Smart city-oriented multi-source heterogeneous address data intelligent modeling method and system

Abstract

The invention discloses a smart city-oriented multi-source heterogeneous address data intelligent modeling method and system, relates to the technical field of multi-source heterogeneous address data processing, and can be applied to smart city address modeling. According to the specific implementation scheme, based on multi-source heterogeneous address data reflecting geographic positioning characteristics in urban resources, the address information adding condition and the text address updating condition are analyzed, and whether address entity and data time-varying association verification is carried out or not is judged according to the analysis result; if yes, whether modeling decision verification is triggered or not is judged; and if the judgment result is no, executing address data time sequence correction based on the analysis result of the address information adding condition and the text address updating condition, and after the address data time sequence correction is finished, performing address entity and data time-varying association verification. According to the technology provided by the invention, the problem of low processing accuracy of the multi-source heterogeneous address data during modeling based on the multi-source heterogeneous address data is solved.

Description

Intelligent Modeling Method and System for Multi-Source Heterogeneous Address Data in Smart Cities Technical Field

[0001] This invention relates to the field of multi-source heterogeneous address data processing technology, and in particular to a method and system for intelligent modeling of multi-source heterogeneous address data for smart cities. Background Technology

[0002] As one of the most core urban public data resources and spatiotemporal infrastructures, addresses have become an indispensable basic information resource and an important tool in the process of precise prevention and control and improving governance. Especially in provinces with a large number of ethnic groups, multi-source heterogeneous address data (such as administrative division addresses, urban planning addresses, highway addresses, etc.) present prominent contradictions due to differences in language, culture, habits and complex historical evolution. The core significance of building models based on multi-source heterogeneous address data is to transform the messy address text into standardized, computable and associative spatial intelligence, thereby unleashing the data potential of cross-department, cross-language and cross-scenario.

[0003] Multi-source heterogeneous address data has practical significance in breaking down information silos, supporting precise governance of public security and crime prevention, empowering cross-domain collaboration, and improving decision-making efficiency. It is also the core data foundation for the digital transformation of smart cities. To achieve precise governance in smart cities, existing methods first perform data preprocessing and standardization. Natural language processing technology is used, combined with a dictionary-based parser and sequence labeling algorithms based on deep learning models such as BERT (Bidirectional Encoder Representations from Transformers), to achieve entity recognition and component analysis of unstructured address text. Address semantic vectors are constructed based on authoritative address databases, and error correction and normalization are performed to form standardized addresses conforming to GIS (Geographic Information System) standards. Next, geocoding and spatial positioning are performed. Using string similarity algorithms and spatial indexing technology, standardized addresses are converted into latitude and longitude coordinates in batches. Through multimodal deep learning, text addresses, map images such as POI (Point of Interest) outlines, and GPS (Global Positioning System) coordinates are integrated. The system (Global Positioning System) trajectory points and spatiotemporal features are used to represent and align data from multiple modalities, thereby completing the accurate mapping from semantic description to spatial entities. Finally, a cross-source data association model is established, which uses multimodal deep learning models, such as heterogeneous graph neural networks and temporal graph neural networks, to learn the complex patterns of multi-source data nodes (standardized address nodes, POI entity nodes, population information nodes, etc.) and their corresponding associations (spatial topological relationships, attribute associations, temporal dependencies).

[0004] The aforementioned technologies suffer from at least the following technical problems: During modeling based on multi-source heterogeneous address data, dynamically changing multi-source heterogeneous address data may cause asynchronous time-series data processing of multimodal data. For example, text addresses may be updated but map images and GPS track points may not be updated synchronously, and dynamically added address information may not be quickly associated with surrounding real-time GPS tracks and POI images. Existing technologies typically only perform basic representation and simple alignment of multimodal data, resulting in insufficient dynamic cross-modal association capabilities. They cannot effectively model the time-varying association relationship between address entities and multimodal data, which may lead to address entity recognition errors, deviation of standardization results from true semantics, and spatial mapping position drift. This results in distortion of semantic error correction and normalization results, failure of alignment logic based on outdated maps or POI information, failure of accurate mapping between text addresses and spatial entities, and distortion of cross-source data node relationship modeling. Consequently, the accuracy of multi-source heterogeneous address data processing is low when modeling based on multi-source heterogeneous address data. Summary of the Invention

[0005] This invention provides an intelligent modeling method and system for multi-source heterogeneous address data in smart cities, which can improve the accuracy of multi-source heterogeneous address data processing, thereby solving the problem of low accuracy in multi-source heterogeneous address data processing in existing technologies. The technical solution provided by this application is as follows: According to the first aspect of this application, an intelligent modeling method for multi-source heterogeneous address data in smart cities is provided. This method includes: analyzing the addition of address information and the updating of text addresses based on multi-source heterogeneous address data reflecting the geographic positioning characteristics of urban resources; determining whether to perform time-varying association verification between address entities and data based on the analysis results, in order to determine the address entity identification error, the deviation of standardization results from semantics, and the spatial mapping location drift corresponding to the multi-source heterogeneous address data; if it is determined to perform, determining whether to trigger modeling decision verification based on the results of the time-varying association verification between address entities and data; if it is determined not to perform, performing address data time-series correction based on the analysis results of the addition of address information and the updating of text addresses, and performing address entity and data time-varying association verification after the address data time-series correction is completed.

[0006] According to another aspect of this application, a multi-source heterogeneous address data intelligent modeling system for smart cities is provided. This system applies a multi-source heterogeneous address data intelligent modeling method for smart cities. The system includes: a time-varying association determination module, a modeling decision verification determination module, and a time-series correction module. The time-varying association determination module analyzes the addition of address information and the updating of text addresses based on multi-source heterogeneous address data reflecting the geographic location characteristics of urban resources. Based on the analysis results, it determines whether to perform time-varying association verification between address entities and data to identify address entity identification errors, deviations from semantics in standardization results, and spatial mapping location drift in the multi-source heterogeneous address data. The modeling decision verification determination module, if determined to perform, determines whether to trigger modeling decision verification based on the results of the time-varying association verification between address entities and data. The time-series correction module, if determined not to perform, performs time-series correction of address data based on the analysis results of the addition of address information and the updating of text addresses, and performs time-varying association verification between address entities and data after the time-series correction is completed.

[0007] The beneficial effects of the technical solution provided by the embodiments of the present invention include at least the following: 1. By analyzing the addition of address information and the updating of text addresses through multi-source heterogeneous address data reflecting the geographic positioning characteristics of urban resources, and determining whether to perform time-varying association verification between address entities and data based on the analysis results, in order to identify address entity recognition errors, deviations of standardization results from semantics, and spatial mapping location drift corresponding to the multi-source heterogeneous address data, this helps to achieve accurate capture of time-varying features of multi-source address data, pre-screening of potential quality defects, and differentiated triggering of association verification; if it is determined that it should be performed, then based on the results of the time-varying association verification between address entities and data, a judgment is made. Determining whether to trigger modeling decision verification helps to achieve dynamic quality control of the modeling process, quantitative evaluation of the credibility of modeling results, and rapid tracing and correction of unqualified modeling data. If it is determined not to proceed, address data time-series correction is performed based on the analysis results of address information addition and text address update. After the address data time-series correction is completed, address entity and data time-varying association verification is performed. This helps to achieve closed-loop repair of defects in multi-source heterogeneous address data, completeness and comprehensiveness of association verification, and improvement of the temporal continuity and spatiotemporal consistency of subsequent modeling data. It also helps to solve the problem of low accuracy in processing multi-source heterogeneous address data in existing technologies.

[0008] 2. By analyzing the addition of address information and the update of text addresses, anomaly judgment values ​​for address addition and text address update are obtained. When the anomaly judgment value for address addition exceeds the preset value, or the anomaly judgment value for text address update exceeds the predefined value, address data timing correction is selectively performed. Compared with existing technologies that suffer from over-correction and poor correction targeting, this method helps to achieve precise triggering of timing correction, deep adaptation of correction strategies and anomaly types, and reduces ineffective correction behavior. This improves the efficiency of processing multi-source heterogeneous address data and provides more accurate and efficient basic data support for subsequent modeling stages.

[0009] 3. By performing time-varying association verification between address entities and data, this method monitors address entity identification errors, deviations from semantics in standardized results corresponding to multi-source heterogeneous address data, and spatial mapping position drift. Compared to existing technologies that have a single data quality monitoring dimension, focusing only on format standardization, easily overlooking core defects such as semantic deviation and spatial drift, and lacking a full-link closed-loop verification mechanism, this method helps to achieve comprehensive monitoring of multi-source address data quality, improves the accuracy of early warning and location of potential defects, and thus ensures the accuracy of rapid tracing of abnormal multi-source heterogeneous address data.

[0010] 4. By monitoring the spatial mapping position drift of multi-source heterogeneous address data, the mapping drift value is obtained. When the mapping drift value is greater than the predefined qualified mapping drift value, the index grid is reduced. Conversely, the average retrieval rate of multi-source heterogeneous address data is monitored. When the average retrieval rate of multi-source heterogeneous address data is greater than the predefined maximum multi-source data retrieval rate, the index grid is increased. Compared with the fixed index grid parameters in the existing technology, this helps to improve the dynamic balance between spatial mapping accuracy and retrieval efficiency of multi-source heterogeneous address data, thereby improving the resource utilization efficiency of address retrieval tasks and the overall performance of multi-source heterogeneous address data fusion processing.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided for a better understanding of this solution and do not constitute a limitation of this application. Specifically: Figure 1 is a flowchart of the intelligent modeling method for multi-source heterogeneous address data for smart cities provided in an embodiment of the present invention; Figure 2 is a general overview diagram of the intelligent modeling method for multi-source heterogeneous address data for smart cities provided in an embodiment of the present invention; Figure 3 is a schematic diagram of address data timing correction in the intelligent modeling method for multi-source heterogeneous address data for smart cities provided in an embodiment of the present invention; Figure 4 is a schematic diagram of multimodal deep learning in the intelligent modeling method for multi-source heterogeneous address data for smart cities provided in an embodiment of the present invention; Figure 5 is a structural schematic diagram of the intelligent modeling system for multi-source heterogeneous address data for smart cities provided in an embodiment of the present invention. Detailed Implementation

[0013] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0014] This invention provides an intelligent modeling method for multi-source heterogeneous address data in smart cities. As shown in Figure 1, the flowchart of this method includes the following steps: Time-varying association determination: In a specified smart city address modeling scenario, based on multi-source heterogeneous address data reflecting the geographic location characteristics of urban resources, the method analyzes the addition of address information and the updating of text addresses. Based on the analysis results, it determines whether to perform time-varying association verification between address entities and data. This is to identify address entity identification errors, deviations from semantics in standardized results, and spatial mapping location drift in the multi-source heterogeneous address data. By performing time-varying association determination, it helps to accurately identify time-varying address characteristics, provide early warnings of potential data quality issues (such as batch address semantic deviations and centralized spatial mapping drift), thereby optimizing the efficiency of verification resource allocation and strengthening the foundation of address modeling data.

[0015] Modeling decision verification judgment: If the judgment is to proceed, the result of the time-varying association verification between address entities and data will be used to determine whether to trigger modeling decision verification, so as to measure the qualification of modeling based on multi-source heterogeneous address data. By conducting modeling decision verification judgment, it is helpful to quantify the compliance of modeling results in dimensions such as recognition accuracy, semantic integrity, and spatial consistency, and ensure that the modeling results meet the accuracy requirements of smart city address management.

[0016] Time-series correction: If determined not to proceed, time-series correction of address data will be performed based on the analysis results of address information addition and text address update. This corrects spatiotemporal misalignments and fills in time-series gaps. After the time-series correction is completed, time-varying association verification between address entities and data will be performed. By performing time-series correction, spatiotemporal misalignment of address data can be eliminated (achieving accurate matching of address text descriptions, latitude and longitude coordinates, and timestamps), significantly improving the consistency and standardization of multi-source heterogeneous address data, and enhancing the dynamic adaptability of address data to urban development.

[0017] Figure 2 shows an overall overview of the intelligent modeling method for multi-source heterogeneous address data in smart cities provided by an embodiment of this invention. As shown in Figure 2: by analyzing the addition of address information and the updating of text addresses, abnormal values ​​for address addition and text address updates are obtained. When the abnormal value for address addition is greater than a preset abnormal value, address data timing correction is performed based on the corresponding abnormal value; otherwise, time-varying association verification between address entities and data is performed. When the abnormal value for text address updates is greater than a predefined abnormal value, address data timing correction is performed based on the corresponding abnormal value; otherwise, time-varying association verification between address entities and data is performed. A mapping drift value is obtained by performing time-varying association verification between address entities and data. When the mapping drift value is greater than a predefined qualified mapping drift value, the index grid is reduced; otherwise, the average retrieval rate of multi-source heterogeneous address data is monitored to see if it is greater than a predefined maximum multi-source data retrieval rate. If not, modeling decision verification is performed; otherwise, the index grid is increased.

[0018] It should be added that a database storing various settings data was established before the design of the intelligent modeling method for multi-source heterogeneous address data for smart cities provided in this application. The database includes, but is not limited to, preset address addition anomaly judgment values, predefined text address update anomaly judgment values, etc. The various values ​​are directly set by technical personnel. The technical personnel will comprehensively consider the actual scenario of smart cities, data characteristics and business needs, and determine the most suitable values ​​through a large number of experiments and data analysis to ensure that the settings data in the database can accurately and effectively support the operation of the intelligent modeling method for multi-source heterogeneous address data.

[0019] In this embodiment, the interconnectedness of time-varying correlation determination, modeling decision verification determination, and time-series correction helps to improve the accuracy of smart city address modeling, reduce the error rate of address entity recognition and spatial mapping error, quickly respond to the time-varying characteristics of address data, reduce the interference of abnormal data on modeling results, ensure the long-term controllability of modeling effects, and thus provide a data foundation for the stable operation and in-depth application of smart city geographic information.

[0020] For example, in provinces with a high concentration of ethnic minorities, a collaborative mechanism is formed through the interaction of time-varying correlation determination, modeling decision verification determination, and time-series correction, from dynamic identification to precise correction, and finally to a quality closed loop. Time-varying correlation determination accurately captures the time-varying risks unique to ethnic scenarios, time-series correction addresses the spatiotemporal misalignment and discontinuity of cross-source data, and modeling decision verification determination quantitatively ensures that data is adapted to governance needs, thereby improving the accuracy, timeliness, collaboration, security, and ethnic adaptability of governance.

[0021] Further, the specific process for analyzing the addition of address information and the update of text addresses is as follows: Monitor the addition of address information to assess the synchronization between the new address information and the GPS trajectory. Specifically, monitor and count the time between the end time of each new address information entry and the moment the captured GPS trajectory contains the new address information. Use the average value as the address addition anomaly judgment value. Determine if the address addition anomaly judgment value is greater than a preset value. If so, mark the corresponding address addition anomaly judgment value as a feedback condition for address data time-series correction; otherwise, perform address entity and data time-varying association verification. The preset address addition anomaly judgment value is determined by analyzing address addition anomalies over historical time periods. The average value of the judgment value is used to monitor the update of text addresses to assess the abnormality of the synchronization between text addresses and GPS track points. Specifically, a counter is used to monitor and count the proportion of addresses corresponding to GPS track points that do not contain newly added address information within a preset effective time window to the total number of addresses of GPS track points, thus obtaining the text address update anomaly judgment value. If the text address update anomaly judgment value is greater than the predefined text address update anomaly judgment value, the corresponding text address update anomaly judgment value is marked as a feedback condition for address data time sequence correction. Otherwise, a time-varying association verification of address entities and data is performed. The predefined text address update anomaly judgment value is represented by the average value of text address update anomaly judgment values ​​over a historical time period.

[0022] The analysis includes address information additions and text address updates, as well as feedback conditions for address data timing correction based on the output. Address data timing correction is selectively performed during the next preset monitoring period. The specific implementation process for selective address data timing correction is as follows: Address addition anomaly judgment values ​​and text address update anomaly judgment values ​​are input into a preset address anomaly association table. Address addition anomaly scores and text address update anomaly scores, representing the degree of impact of address information additions and text address updates, are read. These scores are used as judgment conditions for address data timing correction, and the processing mode for multi-source heterogeneous address data is divided as follows: When the number of newly added addresses in a preset area exceeds the predefined maximum number of addresses in the area, address data timing correction is performed based on stream processing; otherwise, address data timing correction is performed based on batch processing.

[0023] It needs to be specifically explained that, in the implementation scheme of this application, the constructed database includes initially created tables or sets with mapping functions, supporting precise one-to-one mapping between single parameters, as well as many-to-one mapping from multiple parameters to a single parameter. The specific construction process is as follows: Pre-set personnel will collect various key smart city management information within a specific historical time period, including combinations of address addition anomaly judgment values ​​and text address update anomaly judgment values, combinations of address addition anomaly severity scores and the total number of update requests corresponding to multi-source heterogeneous address data, combinations of text address update anomaly judgment values ​​and the data volume of multi-source heterogeneous address data, combinations of address addition anomaly severity scores, text address update anomaly judgment values, the data volume of multi-source heterogeneous address data and the total number of update requests corresponding to multi-source heterogeneous address data, combinations of standardized result deviation values, mapping drift values ​​and the sum of storage rates of multi-source heterogeneous address data, mapping drift values, combinations of storage rates of multi-source heterogeneous address data and the average retrieval rates of multi-source heterogeneous address data, etc.; The collected historical key information is input into a machine learning model with the ability to reveal the importance of features. The decision tree model is a typical example of such a model; the decision tree model, with its unique feature splitting mechanism, according to... According to pre-defined mapping rules, the input information is deeply analyzed. During model operation, it automatically selects the features that have the most significant impact on the results and performs data splitting and classification based on these features. The model accurately extracts results from the input information that can be used as weight values ​​or other related data, such as the combination of address addition anomaly score and text address update anomaly score, the combination of spatial search radius adjustment step size and GPS sampling interval adjustment step size, address activation time adjustment step size, spatial search radius pass value, GPS sampling interval pass value and address activation time pass value, pass standardization similarity threshold, parameter combination corresponding to index grid decrease step size and index grid decrease times, and parameter combination corresponding to index grid increase step size and index grid increase times. The raw data collected within a specific historical time period is matched with the corresponding weights or related data output by the model to ensure that each raw data can find a corresponding weight or related data. Finally, it generates a pre-defined address anomaly association table, a pre-built new address anomaly adjustment table, a pre-defined text address anomaly adjustment table, a pre-defined address data time series adjustment table, a pre-defined address modeling standardization association table, an index grid decrease guidance table, and an index grid increase guidance table.

[0024] During the real-time monitoring phase, if information is detected such as combinations of address addition anomaly judgment values ​​and text address update anomaly judgment values, combinations of address addition anomaly severity scores and the total number of update requests corresponding to multi-source heterogeneous address data, combinations of text address update anomaly judgment values ​​and the data volume of multi-source heterogeneous address data, combinations of address addition anomaly severity scores, text address update anomaly judgment values, the data volume of multi-source heterogeneous address data and the total number of update requests corresponding to multi-source heterogeneous address data, standardized result deviation values, mapping drift values ​​and the storage rate of multi-source heterogeneous address data, or combinations of mapping drift values, the storage rate of multi-source heterogeneous address data and the average retrieval rate of multi-source heterogeneous address data, then this type of real-time information only needs to be input into the pre-built pre-configured database. The system can directly retrieve results from tables such as the address anomaly association table, the pre-built new address anomaly adjustment table, the preset text address anomaly adjustment table, the preset address data time sequence adjustment table, the predefined address modeling standardization association table, the index grid reduction guidance table, and the index grid increase guidance table. These results include combinations of new address anomaly scores and text address update anomaly scores, combinations of spatial search radius adjustment step size and GPS sampling interval adjustment step size, address activation time adjustment step size, spatial search radius pass value, GPS sampling interval pass value and address activation time pass value, pass standardization similarity threshold, parameter combinations corresponding to index grid reduction step size and index grid reduction times, and parameter combinations corresponding to index grid increase step size and index grid increase times.

[0025] Figure 3 shows a schematic diagram of address data timing correction in the intelligent modeling method for multi-source heterogeneous address data for smart cities provided in an embodiment of this invention. As shown in Figure 3, the specific correction rules for address data timing are as follows: Rule 1: When the score for the degree of anomaly in newly added addresses is greater than the score for the degree of anomaly in text address updates, a first timing correction measure for address data is implemented, adjusting the spatial search radius and GPS sampling interval. This first timing correction measure improves the spatiotemporal continuity and detail reproduction of multi-source heterogeneous address data. Rule 2: When the score for the degree of anomaly in newly added addresses is less than the score for the degree of anomaly in text address updates, a second timing correction measure for address data is implemented, adjusting the address activation time. This second timing correction measure improves the temporal fault tolerance of address matching, thereby reducing the timing misalignment of multi-source heterogeneous address data. Rule 3: When the score for the degree of anomaly in newly added addresses is equal to the score for the degree of anomaly in text address updates, a third timing correction measure for address data is implemented, adjusting the spatial search radius, GPS sampling interval, and address activation time. This third timing correction measure ensures the temporal detail of address trajectories while being compatible with the timing misalignment of multi-source heterogeneous address data, thereby reducing the matching error of multi-source heterogeneous address data.

[0026] By selectively correcting the timing of address data, the system obtains scores for the severity of address addition anomalies and text address update anomalies. When the address addition anomaly score is greater than the text address update anomaly score, a first timing correction measure is implemented. When the address addition anomaly score is less than the text address update anomaly score, a second timing correction measure is implemented. When the address addition anomaly score is equal to the text address update anomaly score, a third timing correction measure is implemented. In other words, the first timing correction measure is implemented when address addition anomalies are more severe, and the second timing correction measure is implemented when text address update anomalies are more severe. When the anomalies are of the same degree, the third time-series correction measure for address data is executed. This helps to accurately match the dominant anomaly type with the correction strategy, improve the targeting of the correction, ensure the semantic integrity of the address data after time-series correction, take into account the alignment of newly added data with updated data, and enhance the adaptability of correction under different time-varying scenarios (covering three types of scenarios: newly added dominant, updated dominant, and compound anomalies). This, in turn, improves the standardization level of address data and the accuracy of spatial mapping. Through the interrelation between the first, second, and third time-series correction measures for address data, it helps to achieve synergistic effects of the correction measures and avoid the limitations of a single measure.

[0027] In this embodiment, by analyzing the addition of address information and the update of text addresses, anomaly judgment values ​​for address additions and text updates that are greater than preset anomaly judgment values ​​are obtained. These values ​​are then used as feedback conditions for address data time-series correction (i.e., address data time-series correction is only performed when both feedback conditions are detected simultaneously; if only one anomaly judgment value or one text update judgment value is detected, it is considered qualified, and time-varying association verification between address entities and data is performed). This helps to accurately screen high-risk time-varying scenarios, avoids over-correction of low-risk single anomalies, reduces redundant processes of invalid verification and correction, thereby improving overall modeling efficiency, ensuring the targeted nature of single anomaly data verification, and avoiding excessive intervention of time-series correction on qualified data. By analyzing the close correlation between address information additions, text address updates, and address data time-series correction, it helps to dynamically capture the complex time-varying patterns of address data, strengthen the time-series consistency of multi-source heterogeneous address data, and achieve semantic and spatial synchronization between new and updated data. This provides high-quality time-series data support for modeling decisions, reduces interference from anomaly data on modeling results, and thus improves the dynamic adaptability of address data.

[0028] Furthermore, the specific execution steps of the first time-series correction measure for address data are as follows: The obtained spatial search radius adjustment step size is set as the step size for the gradual increase of the spatial search radius; the obtained GPS sampling interval adjustment step size is set as the step size for the gradual decrease of the GPS sampling interval; the gradual increase of the spatial search radius and the gradual decrease of the GPS sampling interval are executed simultaneously; after each round of operation, it is determined whether the newly obtained address anomaly judgment value meets the set conditions, i.e., the newly obtained address anomaly judgment value is greater than the preset address anomaly judgment value, and greater than the address anomaly judgment value in the initial state when the first time-series correction measure for address data is executed; if this condition is met, a new address adjustment alarm is sent, and a prompt is sent to the preset personnel to return to the initial GPS sampling interval and spatial search radius; otherwise, address entity and data time-varying association verification is performed; the combination of the spatial search radius adjustment step size and the GPS sampling interval adjustment step size is obtained by inputting the address anomaly severity score and the total number of update requests corresponding to the multi-source heterogeneous address data monitored by the counter into the pre-constructed new address anomaly adjustment table.

[0029] In this embodiment, by implementing a first time-series correction measure for address data, the spatial search radius adjustment step size and the GPS sampling interval adjustment step size are obtained and used as the operations of gradually increasing the spatial search radius and gradually decreasing the GPS sampling interval, respectively. Executing these corresponding operations helps improve the spatial positioning accuracy of newly added addresses, reduces spatial mapping drift errors, improves the spatial matching coverage of multi-source heterogeneous address data, helps avoid matching failures caused by insufficient search radius of newly added addresses, and achieves spatial coordinate synchronization between newly added addresses and the existing address database. By simultaneously adjusting the spatial search radius and GPS sampling interval, the spatiotemporal integrity of address data and the accuracy of newly added anomaly identification are improved, while balancing data acquisition efficiency and system load, providing high-quality location reference data for address data time-series correction and dynamic permission allocation. After each round of operations, it is determined whether the newly acquired address anomaly judgment value meets the set conditions. Only if it does not meet the conditions is the address entity and data time-varying association verification performed. This helps to dynamically control the iterative effect of time-series correction, ensuring that each round of adjustment moves towards anomaly mitigation, avoiding invalid verification under insufficient correction conditions, and thus improving the accuracy and effectiveness of time-varying association verification.

[0030] Furthermore, the specific execution steps of the second address data timing correction measure are as follows: The text address update anomaly judgment value and the amount of multi-source heterogeneous address data monitored by the network traffic monitor are used as input parameters for a preset text address anomaly adjustment table; the address activation time adjustment step size output by the preset text address anomaly adjustment table is read; the address activation time adjustment step size is configured as the adjustment amount for executing the address activation time increase operation; the text address update anomaly judgment values ​​of the initial and final states when the address activation time increase operation is executed for the first time are statistically analyzed; the difference between the corresponding initial and final text address update anomaly judgment values ​​is determined to be within the range specified in the preset text address anomaly adjustment table. If the address update is within the range of qualified addresses set in advance by the personnel, the address activation time increase operation continues; otherwise, an activation anomaly prompt is sent, and the preset personnel are fed back the address activation time. After each round of address activation time increase operation, if the newly acquired text address update anomaly judgment value is greater than the predefined text address update anomaly judgment value, and the text address update anomaly judgment value is greater than the text address update anomaly judgment value at the initial state when the second correction measure for address data timing is executed, a text address anomaly adjustment alarm prompt is sent, and the preset personnel are fed back the initial address activation time; otherwise, address entity and data time-varying association verification is performed.

[0031] In this embodiment, the address activation time adjustment step size is obtained by performing a second time-series correction measure on address data, and is used as the adjustment amount for increasing the address activation time (i.e., the duration corresponding to the time-series alignment window). This helps improve the semantic synchronization efficiency after text address updates, ensures that the updated address attributes are fully effective, and improves the adaptability of address activation time to the rhythm of urban address changes (matching update cycles such as road renaming and administrative division adjustments). This helps improve the effective identification coverage of updated addresses, avoids updated addresses not being included in the modeling due to insufficient activation time, and thus ensures the semantic consistency between updated addresses and the standard address database and improves the synchronous response capability of multi-source data to address updates. By judging whether the difference between the text address update anomaly judgment value in the initial and final states of the first execution of the address activation time increase operation is within the qualified address update range, the address activation time increase operation is continued only if it is within the range. This helps to accurately control the adjustment range of address activation time (avoiding semantic redundancy or data conflicts caused by excessive increase), thereby ensuring the semantic integrity and timeliness of text addresses and strengthening the adaptability of address data to urban dynamic updates.

[0032] Furthermore, the specific implementation steps of the third correction measure for address data timing are as follows: The address addition anomaly score, text address update anomaly judgment value, the amount of multi-source heterogeneous address data, and the total number of update requests corresponding to the multi-source heterogeneous address data are used as input parameters for a preset address data timing adjustment table. The qualified values ​​for spatial search radius, GPS sampling interval, and address activation time corresponding to the output multi-source heterogeneous address data are read. Configuration prompts are sent to preset personnel to configure the spatial search radius, GPS sampling interval, and address activation time corresponding to the multi-source heterogeneous address data to their respective qualified values, and after configuration, it is determined whether the requirements are met. If the parameter anomaly rollback condition is met, a timing correction alarm is sent, and feedback is provided to the preset personnel to return to the initial GPS sampling interval, spatial search radius, and address activation time. Otherwise, a time-varying association verification of address entities and data is performed. The parameter anomaly rollback condition means that the newly acquired address addition anomaly judgment value and text address update anomaly judgment value are greater than the corresponding preset address addition anomaly judgment value and predefined text address update anomaly judgment value, respectively, and the address addition anomaly judgment value and text address update anomaly judgment value are greater than the address addition anomaly judgment value and text address update anomaly judgment value of the initial state corresponding to the execution of the third time-series correction measure for address data.

[0033] In this embodiment, by performing a third correction measure for address data timing, qualified values ​​for spatial search radius, GPS sampling interval, and address activation time are obtained. The spatial search radius, GPS sampling interval, and address activation time corresponding to multi-source heterogeneous address data are configured to the corresponding qualified values, which helps to improve the accuracy of multi-parameter collaborative adaptation (taking into account the comprehensive correction needs of addition and update anomalies) and improve the consistency of address data timing characteristics, thereby achieving synchronous matching of spatial positioning and time activation. By judging whether the parameter anomaly rollback conditions are met after configuration, and only performing address entity and data time-varying association verification if they are not met, it helps to achieve comprehensive quality improvement of multi-source heterogeneous address data and improve the accuracy of address entity recognition.

[0034] Furthermore, the time-varying association verification between address entities and data represents an evaluation of the time-varying association between address entities and multimodal data based on three verification levels. Verification level one monitors address entity identification errors, specifically as follows: the ratio of the difference between the total number of labeled entities monitored by the counter and the number of correctly identified address entities to the total number of labeled entities is marked as the address identification error judgment value, reflecting the degree of address entity identification error. A judgment is initiated based on the address identification error judgment value. When the address identification error judgment value is greater than a predefined address identification error judgment value, a multi-source heterogeneous address data re-acquisition prompt is sent; otherwise, monitoring continues to monitor the deviation of the standardized results corresponding to the multi-source heterogeneous address data from semantics. The predefined address identification error judgment value is represented by the average of address identification error judgment values ​​over historical time periods. Verification level two monitors the deviation of the standardized results corresponding to the multi-source heterogeneous address data from semantics. Specifically, the process involves: monitoring the number of standardized results that differ from the preset standardized results after standardization processing based on multi-source heterogeneous address data using a counter, and determining the proportion of these results to the total number of standardized results. This yields a standardization result deviation value, which is then used for judgment. The system checks if the standardization result deviation value exceeds a predefined value. If so, address similarity threshold adjustment is performed in the next preset monitoring period; otherwise, monitoring continues to track the spatial mapping position drift corresponding to the multi-source heterogeneous address data. The predefined standardization result deviation value is represented by the average of standardization result deviation values ​​over historical time periods. This deviation value reflects the degree of semantic deviation of the standardization results corresponding to the multi-source heterogeneous address data. Address similarity threshold adjustment is used to adapt to the characteristics of multi-source heterogeneous address data, filter low-quality address association results, and ensure the accuracy of address mapping.

[0035] Specifically, the process of address similarity threshold adjustment is as follows: First, a standardized similarity threshold is set, specifically: based on the qualified standardized similarity threshold corresponding to the acquired multi-source heterogeneous address data (i.e., the qualified standardized similarity threshold during standardization processing), a target value for address similarity threshold adjustment is set. Second, anomaly judgment is performed on address similarity threshold adjustment, specifically: when the standardized similarity threshold corresponding to the multi-source heterogeneous address data is adjusted to the qualified standardized similarity threshold, it is determined whether the deviation value of the re-acquired standardized result is greater than the predefined deviation value, and whether the deviation value of the re-acquired standardized result is greater than the deviation value of the standardized result in the initial state of address similarity threshold adjustment. Third, anomaly feedback is provided on address similarity threshold adjustment, specifically: if the deviation value of the re-acquired standardized result is greater than the predefined deviation value, or greater than the deviation value of the standardized result in the initial state of address similarity threshold adjustment, an address similarity threshold adjustment alarm is sent, and feedback is provided to the preset personnel to return to the initial standardized similarity threshold; otherwise, the spatial mapping position drift corresponding to the multi-source heterogeneous address data continues to be monitored. The qualified standardized similarity threshold is obtained by inputting the deviation value of the standardized result into a predefined address modeling standardized association table.

[0036] By adjusting the address similarity threshold when the deviation value of the standardized result exceeds the predefined deviation value, and setting the target value corresponding to the address similarity threshold adjustment based on the qualified standardized similarity threshold of the acquired multi-source heterogeneous address data, it helps to improve the semantic fit between the standardized result and the original address, reduce semantic loss caused by over-standardization, narrow the standardization difference of multi-source data, and thus achieve the semantic integrity of the standardized result and the semantic consistency of multi-source heterogeneous address data.

[0037] Address entity and data time-varying association verification also includes verification level three, monitoring the spatial mapping location drift corresponding to multi-source heterogeneous address data. Specifically, it involves: monitoring the number of spatial points with failed geocoding in the drift time series using a counter to obtain a mapping drift value, which reflects the degree of drift of the spatial mapping location corresponding to multi-source heterogeneous address data; classifying the mapping drift risk level based on the mapping drift value, with the following specific process: Operation one, when the mapping drift value is greater than the predefined qualified mapping drift value, the index grid is reduced in the next preset monitoring time period to improve the granularity of the spatial index and narrow the spatial range of address retrieval, thereby correcting the mapping drift deviation. In this context, the predefined acceptable mapping drift value is represented by the average mapping drift value over a historical time period. The index grid reduction adjustment involves using a set index grid reduction step size as the adjustment amount and a set number of index grid reductions as the adjustment count, progressively reducing the index grid corresponding to multi-source heterogeneous address data. The index grid reduction adjustment also includes re-monitoring the mapping drift value at the end of the adjustment process. If the mapping drift value is still greater than the predefined acceptable mapping drift value, an index grid reduction anomaly alarm is sent; otherwise, modeling decision verification is performed. The parameter combination corresponding to the index grid reduction step size and the number of index grid reductions is achieved by combining the mapping drift value with a performance monitor, such as Performance Monitor. Monito monitors the storage rate of multi-source heterogeneous address data and inputs it into the index grid reduction guide table. Operation two involves monitoring whether the average retrieval rate of multi-source heterogeneous address data within a preset time period, monitored by retrieval tools such as Wireshark, exceeds the preset maximum multi-source data retrieval rate. If not, modeling decision verification is performed; otherwise, index grid enlargement is adjusted in the next preset monitoring time period to reduce the computational complexity of the spatial index, thereby improving the retrieval and processing efficiency of multi-source heterogeneous address data. Index grid enlargement adjustment means using a set index grid enlargement step size as the adjustment amount, and using a set index grid enlargement... The index grid corresponding to the multi-source heterogeneous address data is increased step by step using a large number of adjustment steps. The adjustment of index grid increase also includes re-monitoring the mapping drift value, the storage rate of the multi-source heterogeneous address data, and the average retrieval rate of the multi-source heterogeneous address data at the end of the adjustment. If the average retrieval rate of the multi-source heterogeneous address data is not greater than the predefined maximum multi-source data retrieval rate and the mapping drift value is not greater than the predefined qualified mapping drift value, modeling decision verification is performed, that is, modeling is performed based on the corresponding multi-source heterogeneous address data. Otherwise, an index grid increase anomaly alarm is sent. The parameter combination corresponding to the index grid increase step size and the number of index grid increases is obtained by inputting the mapping drift value and the storage rate of the multi-source heterogeneous address data into the index grid increase guidance table.

[0038] It should be added that when applied to provinces with a high concentration of ethnic minorities, the dataset consisting of multi-source heterogeneous address data (such as administrative division addresses, urban planning addresses, highway addresses, etc.) and the predicted results of pre-defined association types (such as spatial topological relationships, attribute relationships, temporal dependencies, etc.) is randomly divided into a training set. The training set is then input into a multimodal deep learning model, such as HGNN (Heterogeneous Graph Neural Network), for training. This model learns the semantic association rules of cross-source address data in multi-ethnic scenarios (including standardized mapping of place names in multi-ethnic languages, and the adaptation logic of ethnic characteristic vocabulary and address attributes), as well as the spatial topological association patterns unique to ethnic minority areas (such as the spatial dependency between addresses within an area and surrounding infrastructure, and the connectivity between addresses across ethnic minority areas). The trained model is then used as a cross-source data association model for provinces with a high concentration of ethnic minorities. By inputting newly collected multi-source heterogeneous address data and spatiotemporal features into the cross-source data association model, the predicted results of association types can be output.

[0039] Figure 4 illustrates a multimodal deep learning diagram of the intelligent modeling method for multi-source heterogeneous address data in smart cities provided by an embodiment of this invention. As shown in Figure 4, address neighbor sampling selects associated neighbor addresses of the target address from a heterogeneous graph containing multiple address types (such as administrative divisions, urban planning addresses, and highway addresses) to construct a local subgraph. Address type-aware aggregation classifies neighbor addresses according to address type (such as spatial attribute type and administrative attribute type), and then performs feature processing based on these types. Address feature extraction encodes address information from different dimensions. NN-1 performs preliminary fusion of features for single-type addresses. NN-2 aggregates features of neighbor addresses of the same type, and NN-3 fuses address features of different types to output the association relationships between addresses. Loss is predicted based on address association relationships, ultimately yielding the association relationship type (spatial topological relationship, attribute association relationship, temporal dependency relationship, etc.).

[0040] By monitoring the spatial mapping position drift of multi-source heterogeneous address data to obtain mapping drift values, and when the mapping drift value exceeds a predefined acceptable mapping drift value, the index grid is gradually reduced by a set step size and a set number of index grid reductions, thus improving the resolution and accuracy of spatial positioning (by narrowing the grid range to focus on real geographic coordinates), reducing matching errors caused by drift, and thus improving the spatial matching accuracy of multi-source address data, ultimately achieving precise alignment between address data and real geographic locations. Furthermore, when the mapping drift value is not greater than a predefined acceptable mapping drift value, the average retrieval rate of multi-source heterogeneous address data continues to be monitored. When the average retrieval rate exceeds a predefined maximum multi-source data retrieval rate, the index grid is gradually increased by a set step size and a set number of index grid increases, thus improving the retrieval efficiency of multi-source heterogeneous address data (by expanding the grid range to reduce retrieval complexity), thereby enhancing the overall operational efficiency of the modeling process and the ability to access multi-source data.

[0041] Figure 5 shows a schematic diagram of the structure of the intelligent modeling system for multi-source heterogeneous address data for smart cities provided in an embodiment of this invention. The intelligent modeling system for multi-source heterogeneous address data for smart cities is used to implement a method for intelligent modeling of multi-source heterogeneous address data for smart cities. The system includes: a time-varying correlation determination module, a modeling decision verification determination module, and a time-series correction module. The time-varying correlation determination module is used to analyze the addition of address information and the updating of text addresses based on multi-source heterogeneous address data reflecting the geographic location characteristics of urban resources in a specified smart city address modeling scenario, and determines whether to perform a time-series correction based on the analysis results. The address entity and data time-varying association verification is used to determine the address entity identification errors, the deviation of the standardization results from the semantics, and the spatial mapping position drift of the multi-source heterogeneous address data. The modeling decision verification judgment module is used to determine whether to trigger the modeling decision verification based on the result of the address entity and data time-varying association verification if the determination is not to proceed, so as to measure the qualification of modeling based on multi-source heterogeneous address data. The time sequence correction module is used to perform address data time sequence correction based on the analysis results of address information addition and text address update if the determination is not to proceed, so as to correct the spatiotemporal misalignment and fill the time sequence gap. After the address data time sequence correction is completed, the address entity and data time-varying association verification is performed.

[0042] In this embodiment, by performing time-varying association verification between address entities and data, and sequentially monitoring address entity identification errors, deviations of standardized results from semantics in multi-source heterogeneous address data, and spatial mapping position drift in multi-source heterogeneous address data, it helps to systematically identify core quality risk points throughout the entire address modeling process (covering the three key dimensions of identification, semantics, and space without omission), accurately locate the root causes of different types of data anomalies (distinguishing between problems caused by defective identification algorithms, insufficient standardization rules, or spatial mapping deviations), provide clear targeted basis for anomaly handling, intercept unqualified data from propagating downstream of modeling in advance, strengthen the closed-loop quality control of multi-source heterogeneous address data, and thereby improve the accuracy of address entity identification, semantic consistency of standardized results, accuracy of spatial mapping, and traceability of data quality.

[0043] By monitoring address entity recognition errors to obtain address recognition error judgment values, and when the address recognition error judgment value is not greater than a predefined address recognition error judgment value, the standardization results corresponding to multi-source heterogeneous address data continue to be monitored for semantic deviations, and the degree of deviation of the standardization results is obtained. This helps to improve the efficiency of the step-by-step control of the inspection process, avoid the ineffective investment of semantic verification in identifying erroneous data, reduce ineffective semantic correction operations, and reduce the complexity of correcting the superposition of recognition errors and semantic deviations. Only when the degree of deviation of the standardization results is not greater than a predefined degree of deviation of the standardization results does the spatial mapping position drift corresponding to multi-source heterogeneous address data continue to be monitored. This helps to ensure that the data entering spatial verification meets both recognition and semantic requirements, eliminates the interference of anomalies in previous dimensions, and ensures the accuracy of spatial mapping verification.

[0044] In summary, the embodiments of this invention analyze the addition of address information and the updating of text addresses using multi-source heterogeneous address data reflecting the geographic location characteristics of urban resources. Based on the analysis results, it determines whether to perform time-varying association verification between address entities and data. This helps to identify address entity recognition errors, deviations from semantics in standardization results, and spatial mapping location drift in the multi-source heterogeneous address data. This facilitates the accurate capture of time-varying features of multi-source address data, the pre-screening of potential quality defects, and the differentiated triggering of association verification. If it is determined that time-varying association verification should be performed, it is then determined whether to trigger modeling decisions based on the results of the time-varying association verification between address entities and data. Policy verification helps to achieve dynamic quality control of the modeling process, quantitative evaluation of the credibility of modeling results, and rapid tracing and correction of unqualified modeling data. If it is determined not to proceed, address data time-series correction is performed based on the analysis results of address information addition and text address update. After the address data time-series correction is completed, address entity and data time-varying association verification is performed. This helps to achieve closed-loop repair of defects in multi-source heterogeneous address data, completeness and comprehensiveness of association verification, and improvement of the temporal continuity and spatiotemporal consistency of subsequent modeling data. It also helps to solve the problem of low accuracy in processing multi-source heterogeneous address data in existing technologies.

[0045] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0046] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device, equipment, and storage medium embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0047] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0048] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent modeling of multi-source heterogeneous address data for smart cities, characterized in that, The method includes: analyzing the addition of address information and the updating of text addresses based on multi-source heterogeneous address data reflecting the geographic positioning characteristics of urban resources; determining whether to perform time-varying association verification between address entities and data based on the analysis results, in order to identify address entity identification errors, deviations of standardization results from semantics, and spatial mapping location drift in the multi-source heterogeneous address data; if determined to perform, determining whether to trigger modeling decision verification based on the results of the time-varying association verification between address entities and data; if determined not to perform, performing time-series correction of address data based on the analysis results of the addition of address information and the updating of text addresses, and performing time-varying association verification between address entities and data after the time-series correction is completed.

2. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 1, characterized in that, The specific process for analyzing the addition of address information and the update of text address is as follows: Monitoring the addition of address information involves calculating the time between the end time of each new address information entry and the time when the captured GPS trajectory contains the new address information, using the average value as the address addition anomaly judgment value; determining whether the address addition anomaly judgment value is greater than a preset address addition anomaly judgment value; if so, marking the corresponding address addition anomaly judgment value as a feedback condition for address data time-series correction; otherwise, performing address entity and data time-varying association verification; Monitoring text address updates involves calculating the proportion of addresses corresponding to GPS trajectory points that do not contain new address information within a preset effective time window to the total number of addresses of GPS trajectory points, obtaining the text address update anomaly judgment value; if the text address update anomaly judgment value is greater than a predefined text address update anomaly judgment value, marking the corresponding text address update anomaly judgment value as a feedback condition for address data time-series correction; otherwise, performing address entity and data time-varying association verification.

3. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 2, characterized in that, The analysis of address information additions and text address updates also includes feedback conditions based on the output address data time-series correction, selectively performing address data time-series correction in the next preset monitoring period. The specific implementation process of selectively performing address data time-series correction is as follows: Input the address addition anomaly judgment value and the text address update anomaly judgment value into a preset address anomaly association table, and read the address addition anomaly degree score and the text address update anomaly degree score, which represent the degree of influence of address information additions and text address updates respectively; use the address addition anomaly degree score and the text address update anomaly degree score as the judgment conditions for address data time-series correction, and perform multi-source heterogeneous address data processing mode division, specifically as follows: When a pre- When the number of newly added addresses in a region exceeds the predefined maximum number of addresses in the region, address data timing correction is performed based on stream processing; otherwise, it is performed based on batch processing. The address data timing correction is specifically as follows: when the address addition anomaly score is greater than the text address update anomaly score, a first address data timing correction measure is implemented to correct the spatial search radius and GPS sampling interval; when the address addition anomaly score is less than the text address update anomaly score, a second address data timing correction measure is implemented to correct the address activation time; when the address addition anomaly score is equal to the text address update anomaly score, a third address data timing correction measure is implemented to correct the spatial search radius, GPS sampling interval, and address activation time.

4. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 3, characterized in that, The specific execution steps of the first address data timing correction measure are as follows: The obtained spatial search radius adjustment step size is set as the step size for the gradual increase of the spatial search radius operation; the obtained GPS sampling interval adjustment step size is set as the step size for the gradual decrease of the GPS sampling interval operation; the gradual increase of the spatial search radius operation and the gradual decrease of the GPS sampling interval operation are executed simultaneously; after each round of operation, it is determined whether the newly obtained address anomaly judgment value meets the set conditions, i.e., the newly obtained address anomaly judgment value is greater than the preset address anomaly judgment value, and greater than the address anomaly judgment value in the initial state when the first address data timing correction measure is executed; if satisfied, a new address adjustment alarm is sent, and the initial GPS sampling interval and spatial search radius are returned; otherwise, address entity and data time-varying association verification is performed; the combination of the spatial search radius adjustment step size and the GPS sampling interval adjustment step size is obtained by inputting the address anomaly degree score and the total number of update requests corresponding to multi-source heterogeneous address data into a pre-constructed new address anomaly adjustment table.

5. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 3, characterized in that, The specific execution steps of the second address data timing correction measure are as follows: The text address update anomaly judgment value and the amount of multi-source heterogeneous address data are used as input parameters of a preset text address anomaly adjustment table; the address activation time adjustment step size output by the preset text address anomaly adjustment table is read; the address activation time adjustment step size is configured as the adjustment amount for executing the address activation time increase operation; the text address update anomaly judgment values ​​of the initial and final states when the address activation time increase operation is executed for the first time are statistically analyzed; it is determined whether the difference between the corresponding initial and final state text address update anomaly judgment values ​​is within the qualified address update range; if so, the address activation time increase operation continues; otherwise, an activation anomaly prompt is sent, and a preset personnel return address activation time is fed back. After each round of address activation time increase operation is completed, if the newly acquired text address update anomaly judgment value is greater than the predefined text address update anomaly judgment value, and the text address update anomaly judgment value is greater than the text address update anomaly judgment value of the initial state when the second correction measure of address data timing is executed, a text address anomaly adjustment alarm is sent and the initial address activation time is returned; otherwise, the address entity and data time-varying association verification is performed.

6. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 3, characterized in that, The specific execution steps of the third correction measure for address data timing are as follows: take the address addition anomaly score, text address update anomaly judgment value, data volume of multi-source heterogeneous address data and the total number of update requests corresponding to multi-source heterogeneous address data as input parameters of the preset address data timing adjustment table, and read the qualified value of spatial search radius, qualified value of GPS sampling interval and qualified value of address activation time corresponding to the output multi-source heterogeneous address data. A configuration prompt is sent to preset personnel to configure the spatial search radius, GPS sampling interval, and address activation time corresponding to the multi-source heterogeneous address data to the corresponding qualified values. After configuration, it is determined whether the parameter anomaly rollback condition is met. If it is met, a timing correction alarm prompt is sent, and the initial GPS sampling interval, spatial search radius, and address activation time are returned. Otherwise, a time-varying association verification between address entities and data is performed. The parameter anomaly rollback condition means that the re-acquired address addition anomaly judgment value and text address update anomaly judgment value are greater than the corresponding preset address addition anomaly judgment value and predefined text address update anomaly judgment value, respectively, and the address addition anomaly judgment value and text address update anomaly judgment value are greater than the address addition anomaly judgment value and text address update anomaly judgment value of the initial state corresponding to the execution of the third time-series correction measure for address data.

7. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 6, characterized in that, The time-varying association verification of address entities and data is carried out as follows: The ratio of the difference between the total number of labeled entities and the number of correctly identified address entities to the total number of labeled entities is marked as the address identification error judgment value, which is used to reflect the degree of address entity identification error. The judgment is initiated based on the address recognition error judgment value. When the address recognition error judgment value is greater than the predefined address recognition error judgment value, a prompt for re-collection of multi-source heterogeneous address data is sent. Otherwise, the monitoring of the deviation of the standardized results corresponding to the multi-source heterogeneous address data from the semantics continues. After standardization processing based on multi-source heterogeneous address data, the proportion of the number of standardized results that are inconsistent with the preset standardized results to the total number of standardized processing results is used to obtain the standardization result deviation value and perform judgment. Determine whether the deviation value of the standardization result is greater than the predefined deviation value of the standardization result. If so, adjust the address similarity threshold in the next preset monitoring period. Otherwise, continue to monitor the spatial mapping position drift of the multi-source heterogeneous address data.

8. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 7, characterized in that, The specific process of address similarity threshold adjustment is as follows: Standardized similarity threshold setting is performed, specifically: based on the qualified standardized similarity threshold corresponding to the acquired multi-source heterogeneous address data, a target value corresponding to address similarity threshold adjustment is set; Address similarity threshold adjustment anomaly judgment is performed, specifically: when the standardized similarity threshold corresponding to the multi-source heterogeneous address data is adjusted to the qualified standardized similarity threshold, it is determined whether the deviation value of the re-acquired standardized result is greater than the predefined deviation value of the standardized result, and whether the deviation value of the re-acquired standardized result is greater than the deviation value of the standardized result in the initial state of address similarity threshold adjustment; The address similarity threshold adjustment anomaly feedback is performed as follows: if the deviation value of the re-acquired standardized result is greater than the predefined deviation value of the standardized result, or greater than the deviation value of the standardized result in the initial state of address similarity threshold adjustment, an address similarity threshold adjustment alarm is sent and the initial standardized similarity threshold is returned; otherwise, the spatial mapping position drift corresponding to the multi-source heterogeneous address data continues to be monitored. The qualified standardized similarity threshold is obtained by inputting the deviation value of the standardized result into the predefined address modeling standardized association table.

9. The intelligent modeling method for multi-source heterogeneous address data for smart cities as described in claim 7, characterized in that, The address entity and data time-varying association verification also includes monitoring the spatial mapping location drift of multi-source heterogeneous address data. Specifically, based on the number of geocoded failed spatial points in the drift time series, a mapping drift value is obtained to reflect the degree of drift of the spatial mapping location corresponding to the multi-source heterogeneous address data. The mapping drift risk level is graded based on the mapping drift value, as follows: When the mapping drift value is greater than a predefined acceptable mapping drift value, the index grid reduction adjustment is performed in the next preset monitoring time period. The index grid reduction adjustment means using a set index grid reduction step size as the adjustment amount and a set number of index grid reductions as the adjustment number, progressively reducing the index grid corresponding to the multi-source heterogeneous address data. At the end of the adjustment, the mapping drift value is re-monitored. If the mapping drift value is still greater than the predefined acceptable mapping drift value, an index grid reduction anomaly alarm is sent; otherwise, modeling decision verification is performed. The parameter combination corresponding to the index grid reduction step size and the number of index grid reductions is obtained by inputting the mapping drift value and the storage rate of the multi-source heterogeneous address data into the index grid reduction guidance table. When the mapping drift value is greater than a predefined acceptable mapping drift value, the index grid reduction anomaly alarm is sent. When the mapping drift value is not greater than the predefined qualified mapping drift value, continue to monitor whether the average retrieval rate of multi-source heterogeneous address data is greater than the predefined maximum multi-source data retrieval rate. If not, perform modeling decision verification; otherwise, perform index grid increase adjustment in the next preset monitoring time period. The index grid increase adjustment means using a set index grid increase step size as the adjustment amount and a set index grid increase number as the adjustment number, gradually increasing the index grid corresponding to the multi-source heterogeneous address data. At the end of the adjustment, re-monitor the mapping drift value, the storage rate of multi-source heterogeneous address data, and the average retrieval rate of multi-source heterogeneous address data. If the average retrieval rate of multi-source heterogeneous address data is not greater than the predefined maximum multi-source data retrieval rate and the mapping drift value is not greater than the predefined qualified mapping drift value, perform modeling decision verification, that is, perform modeling based on the corresponding multi-source heterogeneous address data; otherwise, send an index grid increase anomaly alarm. The parameter combination corresponding to the index grid increase step size and the index grid increase number is obtained by inputting the mapping drift value and the storage rate of multi-source heterogeneous address data into the index grid increase guidance table.

10. A multi-source heterogeneous address data intelligent modeling system for smart cities, used to implement the multi-source heterogeneous address data intelligent modeling method for smart cities as described in any one of claims 1-9, characterized in that, The system includes: a time-varying association determination module, a modeling decision verification determination module, and a time-series correction module. The time-varying association determination module is used to analyze the addition of address information and the updating of text addresses based on multi-source heterogeneous address data reflecting the geographic positioning characteristics of urban resources. Based on the analysis results, it determines whether to perform time-varying association verification between address entities and data to identify address entity identification errors, deviations of standardization results from semantics, and spatial mapping location drift corresponding to the multi-source heterogeneous address data. The modeling decision verification determination module is used to determine whether to trigger modeling decision verification based on the results of the time-varying association verification between address entities and data if the determination is to perform it. The time-series correction module is used to perform time-series correction of address data based on the analysis results of the addition of address information and the updating of text addresses if the determination is not to perform it. After the time-series correction of address data is completed, it performs time-varying association verification between address entities and data.

Images