A cross-regional data acquisition method and system based on intelligent decision-making

Abstract

The present invention discloses a cross-regional data acquisition method and system based on intelligent decision-making, aiming to improve the accuracy and intelligence level of data acquisition between heterogeneous regions. The method comprises: first, obtaining data acquisition instructions issued by the data demand region and decomposing them into rigid demand vectors and elastic demand vectors; then, based on the rigid demand vectors, searching for matching data source features in the cross-regional data source registry, evaluating their data quality, and reconstructing features in combination with quality indicators to generate an enhanced feature vector set; then, jointly constructing a data acquisition strategy using the enhanced feature vectors and the elastic demand vectors; acquiring cross-regional target data according to the strategy and performing integrity testing on it; and if the test results show that the data integrity is lower than a preset value, performing predictive supplementation on the target data, ultimately obtaining optimized cross-regional data.

Description

Technical Field

[0001] The present invention relates to the field of data acquisition technology, and in particular to a cross-regional data acquisition method and system based on intelligent decision-making. Background Art

[0002] With the widespread deployment of information technology and data resources, cross-regional data integration and sharing are playing an increasingly important role in numerous fields, including traffic scheduling, weather monitoring, public safety, and energy management. However, current data acquisition methods, which often rely on static configurations or manual rules, struggle to cope with complex situations such as heterogeneous data sources across regions, dynamic demand changes, and uneven data quality. This results in low data utilization efficiency, insufficient reliability, and weak decision-making support capabilities.

[0003] Existing cross-regional data acquisition methods primarily rely on fixed field matching and static data source registration. These methods lack a deep understanding of data demand semantics and are unable to dynamically distinguish between rigid and flexible demands, limiting the flexibility of data scheduling. Furthermore, faced with inconsistent data source quality and update uncertainty, traditional methods lack effective data quality assessment and completion mechanisms, which can easily lead to data loss and delays, impacting the accuracy of subsequent analysis and decision-making models.

[0004] Furthermore, data acquisition strategies generally neglect collaborative feature modeling between demand and data sources, lacking systematic approaches based on feature reconstruction, quality enhancement, and intelligent strategy generation. This makes it impossible to achieve optimal resource allocation while meeting demand. Especially in multi-source, heterogeneous environments, how to satisfy rigid constraints while maximizing optimization potential under flexible demands remains a challenge and a pain point in technological development.

[0005] Therefore, there is an urgent need for a data acquisition method and system with intelligent understanding capabilities, support for dynamic data quality assessment and strategy optimization, in order to adapt to the current complex and changeable cross-regional data collaboration needs, improve acquisition efficiency and data integrity, and promote the further development of intelligent decision support systems. Summary of the Invention

[0006] In order to solve at least one of the above technical problems, the present invention proposes a cross-regional data acquisition method and system based on intelligent decision-making.

[0007] A first aspect of the present invention provides a cross-regional data acquisition method based on intelligent decision-making, comprising:

[0008] Obtaining a data acquisition instruction of a data demand area, and decomposing the data acquisition instruction into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network;

[0009] Searching a cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set;

[0010] Constructing a data acquisition strategy for the data demand area to the data source area according to the enhanced feature vector and the elastic demand vector;

[0011] Acquire cross-region target data according to the data acquisition strategy, perform integrity detection on the target data, and obtain detection results;

[0012] According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data.

[0013] In this solution, the data acquisition instruction of the data demand area is decomposed into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network, specifically:

[0014] Constructing an adversarial demand decomposition network, wherein the adversarial demand decomposition network includes a pre-trained demand encoding module, a demand discrimination module, and an adversarial training module;

[0015] Obtaining a data acquisition instruction from a data demand area, inputting the data acquisition instruction into the adversarial demand decomposition network, encoding the data acquisition instruction based on a demand encoding module, and generating an initial instruction feature vector;

[0016] Inputting the initial instruction feature vector into an adversarial training module for feature optimization, inputting the optimized initial instruction feature vector into a demand discrimination module to identify mandatory data fields and flexible matching fields of the initial instruction feature vector, and decomposing the initial instruction feature vector into an initial rigid demand vector and an initial elastic demand vector according to the mandatory data fields and the flexible matching fields;

[0017] The rigid demand vector includes the data source type, the required field name, and the field precision type; the flexible demand vector includes the field collection frequency, the data update time range, and the regional level to which the data source belongs.

[0018] In this solution, the cross-regional data source registry is searched based on the rigid demand vector to obtain matching data source features, the data quality of the data source features is evaluated, and features are reconstructed based on the data quality to obtain an enhanced feature vector set, specifically:

[0019] Performing a feature field matching search on a cross-regional data source registry based on the rigid demand vector to obtain a number of data sources with similar data field types, data source levels, and field precision characteristics, and grouping the data sources to construct a set of data sources to be evaluated;

[0020] Constructing a preliminary data quality vector by extracting field integrity, historical availability, update frequency and field consistency indicators for each data source in the set of data sources to be evaluated;

[0021] Perform a multi-dimensional comparison between the constructed preliminary data quality vector and the field precision type in the rigid requirement vector, calculate the matching score between the data source and the rigid requirement, and obtain an initial data source evaluation result;

[0022] When the matching score of any data source in the initial data source evaluation results is greater than a preset reference threshold, a multi-factor fusion reconstruction is performed based on the historical availability and field consistency index of the data source to obtain a data source feature vector with enhanced matching weight with the demand;

[0023] If the matching score is not greater than a preset reference threshold, the historical cross-regional data source evolution model is called to perform fitting inference on the update frequency trend and field missing rate trend of the data source, and the structure vector is predicted and reconstructed based on the inference results to obtain a reconstructed data source feature vector after fitting enhancement;

[0024] The data source feature vector enhanced with the demand matching weight and the reconstructed data source feature vector enhanced by fitting are normalized and aggregated to form an enhanced feature vector set.

[0025] In this solution, the data acquisition strategy for the data demand region to the data source region is constructed based on the enhanced feature vector and the elastic demand vector, specifically:

[0026] Based on the data source field quality, field consistency, and update frequency indicators contained in the enhanced feature vector, combined with the field collection frequency and data update time range in the elastic demand vector, a preliminary set of strategic plans matching the elastic demand vector is constructed, and the collaborative offset between each strategic plan and the enhanced feature vector is calculated;

[0027] When the collaborative offset of any strategy plan is less than the collaborative adaptation threshold, the strategy plan is marked as the preferred strategy, and the field collection frequency is adjusted according to the data update time threshold recorded in the preferred strategy field consistency, thereby generating the first-stage data acquisition strategy and performing a feasibility simulation on the first-stage data acquisition strategy;

[0028] When the field delay risk of the first-stage data acquisition strategy in the simulation is higher than the acquisition frequency change limit in the elastic demand vector, the target data source in the first-stage data acquisition strategy is re-screened for matching based on the enhanced feature vector, and after replacing the original data source, the strategy plan set is updated and the collaborative offset is regenerated. It is again determined whether there is a strategy plan that is less than the collaborative adaptation threshold;

[0029] If none of the strategic plans meet the collaborative adaptation threshold, then based on the field missing trend and historical availability in the enhanced feature vector, a feature evolution simulation is performed on the strategic plan set, and a second-stage optimization strategy is generated by adjusting the field collection frequency and data update time range. The optimization strategy completes the final strategy confirmation through the fitted field update stability curve;

[0030] A data acquisition strategy for the data demand area to the data source area is constructed based on the first-stage data acquisition strategy and the second-stage optimization strategy.

[0031] In this solution, the cross-region target data is acquired according to the data acquisition strategy, and the target data is subjected to integrity testing to obtain the test results, which are specifically:

[0032] Acquire corresponding cross-regional target data according to the preferred data source address and field requirements identified in the data acquisition strategy, wherein the cross-regional target data includes cross-regional traffic data and cross-regional climate data;

[0033] Comparing the target data with the field quality, field consistency, and update frequency indicators in the enhanced feature vector at a field level, extracting information on the field fill rate, data timestamp coverage, and field consistency performance of the target data to form a detection comparison vector;

[0034] According to the field collection frequency and data update time range included in the elastic demand vector, static matching judgment is performed on the detection comparison vector and the field standard indicators in the enhanced feature vector to obtain the matching status of each field;

[0035] By counting the proportion of fields that meet the integrity requirements in the field matching status, and combining the field filling rate, the target data is subjected to integrity detection to obtain a detection result.

[0036] In this solution, according to the detection results, when the integrity of the target data is lower than the preset value, the target data is predictively supplemented to obtain cross-regional optimized data, specifically:

[0037] Based on the insufficient field fill rate and missing data timestamp coverage identified in the detection results, the field consistency and update frequency indicators corresponding to the fields are extracted, and combined with the field missing trend in the enhanced feature vector, the input feature set of the prediction supplement model is constructed;

[0038] Based on the constructed input feature set, combined with the field collection frequency and data update time range in the elastic demand vector, the time series completion network is used to predict and infer missing field values ​​to generate a predicted filled field set.

[0039] Structural fusion of the predicted filling field set and the original target data, and performing supplementary quality judgment on the fusion result based on the consistency and timestamp integrity indicators of the fused fields;

[0040] When the supplementation quality meets the matching requirements of the field standard indicators in the enhanced feature vector, the corresponding cross-region optimization data is generated.

[0041] A second aspect of the present invention further provides a cross-regional data acquisition system based on intelligent decision-making, the system comprising: a memory and a processor, wherein the memory includes a cross-regional data acquisition method program based on intelligent decision-making, and when the cross-regional data acquisition method program based on intelligent decision-making is executed by the processor, the following steps are implemented:

[0042] Obtaining a data acquisition instruction of a data demand area, and decomposing the data acquisition instruction into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network;

[0043] Searching a cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set;

[0044] Constructing a data acquisition strategy for the data demand area to the data source area according to the enhanced feature vector and the elastic demand vector;

[0045] Acquire cross-region target data according to the data acquisition strategy, perform integrity detection on the target data, and obtain detection results;

[0046] According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data.

[0047] The present invention discloses a cross-regional data acquisition method and system based on intelligent decision-making, aiming to improve the accuracy and intelligence level of data acquisition between heterogeneous regions. The method comprises: first, obtaining data acquisition instructions issued by the data demand region and decomposing them into rigid demand vectors and elastic demand vectors; then, based on the rigid demand vectors, searching for matching data source features in the cross-regional data source registry, evaluating their data quality, and reconstructing features in combination with quality indicators to generate an enhanced feature vector set; then, jointly constructing a data acquisition strategy using the enhanced feature vectors and the elastic demand vectors; acquiring cross-regional target data according to the strategy and performing integrity testing on it; and if the test results show that the data integrity is lower than a preset value, performing predictive supplementation on the target data, ultimately obtaining optimized cross-regional data. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flow chart of a cross-regional data acquisition method based on intelligent decision-making according to the present invention is shown;

[0049] Figure 2 A flow chart showing the method of decomposing a data acquisition instruction into a rigid demand vector and an elastic demand vector according to the present invention is shown;

[0050] Figure 3 The flow chart of obtaining the detection result of the present invention is shown;

[0051] Figure 4 A block diagram of a cross-regional data acquisition system based on intelligent decision-making according to the present invention is shown. DETAILED DESCRIPTION

[0052] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0054] Figure 1 A flow chart of a cross-regional data acquisition method based on intelligent decision-making of the present invention is shown.

[0055] like Figure 1 As shown, the first aspect of the present invention provides a cross-regional data acquisition method based on intelligent decision-making, comprising:

[0056] S102, obtaining a data acquisition instruction of a data demand area, and decomposing the data acquisition instruction into a rigid demand vector and a flexible demand vector through an adversarial demand decomposition network;

[0057] S104: searching a cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating the data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set;

[0058] S106, constructing a data acquisition strategy for the data demand area to the data source area according to the enhanced feature vector and the elastic demand vector;

[0059] S108, acquiring cross-region target data according to the data acquisition strategy, performing integrity detection on the target data, and obtaining a detection result;

[0060] S110, according to the detection result, when the integrity of the target data is lower than a preset value, predictively supplement the target data to obtain cross-region optimized data.

[0061] It should be noted that by constructing an adversarial demand decomposition network to perform semantic understanding and structural decomposition of data acquisition instructions, it is possible to accurately distinguish between rigid and flexible demands, and improve the ability to parse complex data requests; secondly, based on the rigid demand vector, the cross-regional data source registry is searched, and through multi-dimensional data quality assessment and feature reconstruction, it is possible to enhance the quality information expression of the data source while ensuring matching accuracy, generate a stable and reliable enhanced feature vector set, and enhance the support capability for subsequent strategy generation; further, the enhanced feature vector and the flexible demand vector are combined to jointly construct the acquisition strategy, so that the strategy can flexibly balance field consistency, update frequency and acquisition cost, and improve the efficiency and adaptability of data scheduling; during the data acquisition process, the integrity of the target data is detected through field-level comparison, and data defects and update delays are effectively identified; finally, when the detection finds that the data integrity is insufficient, the missing fields can be inferred and filled based on the predictive supplement model, key fields are completed and optimized cross-regional data is generated, thereby achieving intelligent and quality-controlled cross-domain data acquisition effects.

[0062] Figure 2 The flowchart of the present invention for decomposing a data acquisition instruction into a rigid demand vector and an elastic demand vector is shown.

[0063] According to an embodiment of the present invention, the data acquisition instruction for acquiring the data demand area is decomposed into a rigid demand vector and an elastic demand vector by an adversarial demand decomposition network, specifically:

[0064] S202, constructing an adversarial demand decomposition network, wherein the adversarial demand decomposition network includes a pre-trained demand encoding module, a demand discrimination module, and an adversarial training module;

[0065] S204, obtaining a data acquisition instruction of the data demand area, inputting the data acquisition instruction into the adversarial demand decomposition network, encoding the data acquisition instruction based on the demand encoding module, and generating an initial instruction feature vector;

[0066] S206: Input the initial instruction feature vector into an adversarial training module for feature optimization, input the optimized initial instruction feature vector into a demand identification module to identify mandatory data fields and flexible matching fields of the initial instruction feature vector, and decompose the initial instruction feature vector into an initial rigid demand vector and an initial elastic demand vector based on the mandatory data fields and the flexible matching fields;

[0067] The rigid demand vector includes the data source type, the required field name, and the field precision type; the flexible demand vector includes the field collection frequency, the data update time range, and the regional level to which the data source belongs.

[0068] It should be noted that by constructing an adversarial demand decomposition network, we achieve intelligent semantic understanding and structural decomposition of data acquisition instructions, capable of breaking down complex natural language or structured instructions into rigid and flexible demand vectors. This network, comprised of a pre-trained demand encoding module, an adversarial training module, and a demand discrimination module, can collaboratively identify key elements and semantic boundaries within instructions. The demand encoding module is used to receive and embed raw data acquisition instructions, extract semantic feature vectors through a multi-layer neural network, and generate a preliminary demand vector representation, thereby generating an initial feature vector with a semantic dimension. The adversarial training module is composed of a main discriminant network and an auxiliary generation network. The main discriminant network is used to distinguish between rigid demand characteristics and flexible demand characteristics based on prior demand classification labels. The auxiliary generation network continuously optimizes the division boundary of the initial demand characteristics based on the loss information fed back by the main discriminant network. Through repeated training, it forms a stable adversarial discrimination capability for rigid and flexible demands, thereby improving the encoding representation's ability to separate field mandatory and substitutable features. The discrimination results of the adversarial training process are returned as feedback to the demand encoding module to further refine the demand encoding representation. Finally, the demand discrimination module performs fine-grained recognition of the optimized feature vectors based on labeled data or rule templates, distinguishing between mandatory data fields (such as data source type, field name, field precision, etc.) and flexibly matching fields (such as collection frequency, update time range, regional level, etc.), thereby accurately decomposing the original instructions into rigid demand vectors and flexible demand vectors. This method uses adversarial structure to enhance the model's sensitivity to boundary fuzzy semantics, effectively avoids feature confusion problems, and improves the accuracy and stability of demand analysis.

[0069] According to an embodiment of the present invention, searching the cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating the data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set is specifically as follows:

[0070] Performing a feature field matching search on a cross-regional data source registry based on the rigid demand vector to obtain a number of data sources with similar data field types, data source levels, and field precision characteristics, and grouping the data sources to construct a set of data sources to be evaluated;

[0071] Constructing a preliminary data quality vector by extracting field integrity, historical availability, update frequency and field consistency indicators for each data source in the set of data sources to be evaluated;

[0072] Perform a multi-dimensional comparison between the constructed preliminary data quality vector and the field precision type in the rigid requirement vector, calculate the matching score between the data source and the rigid requirement, and obtain an initial data source evaluation result;

[0073] When the matching score of any data source in the initial data source evaluation results is greater than a preset reference threshold, a multi-factor fusion reconstruction is performed based on the historical availability and field consistency index of the data source to obtain a data source feature vector with enhanced matching weight with the demand;

[0074] It should be noted that the cross-regional data source registration library refers to a unified resource indexing platform for centrally registering, storing and managing data source information and its characteristics from different regions.

[0075] If the matching score is not greater than a preset reference threshold, the historical cross-regional data source evolution model is called to perform fitting inference on the update frequency trend and field missing rate trend of the data source, and the structure vector is predicted and reconstructed based on the inference results to obtain a reconstructed data source feature vector after fitting enhancement;

[0076] It should be noted that the construction of the historical cross-regional data source evolution model is based on modeling and trend analysis of time series data of quality indicators for each data source over a historical period. Specifically, the update frequency and field missing rate of each cross-regional data source are regularly sampled over the historical period, generating time series data with time as the horizontal axis and quality indicators as the vertical axis. In terms of model selection, time series forecasting methods such as ARIMA can be used based on the stability and periodicity of the data to fit the changing trends of update frequency and the evolution of field missing rate, respectively, to obtain a series of predicted values ​​for future time periods. After trend inference, the predicted update frequency and field missing rate are used as input for feature reconstruction. Vectors are fused with known static field features (such as field type and data source level). A weighting function (such as linear fusion or dynamic weight allocation based on an attention mechanism) is used to construct a new structural vector, resulting in a fitted and enhanced reconstructed data source feature vector. This feature vector not only retains the original structural properties but also incorporates dynamic quality expectations based on historical evolution predictions. This provides a reasonable compensatory enhancement for data sources with currently low quality but trending upward.

[0077] The data source feature vector enhanced with the demand matching weight and the reconstructed data source feature vector enhanced by fitting are normalized and aggregated to form an enhanced feature vector set.

[0078] It should be noted that by performing field-level feature matching retrieval on rigid demand vectors (including field type, data source level, and field precision), candidate data sources that meet basic constraints can be accurately identified from the cross-regional data source registry. These data sources are then grouped by structural similarity or source characteristics to construct a set of data sources to be evaluated. Subsequently, key quality indicators for each candidate data source are extracted, namely field completeness (the inverse of field missing rate), historical availability (the proportion of data available within a historical time period), update frequency (a periodic indicator of data collection or release), and field consistency (the degree of semantic and format alignment with standard fields). This constructs a quantifiable preliminary data quality vector and generates a matching score. For data sources with scores exceeding a preset reference threshold, a multi-factor fusion reconstruction method is employed. This method weights and integrates historical availability and field consistency to highlight their adaptability to the requirements, generating a matching weighted feature vector with stable structure and enhanced semantics. For data sources with insufficient scores, a historical data source evolution model is introduced. Trend analysis is used to predict the evolution of their update frequency and field missing rate. A structural vector prediction reconstruction is then performed to assess their future availability and address any shortcomings. Finally, the above two types of feature vectors are normalized (such as by Z-score or Min-Max normalization) and fused into an enhanced feature vector set.

[0079] According to an embodiment of the present invention, the data acquisition strategy for the data demand region to the data source region is constructed based on the enhanced feature vector and the elastic demand vector, specifically:

[0080] Based on the data source field quality, field consistency, and update frequency indicators contained in the enhanced feature vector, combined with the field collection frequency and data update time range in the elastic demand vector, a preliminary set of strategic plans matching the elastic demand vector is constructed, and the collaborative offset between each strategic plan and the enhanced feature vector is calculated;

[0081] It should be noted that each strategy plan in the strategy plan set includes: the target data source identifier, the collection frequency configuration of the corresponding field, the expected data update time window, the field integrity weight adjustment parameter, and the dynamic tolerance range for adapting to the elastic demand vector. The process of constructing the strategy plan set includes: first, based on the field consistency, update frequency, and historical availability indicators of each data source in the enhanced feature vector, extracting a set of candidate data sources with high stability and medium-to-high matching; then, based on the parameter requirements of the field collection frequency and data update time range in the elastic demand vector, constructing multiple sets of collection configuration plans for each candidate data source; then, combining each set of collection configuration plans with the enhanced feature vector of its corresponding data source and evaluating their collaborative matching with the elastic demand vector; finally, retaining all configuration plans with an acceptable collaborative offset indicator as valid items in the strategy plan set, and adding a dynamic tolerance range configuration to plans with high field update uncertainty to ensure that the strategy has flexible scheduling capabilities under different network conditions and data states. The collaborative offset refers to the multi-dimensional matching difference measure between the data acquisition strategy plan and the enhanced feature vector in terms of each field feature, which is used to evaluate the degree of adaptation between the strategy and the actual data source capabilities.

[0082] When the collaborative offset of any strategy plan is less than the collaborative adaptation threshold, the strategy plan is marked as the preferred strategy, and the field collection frequency is adjusted according to the data update time threshold recorded in the preferred strategy field consistency, thereby generating the first-stage data acquisition strategy and performing a feasibility simulation on the first-stage data acquisition strategy;

[0083] When the field delay risk of the first-stage data acquisition strategy in the simulation is higher than the acquisition frequency change limit in the elastic demand vector, the target data source in the first-stage data acquisition strategy is re-screened for matching based on the enhanced feature vector, and after replacing the original data source, the strategy plan set is updated and the collaborative offset is regenerated. It is again determined whether there is a strategy plan that is less than the collaborative adaptation threshold;

[0084] It should be noted that when the coordination offset of any policy scenario is less than the coordination adaptation threshold, it indicates that the policy closely matches the enhanced feature vector in terms of field quality, field consistency, and update frequency, effectively meeting the collection frequency and update time requirements of the elastic demand. This situation arises from the fact that, faced with multiple candidate policies, the system must determine which options achieve the best match between supply and demand under multiple metrics. The coordination offset, as a measure of adaptability, intelligently selects the best overall policy. By marking such policies as preferred and adjusting the collection frequency based on the data update time threshold reflected in their field consistency, the first-stage data acquisition strategy is generated. However, if the first-stage data acquisition strategy exhibits a field latency risk exceeding the collection frequency variation limit in the elastic demand vector during simulation, this indicates that while the policy design theoretically meets the requirements, there is a risk of update delays or field failure during simulation execution. In environments where data source timeliness is uncertain or potentially volatile, static policies may cause field response lags due to data latency or network instability. At this time, by re-screening the target data source in the original strategy for matching, eliminating data sources with poor response capabilities, and updating the strategy plan set and collaborative offset evaluation, potential data delay problems can be effectively avoided.

[0085] If none of the strategic plans meet the collaborative adaptation threshold, then based on the field missing trend and historical availability in the enhanced feature vector, a feature evolution simulation is performed on the strategic plan set, and a second-stage optimization strategy is generated by adjusting the field collection frequency and data update time range. The optimization strategy completes the final strategy confirmation through the fitted field update stability curve;

[0086] A data acquisition strategy for the data demand area to the data source area is constructed based on the first-stage data acquisition strategy and the second-stage optimization strategy.

[0087] It should be noted that when none of the policy scenarios meet the collaborative adaptation threshold, the system initiates feature evolution simulation based on the field missingness trend and historical availability in the enhanced feature vector. First, the system analyzes the temporal pattern of the field missingness trend and historical availability fluctuations, constructs a dynamic data source behavior model, and simulates changes in field fill rates under different collection frequencies and update time ranges. A sliding window technique is then used to fit the data source update stability, generate a probability density function, and calculate the stability score for each field. Based on this, the system dynamically adjusts the collection frequency, extending the collection interval for low-stability fields and compressing the time window for high-stability fields to optimize transmission load. Multiple rounds of evolutionary simulations are then performed on the adjusted policy using the Monte Carlo method to evaluate its robustness under network delays and data source failures, selecting candidate policies that maintain a smooth transition in the stability curve. Ultimately, the policy is selected that minimizes the decline in historical availability while maximizing the second-order derivative of the stability curve and approaching zero.

[0088] Figure 3 The flowchart of the present invention for obtaining the detection result is shown.

[0089] According to an embodiment of the present invention, the cross-region target data is acquired according to the data acquisition strategy, and the integrity test is performed on the target data to obtain the test result, which is specifically:

[0090] S302, acquiring corresponding cross-regional target data according to the preferred data source address and field requirements identified in the data acquisition strategy, wherein the cross-regional target data includes cross-regional traffic data and cross-regional climate data;

[0091] S304: Perform a field-by-field comparison of the target data with the field quality, field consistency, and update frequency indicators in the enhanced feature vector, extracting information on the field fill rate, data timestamp coverage, and field consistency of the target data to form a detection comparison vector;

[0092] S306: Based on the field collection frequency and data update time range included in the elastic demand vector, statically match the detection comparison vector with the field standard indicators in the enhanced feature vector to obtain the matching status of each field;

[0093] S308 , performing integrity detection on the target data by counting the percentage of fields that meet the integrity requirement in the field matching status and combining the field filling rate to obtain a detection result.

[0094] It should be noted that the test results include the matching status of each field, the proportion of fields that meet the integrity requirements, the field fill rate, and the complete row score.

[0095] According to an embodiment of the present invention, when the integrity of the target data is lower than a preset value based on the detection result, the target data is predictively supplemented to obtain cross-region optimized data, specifically:

[0096] Based on the insufficient field fill rate and missing data timestamp coverage identified in the detection results, the field consistency and update frequency indicators corresponding to the fields are extracted, and combined with the field missing trend in the enhanced feature vector, the input feature set of the prediction supplement model is constructed;

[0097] Based on the constructed input feature set, combined with the field collection frequency and data update time range in the elastic demand vector, the time series completion network is used to predict and infer missing field values ​​to generate a predicted filled field set.

[0098] Structural fusion of the predicted filling field set and the original target data, and performing supplementary quality judgment on the fusion result based on the consistency and timestamp integrity indicators of the fused fields;

[0099] When the supplementation quality meets the matching requirements of the field standard indicators in the enhanced feature vector, the corresponding cross-region optimization data is generated.

[0100] It should be noted that the cross-region optimized data is used to replace the target data whose original integrity is lower than the preset value. The time series completion network builds a cross-region spatiotemporal prediction framework by integrating gradient screening, node prompts and knowledge graph technology. The network first calculates the source region data (representing the source region input feature matrix) and the target region data (representing the gradient of the input feature matrix of the target area), where represents the gradient of the source region loss function L (L uses the mean square error function) to the model parameter θ, Represents the target region gradient. Through the TSB (Target-SkewedBackward) strategy Decomposition into parallel components ( represents the Frobenius inner product) and the vertical component , to achieve beneficial gradient screening. The source area is the data source area, and the target area is the data demand area. and Represent the labels corresponding to the source region and the target region respectively; L represents the partial derivative of the loss function L; θ represents the partial derivative with respect to the parameter vector θ; Represents a preset spatiotemporal prediction model, such as a spatiotemporal graph convolutional network or a graph wave network.

[0101] In the feature processing stage, the network processes node features (N is the number of nodes, T is the time step, D is the feature dimension, and R indicates that the three-dimensional tensor of X is a set of real numbers) Add time embedding E T (E T is the time position code) to get the node input vector The improved self-attention layer replaces the feedforward network with a graph convolutional layer (GNN), (Attention means multi-head attention) extracts spatiotemporal features and connects them through residuals (Norm indicates layer normalization) and secondary graph convolution to obtain node patterns This architecture that alternates between graph convolution and self-attention achieves the collaborative extraction of local spatial features and global temporal patterns.

[0102] The node hint module utilizes a pre-built hint library (K is the number of prompt patterns) Generate prompts (B T is the transposed matrix), is the kth hint, and finally all the node hints With node features It can be spliced ​​in the feature dimension as (∣ represents the concatenation operation). At the same time, the knowledge graph A is constructed K , whose elements ( represents the Frobenius inner product, A represents the node association. K It is a knowledge graph built based on the similarity between node prompts.

[0103] The target bias update strategy is used in the training phase to determine the target gradient (γ1,γ2 are the learning rates of the target region and the source region respectively, dir returns 1 when the directions are the same, otherwise 0). ( is the true value, is the predicted value) and Evaluating the prediction performance, experiments show that the network significantly improves the quality of cross-regional spatiotemporal data completion. RMSE is the root mean square error, Mean is the mean absolute error, and n is the total number of data points for the corresponding data item in the source region.

[0104] According to an embodiment of the present invention, the further embodiment includes:

[0105] Establishing a clock synchronization layer between the data demand area and the data source area according to the preferred data source address identified in the data acquisition strategy, monitoring the transmission delay of the timestamp of the target data through the clock synchronization layer, and generating a dynamic time offset;

[0106] Calculating a tolerance compensation threshold for data integrity detection based on the dynamic time offset and the data update time range in the elastic demand vector, injecting the tolerance compensation threshold into the update frequency indicator in the enhanced feature vector for dynamic correction, and generating a delay compensation feature vector;

[0107] When acquiring cross-region target data according to the data acquisition strategy, delay calibration is performed on the actual arrival timestamp of the target data through the clock synchronization layer to obtain calibrated timestamp data;

[0108] Performing a temporal coverage comparison between the calibrated timestamp data and the update frequency index in the delay compensation feature vector, extracting a timestamp anomaly field set based on the comparison result, and inputting the timestamp anomaly field set into a preset temporal integrity detection model;

[0109] The temporal integrity detection model is combined with the tolerance compensation threshold to perform network delay attribution judgment on the timestamp anomaly field set, and a temporal integrity detection result is output.

[0110] According to an embodiment of the present invention, the clock synchronization layer monitors the transmission delay of the timestamp of the target data to generate the dynamic time offset, specifically:

[0111] Before executing the data acquisition strategy, a synchronous detection request is initiated to the target data source area according to the preferred data source address, and the bidirectional transmission delay of the data packet between the egress gateway of the data source area and the ingress gateway of the data demand area is captured through the synchronous detection request;

[0112] Calculating an average jitter coefficient of the network path based on the bidirectional transmission delay, and generating a transmission delay baseline model in combination with the field collection frequency in the elasticity demand vector;

[0113] When the target data is actually acquired, the server timestamp of the data packet leaving the data source area and the client timestamp of the data packet arriving at the data demand area are recorded in real time through the clock synchronization layer. The server timestamp and the client timestamp are input into the transmission delay baseline model for offset fitting, and a dynamic time offset with a confidence interval is output.

[0114] According to an embodiment of the present invention, performing temporal coverage comparison on the calibrated timestamp data and the delay compensation feature vector is specifically as follows:

[0115] Extracting a time window requirement for updating a frequency indicator from the delay compensation feature vector, forward-expanding the time window requirement according to the dynamic time offset, and generating a time coverage interval after compensation;

[0116] Splitting the calibrated timestamp data into discrete time series by field, and calculating the degree of overlap between each discrete time series point and the compensated time coverage interval;

[0117] If a discrete time series point exceeds the post-compensation time coverage interval, it is marked as a timestamp anomaly field;

[0118] The number and distribution density of timestamp anomaly fields are counted, and a timestamp anomaly field set is constructed in combination with the tolerance compensation threshold.

[0119] It's important to note that in real-time cross-regional data acquisition scenarios, network transmission delays inevitably cause data timestamps to shift relative to the server-generated time. When the offset exceeds the update time range threshold set by the elastic demand, the integrity check mechanism will mistakenly mark valid data as missing. This is especially true for time-sensitive data such as traffic monitoring and weather warnings. Such misjudgments trigger invalid retransmission requests or incorrect data replenishment, which not only wastes network resources but also erodes the reliability of the enhanced feature vector, leading to distorted subsequent policy decisions. This solution builds a dynamic clock synchronization layer to generate a dynamic time offset with confidence intervals based on bidirectional transmission delay and network jitter characteristics, accurately quantifying path delay. This offset is then combined with elastic demand to generate a tolerance compensation threshold, which inversely corrects the update frequency indicator in the enhanced feature vector to form a delay-compensated feature vector that is adaptive to network delays. Finally, through dynamic expansion of the temporal coverage interval and anomaly attribution models, it distinguishes between true data missingness and delay artifacts. This reduces the error rate of time window detection and significantly improves data utilization.

[0120] Figure 4 A block diagram of a cross-regional data acquisition system based on intelligent decision-making according to the present invention is shown.

[0121] A second aspect of the present invention further provides a cross-regional data acquisition system 4 based on intelligent decision-making, the system comprising: a memory 41 and a processor 42, wherein the memory includes a cross-regional data acquisition method program based on intelligent decision-making, and when the cross-regional data acquisition method program based on intelligent decision-making is executed by the processor, the following steps are implemented:

[0122] Obtaining a data acquisition instruction of a data demand area, and decomposing the data acquisition instruction into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network;

[0123] Searching a cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set;

[0124] Constructing a data acquisition strategy for the data demand area to the data source area according to the enhanced feature vector and the elastic demand vector;

[0125] Acquire cross-region target data according to the data acquisition strategy, perform integrity detection on the target data, and obtain detection results;

[0126] According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data.

[0127] The present invention discloses a cross-regional data acquisition method and system based on intelligent decision-making, aiming to improve the accuracy and intelligence level of data acquisition between heterogeneous regions. The method comprises: first, obtaining data acquisition instructions issued by the data demand region and decomposing them into rigid demand vectors and elastic demand vectors; then, based on the rigid demand vectors, searching for matching data source features in the cross-regional data source registry, evaluating their data quality, and reconstructing features in combination with quality indicators to generate an enhanced feature vector set; then, jointly constructing a data acquisition strategy using the enhanced feature vectors and the elastic demand vectors; acquiring cross-regional target data according to the strategy and performing integrity testing on it; and if the test results show that the data integrity is lower than a preset value, performing predictive supplementation on the target data, ultimately obtaining optimized cross-regional data.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0129] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0130] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0131] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0132] Alternatively, if the integrated units described above are implemented as software modules and sold or used as standalone products, they can also be stored on a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product, stored on a storage medium, includes instructions for enabling a computer device (such as a personal computer, server, or network device) to execute all or part of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as removable storage devices, ROM, RAM, magnetic disks, or optical disks.

[0133] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A cross-regional data acquisition method based on intelligent decision-making, characterized in that: The following steps are involved: Obtaining a data acquisition instruction of a data demand area, and decomposing the data acquisition instruction into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network; Searching a cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set; Constructing a data acquisition strategy for the data demand area to the data source area according to the enhanced feature vector and the elastic demand vector; Acquire cross-region target data according to the data acquisition strategy, perform integrity detection on the target data, and obtain detection results; According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data.

2. The cross-regional data acquisition method based on intelligent decision-making according to claim 1 is characterized in that: The data acquisition instruction of the data demand area is decomposed into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network, specifically: Constructing an adversarial demand decomposition network, wherein the adversarial demand decomposition network includes a pre-trained demand encoding module, a demand discrimination module, and an adversarial training module; Obtaining a data acquisition instruction from a data demand area, inputting the data acquisition instruction into the adversarial demand decomposition network, encoding the data acquisition instruction based on a demand encoding module, and generating an initial instruction feature vector; Inputting the initial instruction feature vector into an adversarial training module for feature optimization, inputting the optimized initial instruction feature vector into a demand discrimination module to identify mandatory data fields and flexible matching fields of the initial instruction feature vector, and decomposing the initial instruction feature vector into an initial rigid demand vector and an initial elastic demand vector according to the mandatory data fields and the flexible matching fields; The rigid demand vector includes the data source type, the required field name, and the field precision type; the flexible demand vector includes the field collection frequency, the data update time range, and the regional level to which the data source belongs.

3. The cross-regional data acquisition method based on intelligent decision-making according to claim 1 is characterized in that: The cross-regional data source registry is searched according to the rigid demand vector to obtain matching data source features, the data quality of the data source features is evaluated, and features are reconstructed according to the data quality to obtain an enhanced feature vector set, specifically: Performing a feature field matching search on a cross-regional data source registry based on the rigid demand vector to obtain a number of data sources with similar data field types, data source levels, and field precision characteristics, and grouping the data sources to construct a set of data sources to be evaluated; Constructing a preliminary data quality vector by extracting field integrity, historical availability, update frequency and field consistency indicators for each data source in the set of data sources to be evaluated; Perform a multi-dimensional comparison between the constructed preliminary data quality vector and the field precision type in the rigid requirement vector, calculate the matching score between the data source and the rigid requirement, and obtain an initial data source evaluation result; When the matching score of any data source in the initial data source evaluation results is greater than a preset reference threshold, a multi-factor fusion reconstruction is performed based on the historical availability and field consistency index of the data source to obtain a data source feature vector with enhanced matching weight with the demand; If the matching score is not greater than a preset reference threshold, the historical cross-regional data source evolution model is called to perform fitting inference on the update frequency trend and field missing rate trend of the data source, and the structure vector is predicted and reconstructed based on the inference results to obtain a reconstructed data source feature vector after fitting enhancement; The data source feature vector enhanced with the demand matching weight and the reconstructed data source feature vector enhanced by fitting are normalized and aggregated to form an enhanced feature vector set.

4. The cross-regional data acquisition method based on intelligent decision-making according to claim 1 is characterized in that: The data acquisition strategy of the data demand region to the data source region is constructed according to the enhanced feature vector and the elastic demand vector, specifically: Based on the data source field quality, field consistency, and update frequency indicators contained in the enhanced feature vector, combined with the field collection frequency and data update time range in the elastic demand vector, a preliminary set of strategic plans matching the elastic demand vector is constructed, and the collaborative offset between each strategic plan and the enhanced feature vector is calculated; When the collaborative offset of any strategy plan is less than the collaborative adaptation threshold, the strategy plan is marked as the preferred strategy, and the field collection frequency is adjusted according to the data update time threshold recorded in the preferred strategy field consistency, thereby generating the first-stage data acquisition strategy and performing a feasibility simulation on the first-stage data acquisition strategy; When the field delay risk of the first-stage data acquisition strategy in the simulation is higher than the acquisition frequency change limit in the elastic demand vector, the target data source in the first-stage data acquisition strategy is re-screened for matching based on the enhanced feature vector, and after replacing the original data source, the strategy plan set is updated and the collaborative offset is regenerated. It is again determined whether there is a strategy plan that is less than the collaborative adaptation threshold; If none of the strategic plans meet the collaborative adaptation threshold, then based on the field missing trend and historical availability in the enhanced feature vector, a feature evolution simulation is performed on the strategic plan set, and a second-stage optimization strategy is generated by adjusting the field collection frequency and data update time range. The optimization strategy completes the final strategy confirmation through the fitted field update stability curve; A data acquisition strategy for the data demand area to the data source area is constructed based on the first-stage data acquisition strategy and the second-stage optimization strategy.

5. The cross-regional data acquisition method based on intelligent decision-making according to claim 1 is characterized in that: The cross-region target data is obtained according to the data acquisition strategy, and the integrity test is performed on the target data to obtain the test result, which is specifically: Acquire corresponding cross-regional target data according to the preferred data source address and field requirements identified in the data acquisition strategy, wherein the cross-regional target data includes cross-regional traffic data and cross-regional climate data; Comparing the target data with the field quality, field consistency, and update frequency indicators in the enhanced feature vector at a field level, extracting information on the field fill rate, data timestamp coverage, and field consistency performance of the target data to form a detection comparison vector; According to the field collection frequency and data update time range included in the elastic demand vector, static matching judgment is performed on the detection comparison vector and the field standard indicators in the enhanced feature vector to obtain the matching status of each field; By counting the proportion of fields that meet the integrity requirements in the field matching status, and combining the field filling rate, the target data is subjected to integrity detection to obtain a detection result.

6. The cross-regional data acquisition method based on intelligent decision-making according to claim 1 is characterized in that: According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data, specifically: Based on the insufficient field fill rate and missing data timestamp coverage identified in the detection results, the field consistency and update frequency indicators corresponding to the fields are extracted, and combined with the field missing trend in the enhanced feature vector, the input feature set of the prediction supplement model is constructed; Based on the constructed input feature set, combined with the field collection frequency and data update time range in the elastic demand vector, the time series completion network is used to predict and infer missing field values ​​to generate a predicted filled field set. Structural fusion of the predicted filling field set and the original target data, and performing supplementary quality judgment on the fusion result based on the consistency and timestamp integrity indicators of the fused fields; When the supplementation quality meets the matching requirements of the field standard indicators in the enhanced feature vector, the corresponding cross-region optimization data is generated.

7. A cross-regional data acquisition system based on intelligent decision-making, characterized in that: The cross-regional data acquisition system based on intelligent decision-making includes a storage device and a processor. The storage device includes a cross-regional data acquisition method program based on intelligent decision-making. When the cross-regional data acquisition method program based on intelligent decision-making is executed by the processor, the following steps are implemented: Obtaining a data acquisition instruction of a data demand area, and decomposing the data acquisition instruction into a rigid demand vector and an elastic demand vector through an adversarial demand decomposition network; Searching a cross-regional data source registry based on the rigid demand vector to obtain matching data source features, evaluating data quality of the data source features, and reconstructing features based on the data quality to obtain an enhanced feature vector set; Constructing a data acquisition strategy for the data demand area to the data source area according to the enhanced feature vector and the elastic demand vector; Acquire cross-region target data according to the data acquisition strategy, perform integrity detection on the target data, and obtain detection results; According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data.

8. The cross-regional data acquisition system based on intelligent decision-making according to claim 7 is characterized in that: The data acquisition strategy of the data demand region to the data source region is constructed according to the enhanced feature vector and the elastic demand vector, specifically: Based on the data source field quality, field consistency, and update frequency indicators contained in the enhanced feature vector, combined with the field collection frequency and data update time range in the elastic demand vector, a preliminary set of strategic plans matching the elastic demand vector is constructed, and the collaborative offset between each strategic plan and the enhanced feature vector is calculated; When the collaborative offset of any strategy plan is less than the collaborative adaptation threshold, the strategy plan is marked as the preferred strategy, and the field collection frequency is adjusted according to the data update time threshold recorded in the preferred strategy field consistency, thereby generating the first-stage data acquisition strategy and performing a feasibility simulation on the first-stage data acquisition strategy; When the field delay risk of the first-stage data acquisition strategy in the simulation is higher than the acquisition frequency change limit in the elastic demand vector, the target data source in the first-stage data acquisition strategy is re-screened for matching based on the enhanced feature vector, and after replacing the original data source, the strategy plan set is updated and the collaborative offset is regenerated. It is again determined whether there is a strategy plan that is less than the collaborative adaptation threshold; If none of the strategic plans meet the collaborative adaptation threshold, then based on the field missing trend and historical availability in the enhanced feature vector, a feature evolution simulation is performed on the strategic plan set, and a second-stage optimization strategy is generated by adjusting the field collection frequency and data update time range. The optimization strategy completes the final strategy confirmation through the fitted field update stability curve; A data acquisition strategy for the data demand area to the data source area is constructed based on the first-stage data acquisition strategy and the second-stage optimization strategy.

9. The cross-regional data acquisition system based on intelligent decision-making according to claim 7 is characterized in that: The cross-region target data is obtained according to the data acquisition strategy, and the integrity test is performed on the target data to obtain the test result, which is specifically: Acquire corresponding cross-regional target data according to the preferred data source address and field requirements identified in the data acquisition strategy, wherein the cross-regional target data includes cross-regional traffic data and cross-regional climate data; Comparing the target data with the field quality, field consistency, and update frequency indicators in the enhanced feature vector at a field level, extracting information on the field fill rate, data timestamp coverage, and field consistency performance of the target data to form a detection comparison vector; According to the field collection frequency and data update time range included in the elastic demand vector, static matching judgment is performed on the detection comparison vector and the field standard indicators in the enhanced feature vector to obtain the matching status of each field; By counting the proportion of fields that meet the integrity requirements in the field matching status, and combining the field filling rate, the target data is subjected to integrity detection to obtain a detection result.

10. The cross-regional data acquisition system based on intelligent decision-making according to claim 7, characterized in that: According to the detection results, when the integrity of the target data is lower than a preset value, the target data is predictively supplemented to obtain cross-regional optimized data, specifically: Based on the insufficient field fill rate and missing data timestamp coverage identified in the detection results, the field consistency and update frequency indicators corresponding to the fields are extracted, and combined with the field missing trend in the enhanced feature vector, the input feature set of the prediction supplement model is constructed; Based on the constructed input feature set, combined with the field collection frequency and data update time range in the elastic demand vector, the time series completion network is used to predict and infer missing field values ​​to generate a predicted filled field set. Structural fusion of the predicted filling field set and the original target data, and performing supplementary quality judgment on the fusion result based on the consistency and timestamp integrity indicators of the fused fields; When the supplementation quality meets the matching requirements of the field standard indicators in the enhanced feature vector, the corresponding cross-region optimization data is generated.

Images