, characterized by the following steps: Value prediction steps: Based on the **intrinsic characteristics** of the acquired data set, analyze and predict the expected utility of the unacquired data. The "intrinsic characteristics" refer to the semantic and logical attributes inherent in the data content itself, or the structural patterns possessed by the data identifiers, excluding external attributes such as data source and acquisition time. Resource allocation steps: Based on the expected utility predicted above, dynamically allocate the system's data collection resources, prioritizing the collection channels for data predicted to be highly useful.

[0010] II. Core System: A data acquisition and optimization system, characterized by including functionally coordinated devices: Utility prediction device: used to perform the value prediction step, that is, to predict the expected utility of the unacquired data based on the inherent characteristics of the data; Resource allocation device: used to perform the resource allocation step, that is, to dynamically allocate the collected resources according to the predicted utility; Anomaly handling device (optional): used to respond to abnormal states during the data acquisition process and execute reliability assurance strategies such as task retry or resource rescheduling.

[0011] III. Fundamental Difference from Existing Technologies: Compared with existing technologies that rely on fixed rules or ex post facto evaluations (such as CN202411787618.7), the inventive contribution of this invention lies in achieving a **fundamental shift in technological paradigm**: 1. From “static rules” to “dynamic prediction”: The basis for resource allocation has shifted from preset, static rules to dynamic predictions of utility calculated in real time based on the inherent characteristics of the data.

[0012] 2. From “post-event evaluation” to “pre-event prediction”: The value assessment point has been moved from after the data collection is completed to before resource investment, thereby avoiding the consumption of resources for low-value data from the source.

[0013] 3. From “Scenario-Specific” to “General Paradigm”: By stripping away specific implementation details and focusing on the functional essence of “intrinsic feature analysis” and “utility-driven allocation”, this invention provides a general framework that can be adapted to a wide range of data collection scenarios, rather than a specific tool for solving a single problem. Beneficial effects

[0014] Through the above-described open technical solution, this invention can bring the following reasonably foreseeable and fundamental improvements to various data acquisition systems: 1. Fundamental improvement in resource utilization: Through forward-looking value prediction, the system can proactively focus resources on high-value data, thereby significantly reducing resource waste on low-value or redundant data, and theoretically greatly reducing ineffective resource expenditure.

[0015] 2. Synergistic optimization of data acquisition quality and efficiency: Under the same resource constraints, the system can collect data sets with higher overall value; to achieve specific value goals, the system requires lower resource and time costs.

[0016] 3. Enhanced System Adaptability and Robustness: The dynamic allocation mechanism enables the system to flexibly respond to changes in the distribution of data source value and fluctuations in the system's own load. Combined with an optional feedback optimization mechanism, the system possesses the ability to continuously learn and self-optimize.

[0017] 4. Technical universality and forward-looking nature: Based on the innovative architecture design of this invention, this solution is naturally compatible with any new data source (such as photonic chip data streams, neuromorphic computing data, etc.) and any new analysis algorithm that may emerge in the future. The scope of protection covers the technological evolution path and has long-term technological vitality. Attached Figure Description

[0018] Figure 1 This is a flowchart of the closed-loop intelligent data acquisition method provided in this embodiment of the invention.

[0019] Figure 2 This is a schematic diagram of the multi-feature fusion intelligent scheduling model in an embodiment of the present invention.

[0020] Figure 3 This is a schematic diagram illustrating the principle of value prediction based on data pattern aggregation in an embodiment of the present invention.

[0021] Figure 4 This is a flowchart of the distributed collaboration and continuous learning optimization mechanism in an embodiment of the present invention.

[0022] Figure 5 This is a schematic diagram of the hardware structure of an electronic device for implementing the method according to an embodiment of the present invention. Detailed Implementation

[0023] Example 1 Closed-loop intelligent data acquisition method and system This embodiment describes the core architecture. Its purpose is to define and protect the general technical paradigm of "value prediction and dynamic resource allocation based on intrinsic characteristics," without involving any specific algorithm implementation, hardware form, data format, or application scenario.

[0024] I. System Initialization and Environment Abstraction The system operates within an abstract computing environment comprised of a processor, memory, and a network interface. Upon startup, the system establishes a historical knowledge base. This knowledge base is functionally defined as recording the mapping relationship between the "intrinsic characteristics" and "quality feedback" of acquired data samples, thus establishing a logical benchmark for subsequent value prediction.

[0025] II. Abstract Definition of Core Functional Modules The system forms a closed loop of "analysis-prediction-allocation-optimization" through the following three functionally coordinated logical modules.

[0026] 1. Value Prediction Module (Function Definition): Function 1: Analyze the inherent characteristics of acquired data. "Inherent characteristics" refers to the set of attributes extracted from the data content itself or its identifiers that characterize its potential value; Function 2: Based on the above analysis, an association inference mechanism is used to predict the expected utility value of data that has not yet been acquired. The core logic of this prediction is: if the unacquired data is similar in intrinsic characteristics to a certain type of acquired data, it is inferred that it may have similar utility.

[0027] 2. Dynamic Scheduling Module (Function Definition): Function 1: Receive the expected utility value from the value prediction module; Function 2: Based on the utility value, dynamically decide and allocate system resources available for data acquisition (such as communication bandwidth, storage space, and the priority of computing unit usage). Its core principle is to allocate more resources to data acquisition tasks with higher expected utility values.

[0028] 3. Feedback Optimization Module (Function Definition): Function 1: To assess the quality of successfully collected data and generate feedback on actual utility. Function 2: Feed the new relationship between "intrinsic characteristics and actual utility" back to the historical knowledge base to optimize the subsequent inference logic of the value prediction module and make its predictions more accurate.

[0029] III. Open-loop and closed-loop logic flow The modules described above form a self-optimizing closed loop through open data flow: Step S1: System initialization, loading the historical knowledge base; Step S2: For the target to be collected, the value prediction module analyzes its inherent characteristics and predicts its expected utility by referring to the historical knowledge base; Step S3: The dynamic scheduling module generates a resource allocation strategy based on the expected utility; Step S4: The system performs the data acquisition action; Step S5: The feedback optimization module evaluates the actual utility of the newly collected data and updates the historical knowledge base; Step S6: The process jumps to step S2 to achieve continuous loop optimization. Example 2

[0030] Adaptive data acquisition system with multi-dimensional feature fusion This embodiment expands upon the closed loop established in Embodiment 1 for the first time. Its core lies in introducing "multi-dimensional feature fusion" and "adaptive scheduling" to demonstrate how the invention enhances versatility through richer intrinsic feature analysis and more flexible decision-making logic, while maintaining the neutrality of the algorithm, data format, and hardware.

[0031] I. System Function Expansion and Positioning This embodiment aims to illustrate that the "analyze intrinsic features" and "dynamic scheduling" functions in Embodiment 1 can and typically be implemented by fusing multiple complementary feature dimensions and introducing an adaptive mechanism. This is an important and common technical path for optimizing system performance. This description is merely an exemplary way of implementing the claims and does not constitute any limitation.

[0032] II. Extended Definition of Core Functional Modules 1. Expansion of the value prediction module (multi-dimensional feature fusion analysis) Functional extension: The "analyzing the intrinsic characteristics of the acquired data" specifically includes extracting at least two different types of intrinsic feature dimensions in parallel or in series. Exemplary dimensions (not exhaustive): Dimension A: Content semantic features. These refer to the attributes that are extracted from the data content itself and reflect its theme, logic, information density, and other meanings.

[0033] Dimension B: Identifier structural features. These are attributes parsed from the identifiers or metadata of data objects that reflect their organization, hierarchical relationships, or pattern patterns.

[0034] Functional extension: The aforementioned "predicting expected utility" correlation inference mechanism can be operationally defined as a feature fusion function. This function accepts the aforementioned multiple feature dimensions as input and outputs a comprehensive utility prediction value. The specific form of this function (such as linear weighting, neural networks, etc.) is open.

[0035] 2. Extension of the dynamic scheduling module (multi-objective adaptive decision-making) Functional extension point: The decision logic of "dynamically allocating resources based on utility value" can be further incorporated into at least one system or environment status indicator as a common decision basis.

[0036] Exemplary status indicators (not exhaustive): real-time access load of the data source, current latency of the network channel, and remaining computing resources of the system itself.

[0037] Functional expansion points: Scheduling decision-making is a process of adaptive trade-offs among multiple objectives. The core objective is to maximize the overall expected utility of the collected data, while also considering objectives such as balancing the load of different data sources, ensuring the real-time performance of collection tasks, or avoiding excessive consumption of a single resource. Decision-making strategies (such as optimization algorithms and heuristic rules) are open.

[0038] III. Extended Adaptive Closed-Loop Logic The system adds adaptability to the closed-loop design of Example 1: In step S2 (value prediction), the system performs multi-dimensional feature fusion analysis to generate more robust utility predictions.

[0039] In step S3 (dynamic scheduling), the system simultaneously senses environmental status indicators and generates the current optimal resource allocation strategy based on multi-objective trade-offs.

[0040] When environmental status indicators change (e.g., a sudden increase in the load of a data source), the system can immediately trigger a re-decision starting from step S3 to adjust resource allocation without waiting for the complete next cycle.

[0041] IV. Legal and Technical Value of this Implementation Example Any technical solution that uses a combination of multi-dimensional features (such as content + structure) for value judgment, or considers multi-state environmental indicators (such as load and latency) for resource scheduling, will fall into the general technical path of 'multi-dimensional feature fusion + multi-objective adaptation' described in this embodiment. This embodiment further broadens the interpretative space of the claims, ensuring that the scope of protection covers all equivalent implementations, and naturally resists the risk of evasion caused by 'simplifying feature dimensions' or 'adopting a single static strategy'. Example 3

[0042] Federated collaborative distributed data acquisition network This embodiment is a second specific extension based on embodiment 2, applying the core functional paradigm of the present invention to a distributed network architecture.

[0043] Its core lies in defining a "federated collaboration" mechanism that enables multiple geographically or logically dispersed data collection nodes to share knowledge, collaboratively predict and optimize global resources without centralizing all raw data, while maintaining openness to specific communication protocols, node types and aggregation algorithms.

[0044] I. System Architecture Expansion and Positioning This embodiment aims to illustrate that the value prediction and dynamic scheduling functions described in this invention can and typically be deployed in a distributed system composed of multiple independent nodes. This demonstrates the universality and scalability of this invention in complex, large-scale real-world scenarios (such as edge computing and cross-regional data acquisition). This description focuses on the functional collaboration relationships between nodes and does not limit the network topology, communication protocol, or node hardware.

[0045] II. Distributed Extension Definition of Core Functional Modules 1. Distributed Value Prediction Module (Federated Knowledge Sharing and Aggregation) Functional expansion points: The system includes multiple data collection nodes, each maintaining a subset of the historical knowledge base locally (based on its locally collected data). The function of "analyzing intrinsic features and predicting utility" is achieved through collaboration between nodes.

[0046] Collaboration mechanism: Step 1 (Local Calculation): Each node generates a preliminary local utility prediction for the target to be collected based on its local knowledge base.

[0047] Step 2 (Knowledge Sharing): Each node uploads local prediction-related summary information (such as feature vectors and prediction confidence, rather than the original data) to a coordination unit via a communication network. The coordination unit can be a dedicated server or a peer node. Step 3 (Global Aggregation): The coordination unit uses an open aggregation algorithm (such as weighted averaging, model parameter fusion, secure multi-party computation, etc.) to synthesize the summary information from each node, generating a more comprehensive and robust global utility prediction.

[0048] Step 4 (Result Distribution): The coordination unit distributes the global prediction results to each relevant node to guide its local scheduling.

[0049] 2. Distributed dynamic scheduling module (global resource view and collaborative scheduling) Functional extension point: The decision to "dynamically allocate resources" is based on a global view consisting of the resource status of multiple nodes.

[0050] Collaboration mechanism: Step 1 (Status Reporting): Each node periodically reports its local resource status (such as remaining bandwidth, storage space, and CPU load) to the coordination unit.

[0051] Step 2 (Global Optimization): Based on global utility prediction and a global resource view, the coordination unit runs a global optimization model to calculate a resource allocation scheme aimed at maximizing the overall acquisition efficiency of the network. This scheme may instruct a node to contribute its idle resources to high-value global tasks.

[0052] Step 3 (Instruction Distribution): The coordination unit distributes the scheduling instructions to each node.

[0053] Step 4 (Local Execution): Each node adjusts the priority and resource usage of its local data collection tasks according to the received instructions.

[0054] III. Federal Collaborative Closed-Loop Logic While retaining the core steps of Examples 1 and 2, the closed-loop logic of the distributed system introduces a cross-node collaborative loop: Node-level loop: Each node independently completes a local closed loop from local prediction (based on shared knowledge) to local scheduling, collection, and feedback.

[0055] Network-level loop: The coordination unit periodically executes a network-level closed loop of "knowledge sharing and aggregation" -> "global optimization and scheduling" -> "instruction distribution", continuously synchronizing and optimizing the cognition and decision-making of the entire network.

[0056] These two loops operate asynchronously in parallel and reinforce each other, enabling the system to possess collective intelligence.

[0057] IV. Legal and Technical Value of this Implementation Example This embodiment explicitly extends the scope of protection of the present invention from single-machine systems to distributed network systems. By defining a set of open functional collaborative mechanisms—"federated knowledge sharing" and "global resource collaborative scheduling"—the present invention covers all data acquisition scenarios involving multi-node intelligent collaboration. If competitors construct any form of acquisition network that shares value perceptions and optimizes global resource allocation through intermediary coordination, regardless of the distributed algorithm they employ (such as federated learning or consensus mechanisms), its architecture essentially falls within the functional paradigm described in this embodiment. This provides a high-level patent barrier against circumvention through architectural designs that "split the system into multiple independent nodes but lack collaboration." Example 4

[0058] Integrated Operation Verification of Closed-Loop Intelligent Data Acquisition System This embodiment is illustrated in conjunction with the accompanying drawings ( Figures 1 to 5 The system integrates the higher-level functional logic described in Examples 1 to 3 into a specific, operable system instance to verify and fully support all the technical features defined in Claims 1 to 12, and to prove the complete feasibility of its technical solution.

[0059] I. System Hardware and Initial Deployment The system is deployed on a general-purpose electronic device equipped with a processor, memory, storage device, and network interface (see...). Figure 5 After startup, the system initializes and establishes a historical quality feedback database in the storage device. Its logical structure includes: 1. Data intrinsic feature library: used to store the "content semantic feature vector" and "identifier structure encoding" of the collected data.

[0060] 2. Utility Feedback Library: This library stores the "quality scores" and "task contribution records" corresponding to the aforementioned feature data. This infrastructure supports the implementation basis of the methods, systems, devices, and media described in claims 1, 2, 11, and 12.

[0061] II. Execution of the core closed-loop process in conjunction with the attached diagram The system strictly follows Figure 1 The overall process shown fully maps to the core "analysis-prediction-allocation" method defined in claim 1: Step 1: Value Prediction (corresponding to) Figure 1 , Figure 2 and Figure 3 ) When there is new data to be collected, the **utility prediction device** is activated.

[0062] The device first parses the identifier of the new target and extracts its **structural pattern features** (e.g., an encoding format that conforms to a specific rule). Simultaneously, if partial metadata or a summary is available, its **content semantic features** are analyzed using a **machine learning model**. For example, in a specific test verification, a **pre-trained language model** (such as a Transformer-based model) was used to perform this analysis, achieving good results. Those skilled in the art will understand that other types of machine learning models (such as deep neural networks, decision tree ensemble models, etc.) can also be used to achieve the same functionality.

[0063] The device then compares and performs correlation analysis between the characteristics of the new target and records in the historical database (e.g., ...). Figure 3 As shown, the new identifier is aggregated and matched with identifier patterns from historical high-quality data. A built-in machine learning model (such as a linear regression model) is used for comprehensive calculation to predict the expected utility value of the new data target. Specific values ​​such as '95 points' in the figure are merely examples to illustrate the principle and do not constitute any limitation on the actual effect of the invention.

[0064] This process is entirely based on the data's own "intrinsic characteristics" (content or structure). Step 2: Dynamic scheduling (corresponding to...) Figure 1 and Figure 2 ) The expected utility value is sent to the **resource allocation device**: this device simultaneously monitors the current **status indicators** of each data source, such as network latency (access pressure) and server response time (service indicators).

[0065] like Figure 2 The **decision engine** shown executes a **dynamic allocation** strategy based on utility values ​​and status indicators. For example, it allocates 70% of available bandwidth to the data task with the highest expected utility, and fine-tunes it in real time based on network latency. This logic involves real-time adjustments based on multi-dimensional input.

[0066] Step 3: Perform data collection and feedback optimization The system executes data collection tasks according to the scheduling strategy: if a data source connection fails during the process (**computing unit failure**), the system's **error handling device** will immediately trigger **task migration**, scheduling the data source to a backup data source.

[0067] Successfully collected data will be evaluated again by the semantic analysis model to generate an actual quality score, which will then be **feedback** to the historical database for updating and optimizing the prediction model, achieving closed-loop learning. This "feedback optimization" step enhances the system's adaptability and is an important component of the complete technical solution.

[0068] III. Examples of Distributed Collaboration Scenarios (corresponding) Figure 4 ) To demonstrate the universality of the claims, this embodiment further describes an extended scenario: the system can be deployed as follows Figure 4 The federal architecture shown.

[0069] Multiple edge data acquisition nodes (e.g., **edge nodes 1, 2, and 3**) execute the aforementioned "analysis-prediction" process locally and periodically upload feature model parameters (rather than raw data) to the **central coordinator**. The central coordinator aggregates these parameters to generate a more powerful global prediction model, which is then distributed to all nodes. This allows each node to benefit from global knowledge and achieve collaborative optimization. This scenario validates the feasibility and effectiveness of the method described in the claims within a distributed system.

[0070] IV. Summary of Comprehensive Support for the Claims of this Patent This embodiment, through the above specific and feasible integrated description, proves that all technical solutions in the claims are not abstract concepts, but can be implemented by those skilled in the art. The specific supporting relationships are as follows: Support for claims 1 and 2: The entire method flow and system structure of "predicting the expected utility of unacquired data based on the analysis of acquired data; allocating acquisition resources according to the expected utility" are fully reproduced.

[0071] Support for claims 3-7: It explicitly demonstrates that "analysis is based solely on intrinsic features" (such as semantic features, identifier structural patterns), and specifies two paths: using a **pre-trained language model** when the intrinsic features are content, and performing **association recognition and set utility evaluation** when they are patterns.

[0072] *Support for claims 8-10: Demonstrates how **dynamic allocation** is executed based on data source status indicators, and includes the complete process of responding to abnormal execution **reliability strategies** (such as task migration).

[0073] Support for claims 11-12: It is clarified that all the above functions can be implemented by an electronic device (claim 11) executing a program stored in a computer-readable storage medium.

[0074] In summary, this embodiment provides a complete implementation chain from abstract to concrete and from single to multiple, **strongly and fully supporting all the contents of claims 1 to 12**, ensuring that the generalization of the claims is substantially supported by the specification.

[0075] Disclaimer: Those skilled in the art should understand that the above embodiments are merely examples illustrating the core principles of this invention (i.e., "value prediction and dynamic resource allocation based on the inherent characteristics of data"). Any specific technical means mentioned in the specification (such as specific algorithm models, data structure examples, hardware configurations, or application scenarios) are not intended to limit this invention. As long as it contains the core logical closed loop of "analyzing inherent characteristics → predicting utility → dynamically allocating resources," any technical solution based on the same inventive concept disclosed in this invention, using different but functionally equivalent technical solutions, or modifications adapted to other application fields, can be naturally conceived and implemented by those skilled in the art after reading this specification without creative effort. The scope of protection of this invention is uniquely defined by the appended claims and by reasonable interpretation.

Claims

1. A data acquisition optimization method, characterized in that, include: Based on the analysis of the acquired data, predict the expected utility of the unacquired data; allocate collection resources according to the expected utility.

2. A data acquisition optimization system, characterized in that, include: A utility prediction device for predicting the expected utility of unacquired data based on the analysis of acquired data. A resource allocation device for allocating acquisition resources according to the expected utility.

3. The method or system according to claim 1 or 2, characterized in that, The analysis is based solely on the intrinsic characteristics of the data.

4. The method or system according to claim 3, characterized in that, The intrinsic characteristics include data content or data patterns.

5. The method or system according to claim 4, characterized in that, When the intrinsic feature is data content, the analysis is performed using a machine learning model.

6. The method or system according to claim 5, characterized in that, The machine learning model is a pre-trained language model.

7. The method or system according to claim 4, characterized in that, When the intrinsic feature is a data pattern, the analysis includes identifying the association patterns between data items and evaluating their collective utility.

8. The method or system according to claim 1 or 2, characterized in that, The allocation is dynamic.

9. The method or system according to claim 8, characterized in that, The dynamic allocation is based on the status indicators of the data source.

10. The method or system according to claim 1 or 2, characterized in that, It also includes responding to data collection anomalies and executing reliability strategies.

11. An electronic device comprising a processor and a memory, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 10.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 10.

Images