Intelligent traffic asset data management method and system

Abstract

The invention discloses an intelligent traffic asset data management method and system, and relates to the technical field of data management, and the method comprises the steps: carrying out multi-dimensional analysis to determine a data classification target label, constructing a machine learning model, obtaining an intelligent data classification model, and carrying out the recognition and classification of asset data. Generating a data label to label data in the traffic asset database; establishing a response association relationship between the traffic asset data and a data application scene, and performing data label dependency and transparency analysis to obtain a data calling constraint; and searching and filtering in the traffic asset database according to the data label in the data calling constraint, and outputting scene application data. The technical problems that in the prior art, due to the fact that the incidence relation between data and application scenes cannot be preset in the data management process, search condition setting is complex, data matching is not accurate, and omission exists during data query and application are solved, and the technical effect of improving data management efficiency is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and in particular to an intelligent transportation asset data management method and system. Background Art

[0002] Existing transportation asset data management technologies usually focus on the security and accuracy of data, but often ignore the effective connection between data and specific application scenarios in the data management process. Traditional management methods often start setting search conditions for database queries only when encountering application needs for transportation asset data. This method increases the complexity of processing, because the data is not pre-set with the corresponding relationship with the application scenario when it is managed, which may result in the inability to accurately match the data required for the scenario during actual search. In addition, differences in the feature format and content of the data may also lead to incomplete or erroneous search results, resulting in incomplete data and even difficulty in searching. The traditional method requires analysis after obtaining the search conditions and application scenarios to determine which data are relevant, and then conduct a secondary search. This process is prone to missing key data, resulting in incomplete analysis. Summary of the invention

[0003] The present application provides an intelligent transportation asset data management method and system, which is used to solve the technical problem that the prior art fails to preset the relationship between data and application scenarios during the data management process, resulting in complex search condition settings, inaccurate data matching, and omissions in data query and application.

[0004] In view of the above problems, the present application provides an intelligent transportation asset data management method and system.

[0005] The first aspect of the present application provides an intelligent traffic asset data management method, the method comprising: Perform multi-dimensional analysis on the traffic asset data to determine the data classification target label, which is used to characterize the multi-dimensional data type, and each traffic asset data has a label for each dimension; based on the data classification target label, build a machine learning model, and perform deep learning convergence on the machine learning model by building a training sample set to obtain an intelligent data classification model; connect to the traffic asset database, identify and classify the asset data through the intelligent data classification model, generate data labels, and use the data labels to annotate the data in the traffic asset database; establish a response association relationship between the traffic asset data and the data application scenario; perform data label dependency and transparency analysis based on the response association relationship to obtain data call constraints; search and filter in the traffic asset database based on the data labels in the data call constraints, and output scenario application data.

[0006] The second aspect of the present application provides an intelligent traffic asset data management system, the system comprising: The parsing module is used to perform multi-dimensional parsing on the traffic asset data and determine the data classification target label, which is used to characterize the multi-dimensional data type, and each traffic asset data has a label for each dimension; the model building module is used to build a machine learning model based on the data classification target label, and obtain an intelligent data classification model by deep learning convergence of the machine learning model through building a training sample set; the identification and classification module is used to connect to the traffic asset database, identify and classify the asset data through the intelligent data classification model, generate data labels, and use the data labels to annotate the data in the traffic asset database; the association relationship establishment module is used to establish a response association relationship between the traffic asset data and the data application scenario; the data call constraint acquisition module is used to perform data label dependency and transparency analysis according to the response association relationship to obtain data call constraints; the scenario application data acquisition module is used to search and filter in the traffic asset database according to the data labels in the data call constraints, and output scenario application data.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: The present application performs multi-dimensional analysis on traffic asset data to determine data classification target labels, wherein the data classification target labels are used to characterize multi-dimensional data types, and each traffic asset data has a label for each dimension; based on the data classification target labels, a machine learning model is constructed, and deep learning convergence of the machine learning model is performed by constructing a training sample set to obtain an intelligent data classification model; a traffic asset database is connected, and asset data is identified and classified through the intelligent data classification model to generate data labels, and the data in the traffic asset database is annotated using the data labels; a response association relationship between traffic asset data and data application scenarios is established; data label dependency and transparency analysis is performed based on the response association relationship to obtain data call constraints; and search and filter processing is performed in the traffic asset database based on the data labels in the data call constraints to output scenario application data. The present invention solves the technical problem that the prior art fails to preset the association between data and application scenarios in the process of data management, resulting in complex search condition setting, inaccurate data matching, and omissions in data query and application. By performing multi-dimensional analysis on traffic asset data, data classification target labels are determined and an intelligent data classification model is constructed. The model is used to identify and label data, and the association between data and scenarios is established in combination with application scenarios. Through transparency analysis and dependency analysis, data call constraints are obtained to ensure data privacy protection. Data search and filtering are performed according to the constraints, and scenario application data that meets business needs is output, thereby achieving the technical effect of improving data management efficiency, ensuring efficient matching of data and scenarios, and ensuring the security of data privacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A flow chart of an intelligent traffic asset data management method provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of an intelligent transportation asset data management system provided in an embodiment of the present application.

[0010] Explanation of the accompanying drawings: parsing module 11, model building module 12, identification and classification module 13, association relationship establishment module 14, data call constraint acquisition module 15, scenario application data acquisition module 16. DETAILED DESCRIPTION

[0011] The present application provides an intelligent transportation asset data management method and system to solve the technical problem that the prior art fails to preset the relationship between data and application scenarios in the data management process, resulting in complex search condition settings, inaccurate data matching, and omissions in data query and application. By performing multi-dimensional analysis on transportation asset data, data classification target labels are determined and an intelligent data classification model is constructed. The model is used to identify and annotate data, and the relationship between data and scenarios is established in combination with application scenarios. Through transparency analysis and dependency analysis, data call constraints are obtained to ensure data privacy protection. Data search and filtering are performed according to the constraints, and scenario application data that meets business needs is output, thereby achieving the technical effect of improving data management efficiency, ensuring efficient matching of data and scenarios, and ensuring the security of data privacy.

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0013] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0014] Embodiment 1, as Figure 1 As shown, the present application provides an intelligent traffic asset data management method, the method comprising: Step S100: Perform multi-dimensional analysis on the traffic asset data to determine a data classification target label, where the data classification target label is used to characterize the multi-dimensional data type, and each traffic asset data has a label for each dimension.

[0015] In an embodiment of the present application, in order to perform multi-dimensional analysis of traffic asset data, traffic asset data is first obtained from multiple data sources and integrated, and the data source is labeled to ensure the integrity and traceability of the data. Next, by setting data clustering features, including collection equipment, time features, spatial features, geographic location, data structure, data content features, and environmental features, the attributes of each piece of data are fully described. Based on these features, a corresponding data distance metric is set to measure the similarity between different data. Then, a data clustering algorithm is used to perform hierarchical clustering to classify similar data points into the same category, and a data classification target label is determined for each category of data based on the clustering results. The determined data classification target label is used to characterize multi-dimensional data types, and each traffic asset data has a label for each dimension.

[0016] Furthermore, in the method provided in the embodiment of the application, multi-dimensional analysis of traffic asset data is performed to determine data classification target labels, and further includes: Acquire all sources of traffic asset data, integrate the collected data from all sources, and mark the data sources; set data clustering features, which include collection equipment, time features, spatial features, geographic location, data structure, data content features, and environmental features; set data distance metrics based on the data clustering features; hierarchically cluster the traffic asset data according to the data clustering features and data distance metrics, and determine the data classification target label according to the clustering results.

[0017] In the embodiment of the present application, all sources of traffic asset data are first obtained and integrated. The sources of traffic asset data are wide and diverse, mainly including traffic infrastructure data (such as road type, pavement structure, traffic signs and markings, bridge type, tunnel information, etc.), traffic equipment data (such as traffic lights, dynamic weighing sensors, surveillance cameras, parking facilities, etc.), traffic flow and vehicle data (such as traffic flow, speed, vehicle type classification, seasonal fluctuations, etc.) and environmental and external data (such as weather, road conditions, holidays, etc.). Each data source is stored in a different system, such as a traffic monitoring platform, a geographic information system (GIS), a vehicle management database, and an environmental monitoring system. In order to integrate data from these multiple sources, the ETL (Extract, Transform, Load) tool is used to standardize data of different formats and structures by extracting, transforming, and loading, and store them in a unified database. At the same time, during the data integration process, the data source is labeled for each piece of data, and its acquisition equipment type, timestamp, geographic location, and environmental information are recorded to ensure the traceability of the data.

[0018] Next, set the data clustering features. Data clustering features include multiple dimensions, specifically, including collection equipment features (such as equipment type, collection equipment model, etc.), time features (such as data timestamp, collection period, time period, etc.), spatial features (such as the distribution of data collection areas, such as urban areas, road section information, etc.), geographic location (using GPS or GIS technology to mark the specific location of the data, such as longitude and latitude information), data structure features (such as the organization of data, such as tabular data, image data or time series data, etc.), data content features (such as traffic flow, vehicle speed, accident conditions and other content information) and environmental features (such as weather conditions, road conditions, holidays and other external factors).

[0019] After setting the clustering features, set the data distance metric, which is a method used to calculate the similarity between data points. For numerical data, use Euclidean distance to measure the difference between data points. For categorical data, such as equipment type or weather conditions, use Jaccard similarity coefficient or Hamming distance to evaluate similarity. To ensure that different features contribute equally to the clustering analysis, standardize the numerical data, such as using Min-Max normalization or Z-score normalization method, to ensure that all feature data are compared at the same scale.

[0020] Then, based on the set clustering features and distance metrics, a hierarchical clustering algorithm is used to perform cluster analysis on the traffic asset data. Hierarchical clustering is a commonly used unsupervised learning algorithm that gradually merges similar data points into clusters by calculating the similarity between data points. Specifically, first, each data point is regarded as an independent cluster, and then the most similar clusters are gradually merged by calculating the similarity between clusters until all data points are merged into a large cluster. The advantage of hierarchical clustering is that it does not require a preset number of clusters, and can form a hierarchical cluster structure by calculating the distance relationship between different clusters. The clustering results are usually represented by a tree diagram to intuitively display the hierarchical structure of the data. In this process, the traffic asset data is divided into multiple groups according to the similarity of the clusters, and each group contains traffic data with similar attributes.

[0021] Finally, based on the clustering results, a data classification target label is assigned to each cluster. These labels are named according to the characteristics of each cluster, such as "high traffic cluster", "low traffic cluster", "normal equipment cluster", "equipment failure cluster", etc. Each piece of data will be assigned the label of its cluster, so that the data can be identified and classified in multiple dimensions. These data classification target labels not only describe the type of data, but also reflect the clustering hierarchical relationship of the data. For example, some labels may represent the basic categories of data (such as "normal state" or "fault state"), while some labels may further distinguish the specific characteristics of the data (such as "high traffic" or "low traffic"). Through this process, the traffic asset data is finally labeled with multiple dimensions, which is convenient for subsequent management, query and application.

[0022] Through the above steps, traffic asset data is acquired and integrated from multiple sources, and then through steps such as feature extraction, distance measurement and hierarchical clustering, finally a data classification target label with a hierarchical relationship is formed.

[0023] Furthermore, in the method provided in the embodiment of the application, hierarchical clustering is performed on the traffic asset data according to the data clustering characteristics and the data distance metric, and the data classification target label is determined according to the clustering result, and further includes: According to the data clustering characteristics, a clustering feature level is set, and the clustering feature level corresponds to the feature range relationship of the data clustering characteristics; according to the clustering hierarchical relationship of the clustering results, a hierarchical label is set according to the cluster center characteristics of the clustering results; the hierarchical label of the clustering feature level is merged with the hierarchical label of the corresponding clustering result to determine the data classification target label, and the data classification target label has a hierarchical relationship.

[0024] In the embodiment of the present application, firstly, the clustering feature level is set according to the set data clustering features, that is, a hierarchical structure is set for different features. The setting of the feature level is based on the feature range relationship of the data clustering features. The feature range of the high level is larger, indicating macro and broader attributes; while the feature range of the low level is smaller, indicating specific and refined attributes. For example, facility types (such as road sections, intersections, bridges, etc.) belong to higher-level features, and their range is wider, covering major categories of traffic assets; while features such as high flow, medium flow, and low flow belong to lower levels, and their range is smaller, focusing on specific traffic conditions.

[0025] After completing the hierarchical setting, the traffic asset data is classified according to the hierarchical clustering results. Hierarchical clustering gradually aggregates the data into multiple clusters based on the similarities between data points, and finally forms a tree structure to represent the hierarchical relationship of the data. Each cluster represents a group of similar data points, and the hierarchical relationship between clusters shows their differences in the feature space. For example, in the analysis of traffic flow data, the top-level label may be "traffic status", which includes a wide range of categories such as "normal", "congested", "under construction", etc.; while the bottom-level labels are specific to the detailed features of each data, such as "high flow", "medium flow", and "low flow".

[0026] According to the results of hierarchical clustering, hierarchical labels are set for the central features of each cluster, and the data is further refined according to these hierarchical labels. For example, in the cluster analysis of traffic flow, the traffic feature is an intermediate label (secondary label), which can be subdivided into the bottom-level labels of "high flow", "medium flow" and "low flow". These bottom-level labels represent the actual flow situation in the data and describe the microscopic characteristics of the traffic state.

[0027] Next, the hierarchical labels of the clustering feature level are merged with the hierarchical labels of the corresponding clustering results to finally determine the target labels for data classification. These target labels not only characterize the multi-dimensional types of data, but also reflect the hierarchical relationship between data. Through the hierarchical label structure, data classification not only has breadth (such as facility type, traffic status, etc.) and depth (such as high or low traffic, fast or slow speed, etc.), but also can describe the specific status and behavior of traffic assets in a refined manner. These labels provide a clear classification basis for subsequent intelligent decision-making, analysis and application.

[0028] Specifically, data classification target labels can be divided into three levels: top-level labels (macro labels), middle-level labels (secondary labels), and bottom-level labels (micro labels). Top-level labels are used to perform the most preliminary classification of data, usually based on the main attributes or large-scale characteristics of the data, such as "facility type" (road section, intersection, bridge, etc.) and "traffic status" (normal, congestion, accident, construction, etc.). Middle-level labels are labels that are further subdivided on the basis of top-level labels, and are used to describe the behavior of traffic data under specific conditions, such as "flow characteristics" (high flow, medium flow, low flow) and "environmental conditions" (sunny, rainy, snowy, etc.). Bottom-level labels are the identification of specific and detailed characteristics of traffic asset data, such as "time period labels" (morning, afternoon, morning peak, evening peak, etc.), "vehicle speed labels" (high speed, low speed), and "facility performance labels" (equipment failure, sensor abnormality, etc.). These hierarchical labels gradually refine the classification of data through hierarchical clustering to ensure that data can be clearly and specifically described at different levels.

[0029] In summary, by setting the clustering feature level and assigning hierarchical labels according to the hierarchical clustering results, these hierarchical labels are finally fused into data classification target labels.

[0030] Step S200: Based on the data classification target label, a machine learning model is constructed, and deep learning convergence of the machine learning model is performed by constructing a training sample set to obtain an intelligent data classification model.

[0031] In an embodiment of the present application, intelligent data classification is performed by constructing a multi-layer machine learning model based on the data classification target label. Specifically, a multi-layer machine learning model is first set according to the hierarchical relationship of the data classification target label. The model includes multiple label processing layers, each layer corresponds to a specific classification target label, for example, a top-level label (such as "traffic status"), an intermediate layer label (such as "traffic characteristics"), and a bottom-level label (such as "specific time period" or "vehicle speed label"). Each layer processes a specific label, and through a multi-level model structure, the classification of the data is gradually refined.

[0032] Next, we construct training and test datasets at different levels based on the characteristics of the labels at each level. Each training dataset contains the input data and corresponding target labels of the labels at that level, while the test dataset is used to evaluate the classification effect of the model on the labels at that level. These datasets are derived from the previously labeled traffic asset data to ensure that the training and test data at each level are relevant and representative.

[0033] Each label processing layer is then trained using training and test data sets at each level. Through deep learning methods, the model is continuously optimized and gradually converged, so that each layer can accurately classify specific labels. During the training process, the back propagation algorithm and gradient descent optimization strategy are used to continuously adjust the weights and biases of each layer to ensure that the classification results of each layer are close to the true label. Finally, after multiple training cycles and optimization steps, an intelligent data classification model is obtained, which can accurately classify data through multi-layer label processing based on the multi-dimensional characteristics of the input data.

[0034] Furthermore, in the method provided in the embodiment of the application, based on the data classification target label, a machine learning model is constructed, and deep learning convergence of the machine learning model is performed by constructing a training sample set to obtain an intelligent data classification model, and further includes: According to the hierarchical relationship of the data classification target labels, a multi-layer machine learning model is set, and the multi-layer machine learning model has multiple label processing layers; according to the data classification target labels, training data sets and test data sets at each level are constructed; and the training data sets and test data sets at each level are used to train and converge each label processing layer respectively to obtain the intelligent data classification model.

[0035] In an embodiment of the present application, accurate classification of data is achieved by constructing a multi-layer machine learning model based on the hierarchical relationship of the data classification target labels. Specifically, a multi-layer machine learning model is first set according to the hierarchical relationship of the data classification target labels. Each label level corresponds to a label processing layer of the model. The top-level label (such as "traffic status") is responsible for a wide range of classifications, the middle-level labels (such as "traffic characteristics") refine the classification, and the bottom-level labels (such as "vehicle speed" or "time period") perform the most detailed processing of the data. Through this hierarchical design, data is processed layer by layer and the classification results of each level are accurately output.

[0036] Next, training data sets and test data sets for each level are constructed based on these hierarchical labels. Through data extraction and preprocessing methods, data related to each label level is extracted from preset databases (such as SQL databases, NoSQL databases, etc.). Each training data set contains the input features and target labels required for the level label, and the test data set is used for subsequent model verification. During the model training process, deep learning methods such as multi-layer perceptron (MLP) are used to learn the training data set. The label processing layer of each layer learns layer by layer according to the data and target labels in the training set, and adjusts the model parameters through the back propagation algorithm and gradient descent optimization method. The model of each layer updates its parameters through multiple iterations until the model reaches the preset number of training times on the training data set, and it is considered that the model has converged and can stably output prediction results. The system uses the test data set to evaluate the model during the training process to ensure that the model can generalize and prevent overfitting.

[0037] Finally, after multiple trainings, an intelligent data classification model was obtained. The model can process the input data layer by layer and output accurate classification results, reflecting the multi-dimensional characteristics and hierarchical labels of traffic asset data.

[0038] Step S300: connecting to the traffic asset database, identifying and classifying the asset data through the intelligent data classification model, generating data labels, and using the data labels to label the data in the traffic asset database.

[0039] In the embodiment of the present application, the data of traffic assets (such as the operating status of roads, bridges, monitoring equipment, etc.) are first combined with the constructed intelligent data classification model by connecting to the traffic asset database. The traffic asset database stores a large amount of traffic-related information, such as traffic flow, vehicle data, traffic facility status, environmental conditions, etc. In order to realize the intelligent classification and management of these data, the database is connected to ensure that the relevant data of traffic assets can be accessed and read in real time.

[0040] Once the database is successfully connected, the intelligent data classification model is used to identify and classify these traffic asset data. The intelligent data classification model outputs the corresponding classification label based on the input characteristics of each data (such as flow, speed, time, location, etc.). Each traffic asset data will be analyzed by the model and assigned one or more data labels, which correspond to different levels of classification results, such as "high flow", "congestion", "equipment failure", etc.

[0041] Through these tags, the traffic asset data in the traffic asset database are annotated. The annotation process is to associate the tags generated by the model with the original data in the traffic asset database and add specific classification information to each piece of data.

[0042] Step S400: Establishing a responsive association relationship between traffic asset data and data application scenarios.

[0043] In the embodiment of the present application, the association rule learning algorithm is first used to mine the association rules between data features and application scenarios from the traffic asset data. These rules reflect the potential connection between data features and scenario requirements. Then, based on the mined association rules, the specific association relationship between data and scenarios is identified. Then, the key data tags in the traffic asset data are extracted and combined with the identified association relationship; finally, based on these data tags and association relationships, the response association relationship between the data application scenario and the data tags is established.

[0044] Furthermore, in the method provided in the embodiment of the application, the establishment of a responsive association relationship between traffic asset data and data application scenarios further includes: The association rules between the traffic asset data and the data application scenarios are mined through an association rule learning algorithm; the association relationship is identified based on the association rules; the data tags of the traffic asset data are extracted based on the association relationship, and based on the data tags and the association relationship, a response association relationship between the data application scenario and the data tags is established.

[0045] In the embodiment of the present application, the traffic asset data is first analyzed using an association rule learning algorithm (such as the Apriori algorithm) to explore the potential correlation between data features and generate frequent item sets and association rules. The data is stored in the form of a transaction table, where each row represents a transaction (such as the traffic status in a certain time period) and each column represents a data item (such as "vehicle speed>60km / h", "flow rate>1000", etc.). During the analysis process, the minimum support and minimum confidence thresholds are set to filter out rules with sufficient frequency and reliability. The support calculation formula is: the support of a certain item set is equal to the number of times the item set appears in all transactions divided by the total number of transactions. For example, if "vehicle speed>60km / h" appears 8 times in 10 transactions and the total number of transactions is 10, its support is 8 / 10=0.8. By calculating the support for each feature item, the frequent item sets with support greater than the set threshold are filtered out as the basis for subsequent rule generation.

[0046] Then, higher-order candidate item sets are generated based on the first-level frequent item sets. For example, starting from the two first-level item sets of "vehicle speed > 60km / h" and "flow rate > 1000", the second-level item set "vehicle speed > 60km / h ∧ flow rate > 1000" is generated, and its support is calculated to determine whether the item set meets the set minimum support threshold. If the support meets the requirements, it will be retained as a frequent item set; otherwise, the item set and all possible supersets will be pruned to reduce the amount of calculation. By repeating the generation and pruning steps, all frequent item sets that meet the minimum support are finally obtained. These item sets reflect the high-frequency co-occurrence patterns of feature items in traffic asset data.

[0047] Based on the frequent item sets, association rules are further generated. Association rules are expressed in the form of "premise → conclusion", for example, "vehicle speed > 60km / h → the probability of an accident increases". In order to measure the reliability of the rules, the confidence of each rule is calculated, that is, the proportion of transactions that meet the premise of the rule and the conclusion is also satisfied. The confidence calculation formula is: the confidence of a rule is equal to its support divided by the support of the premise item set. For example, the support of "vehicle speed > 60km / h ∧ traffic > 1000 → accident occurs" is 0.5, and the support of "vehicle speed > 60km / h ∧ traffic > 1000" is 0.6, then the confidence of the rule is 0.5 / 0.6=0.83. Rules with confidence greater than the set threshold (such as 0.7) are retained, and invalid rules with low confidence are eliminated.

[0048] After identifying valid association rules, data labels are extracted according to the rules. Data labels reflect the multi-dimensional characteristics of transactions, such as "high traffic", "high speed", "accident risk", etc. For example, the label corresponding to the rule "high risk of accident when speed > 60km / h and traffic > 1000" is "high traffic, high speed". The generated labels are attached to each data record in the traffic asset database to form structured annotated data.

[0049] Then, data labels are extracted from the traffic asset data based on the association relationship. Data labels are high-level identifiers of traffic asset data characteristics, such as "high traffic", "high speed", "accident risk", etc. Through the premise or conclusion conditions of the association rules, corresponding labels are attached to the traffic asset data that meets specific rules. For example, when the transaction data meets the conditions of "vehicle speed>60km / h∧flow rate>1000", "high traffic" and "high speed" labels are generated for the transaction, and these labels are written into the traffic asset database, and classification information is attached to each data record.

[0050] Finally, based on data tags and association relationships, the response association relationship between data application scenarios and data tags is established. By building a scenario response rule library, data tags are dynamically associated with data application scenarios. For example, when "high traffic" and "high speed" tags are detected in real-time data, the response rules in the rule library are automatically matched to trigger the corresponding application scenarios (such as "adjusting the traffic light cycle" or "deploying traffic diversion").

[0051] Step S500: performing data tag dependency and transparency analysis according to the response association relationship to obtain data call constraints.

[0052] In an embodiment of the present application, the dependency relationship between the data tag and the application scenario is first identified. This means that certain data tags are prerequisites for specific application scenarios, and only when these conditions are met can the corresponding scenario response be triggered. Next, based on the dependency relationship, the transparency of the data tags is identified, especially the sensitive data therein is identified, and these sensitive data require additional protection to ensure privacy and security. On this basis, sensitive data is used as a filtering constraint to filter and control data calls; and the dependency relationship is used as an extraction constraint to ensure that data calls can only be made when specific conditions are met.

[0053] Through this series of processes, data call constraints are ultimately generated to ensure that data calls not only comply with business logic, but can also be executed efficiently while ensuring data security and compliance.

[0054] Furthermore, in the method provided in the embodiment of the application, data label dependency and transparency analysis are performed according to the response association relationship to obtain data call constraints, and the method also includes: According to the response association relationship, the dependency relationship between the data tag and the data application scenario is identified; based on the dependency relationship, the traffic asset data of the data tag is identified for content transparency to determine sensitive data; the sensitive data is used as a filtering constraint and the dependency relationship is used as an extraction constraint to generate the data call constraint.

[0055] In an embodiment of the present application, firstly, the dependency between the data tag and the data application scenario is identified based on the response association relationship. Specifically, by analyzing historical data, identify which data tags are prerequisites for specific application scenarios. For example, if "vehicle speed>60km / h" and "traffic volume>1000" frequently appear in the "high-traffic road section" scenario, clarify the dependency between these tags and mark them as key conditions for the scenario. This step determines which data tags are necessary in the business scenario and which can be considered optional.

[0056] Then, based on the identified dependencies, transparency identification is performed, that is, the content transparency of the data labels is evaluated. Transparency indicates the degree of openness of the data. Data labels with high transparency (such as vehicle speed and traffic) usually do not involve privacy, while data labels with low transparency (such as license plate number and user location) may involve privacy or security issues and therefore need to be protected. Through the rule-based classification method, the privacy and security analysis of each label is performed, and the labels are divided into two categories: high transparency and low transparency. High transparency labels can be used publicly, while low transparency labels require further protection.

[0057] Next, based on the results of the transparency analysis, sensitive data identification is performed. For labels with low transparency (such as "license plate number" and "user location"), they are marked as sensitive data. The core of this step is to determine whether encryption, de-identification or other protection measures are required by checking whether the data label involves personal privacy or sensitive information.

[0058] After identifying sensitive data, protect it as a filtering constraint. Set sensitive data as objects that need to be desensitized based on the transparency and privacy attributes of the data label. For sensitive data, take desensitization measures, such as encryption or de-identification, to ensure that personal privacy or security information is not leaked during the data call and storage process. By setting filtering constraints, ensure that sensitive data is always properly protected when the data is called and will not be exposed to unauthorized visitors.

[0059] At the same time, extraction constraints are generated based on the identified dependencies. Extraction constraints ensure that when data is called, it will only be extracted and applied when the dependency conditions of a specific scenario are met. For example, the "high-traffic road section" application scenario requires the two conditions of "vehicle speed>60km / h" and "traffic volume>1000". Only when these two conditions are met at the same time will the corresponding data be extracted for application.

[0060] Finally, the filtering constraints and extraction constraints are combined to generate complete data call constraints. These constraints not only ensure the privacy and security of sensitive data but also ensure that data calls are made under the condition of meeting the dependency conditions, thereby improving the security and compliance of data management.

[0061] Furthermore, in the method provided by the application embodiment, using the sensitive data as a filtering constraint further includes: Evaluating the sensitivity of the sensitive data; according to the sensitivity, setting the content desensitization gradient of the sensitive data, desensitizing the sensitive data according to the desensitization gradient to obtain desensitized data; evaluating the impact of the desensitized data on the data application scenario, determining the desensitization execution gradient, and setting the filtering constraint based on the desensitization execution gradient.

[0062] In the embodiment of the present application, first, the sensitivity of the sensitive data is evaluated. Specifically, this process determines its privacy by analyzing the content of the data label, and then sets appropriate measures for desensitization processing. The specific method is to analyze according to the content of the data label. For example, data involving personal privacy such as license plate numbers, ID card numbers, and user locations are rated as highly sensitive data; while relatively public data such as traffic flow, weather, and road conditions are rated as low-sensitive data. Through a rule engine, each data label is classified and assigned a sensitivity level.

[0063] Next, a desensitization gradient is set for each type of data according to its sensitivity. The desensitization gradient refers to the depth and method of desensitization processing. Different desensitization techniques are applied to data with different sensitivities. For highly sensitive data, such as the ID card number and license plate number of the vehicle owner, complete desensitization or encryption methods are adopted. The encryption method ensures that the data is fully encrypted during storage and transmission and the original information cannot be leaked. For example, the license plate number may be fully encrypted or protected by full-field desensitization (such as replacing "沪A12345" with "沪A*****"). For moderately sensitive data, such as traffic flow and accident locations, fuzzification processing or partial desensitization is adopted. The data accuracy is reduced through fuzzification techniques, such as converting specific traffic flow data into a range value (such as "1000 - 1500") instead of the exact value "1200". For the accident location, only the general area is retained while hiding the specific street name. Low-sensitive data, such as weather and road conditions, usually do not require desensitization processing or are simply processed using de-identification techniques, that is, removing the identification information related to individuals to ensure the anonymity of the data.

[0064] After the desensitization gradient is set, the data is actually desensitized. For highly sensitive data, encryption or full-field desensitization is used to ensure that it is not leaked at all; for medium-sensitive data, fuzzification or partial encryption is used so that the data does not reveal specific details but still retains a certain business value; for low-sensitivity data, the simplest de-identification method is used to remove the identification information related to the individual. After desensitization is completed, the system obtains a batch of desensitized data, which can continue to support business analysis and applications on the basis of protecting privacy.

[0065] Then, we conduct an impact assessment to evaluate the impact of desensitized data on application scenarios. We use business rules and scenario requirements to determine whether the desensitized data can still meet business needs. For example, after the license plate number is encrypted, it may not be used directly for vehicle identification, but it can still be used for traffic analysis. For traffic flow data, the accuracy of the data is reduced through fuzzification, but it can still provide support for traffic analysis. By evaluating the impact of desensitized data on various application scenarios, we ensure that the desensitized data can continue to meet business needs and effectively support scenario analysis.

[0066] Finally, based on the impact assessment results of the desensitized data, determine the desensitization execution gradient, that is, the depth and method of the desensitization process. For highly sensitive data, full encryption or full-field desensitization is used; for medium-sensitive data, a lighter fuzzy process is used; and for low-sensitive data, only de-identification is performed. Based on these desensitization execution gradients, filter constraints are determined and applied to the data call and storage process to ensure that sensitive data will not be leaked during the processing and use of the data, and that the business scenario can still operate normally. Filter constraints ensure that when the data is called, it can only be used if it meets specific privacy protection requirements, while ensuring that the data can still meet business needs.

[0067] Through the above steps, it is ensured that sensitive data can be effectively used in various application scenarios after desensitization.

[0068] Step S600: Search and filter the traffic asset database according to the data tags in the data call constraint, and output scenario application data.

[0069] In an embodiment of the present application, when searching and filtering in the traffic asset database according to the data tags in the data call constraints, the tags that match the current business scenario are identified according to the set constraints and data tags. For example, if the business scenario requires the analysis of traffic flow, relevant tags such as "vehicle speed" and "flow" are identified as filtering conditions. Next, the data in the traffic asset database is retrieved using a database query language (such as SQL) to ensure that records that meet the requirements are extracted from the database. When the data that meets the conditions is retrieved, further data filtering is performed. At this point, the data is desensitized according to the preset privacy protection requirements. Specifically, sensitive fields are encrypted, obfuscated or de-identified according to the privacy sensitivity of the data tag. For example, if the retrieved license plate number data is highly sensitive data, the entire field will be desensitized or encrypted according to the previously set desensitization gradient to ensure that the data does not leak any personal information.

[0070] After completing the screening, searching and filtering of data tags, the scenario application data that finally meets the business needs will be output.

[0071] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects: The present application performs multi-dimensional analysis on traffic asset data to determine data classification target labels, wherein the data classification target labels are used to characterize multi-dimensional data types, and each traffic asset data has a label for each dimension; based on the data classification target labels, a machine learning model is constructed, and deep learning convergence of the machine learning model is performed by constructing a training sample set to obtain an intelligent data classification model; a traffic asset database is connected, and asset data is identified and classified through the intelligent data classification model to generate data labels, and the data in the traffic asset database is annotated using the data labels; a response association relationship between traffic asset data and data application scenarios is established; data label dependency and transparency analysis is performed based on the response association relationship to obtain data call constraints; and search and filter processing is performed in the traffic asset database based on the data labels in the data call constraints to output scenario application data. The present invention solves the technical problem that the prior art fails to preset the association between data and application scenarios in the process of data management, resulting in complex search condition setting, inaccurate data matching, and omissions in data query and application. By performing multi-dimensional analysis on traffic asset data, data classification target labels are determined and an intelligent data classification model is constructed. The model is used to identify and label data, and the association between data and scenarios is established in combination with application scenarios. Through transparency analysis and dependency analysis, data call constraints are obtained to ensure data privacy protection. Data search and filtering are performed according to the constraints, and scenario application data that meets business needs is output, thereby achieving the technical effect of improving data management efficiency, ensuring efficient matching of data and scenarios, and ensuring the security of data privacy.

[0072] Embodiment 2, based on the same inventive concept as the intelligent traffic asset data management method in the above embodiment, Figure 2 As shown, the present application provides an intelligent traffic asset data management system, and the system and method embodiments in the present application are based on the same inventive concept. The system includes: The parsing module 11 is used to perform multi-dimensional parsing on the traffic asset data and determine the data classification target label, which is used to characterize the multi-dimensional data type, and each traffic asset data has a label for each dimension; the model building module 12 is used to build a machine learning model based on the data classification target label, and to obtain an intelligent data classification model by deep learning convergence of the machine learning model by building a training sample set; the identification and classification module 13 is used to connect to the traffic asset database, identify and classify the asset data through the intelligent data classification model, generate data labels, and use the data labels to annotate the data in the traffic asset database; the association relationship establishment module 14 is used to establish a response association relationship between the traffic asset data and the data application scenario; the data call constraint acquisition module 15 is used to perform data label dependency and transparency analysis based on the response association relationship to obtain data call constraints; the scenario application data acquisition module 16 is used to search and filter in the traffic asset database according to the data labels in the data call constraints, and output scenario application data.

[0073] Furthermore, the system is also used to implement the following functions: Acquire all sources of traffic asset data, integrate the collected data from all sources, and mark the data sources; set data clustering features, which include collection equipment, time features, spatial features, geographic location, data structure, data content features, and environmental features; set data distance metrics based on the data clustering features; hierarchically cluster the traffic asset data according to the data clustering features and data distance metrics, and determine the data classification target label according to the clustering results.

[0074] Furthermore, the system is also used to implement the following functions: According to the data clustering characteristics, a clustering feature level is set, and the clustering feature level corresponds to the feature range relationship of the data clustering characteristics; according to the clustering hierarchical relationship of the clustering results, a hierarchical label is set according to the cluster center characteristics of the clustering results; the hierarchical label of the clustering feature level is merged with the hierarchical label of the corresponding clustering result to determine the data classification target label, and the data classification target label has a hierarchical relationship.

[0075] Furthermore, the system is also used to implement the following functions: According to the hierarchical relationship of the data classification target labels, a multi-layer machine learning model is set, and the multi-layer machine learning model has multiple label processing layers; according to the data classification target labels, training data sets and test data sets at each level are constructed; and the training data sets and test data sets at each level are used to train and converge each label processing layer respectively to obtain the intelligent data classification model.

[0076] Furthermore, the system is also used to implement the following functions: The association rules between the traffic asset data and the data application scenarios are mined through an association rule learning algorithm; the association relationship is identified based on the association rules; the data tags of the traffic asset data are extracted based on the association relationship, and based on the data tags and the association relationship, a response association relationship between the data application scenario and the data tags is established.

[0077] Furthermore, the system is also used to implement the following functions: According to the response association relationship, the dependency relationship between the data tag and the data application scenario is identified; based on the dependency relationship, the traffic asset data of the data tag is identified for content transparency to determine sensitive data; the sensitive data is used as a filtering constraint and the dependency relationship is used as an extraction constraint to generate the data call constraint.

[0078] Furthermore, the system is also used to implement the following functions: Evaluate the sensitivity of the sensitive data; according to the sensitivity, set a content desensitization gradient for the sensitive data, desensitize the sensitive data according to the desensitization gradient, and obtain desensitized data; perform an impact evaluation on the data application scenario according to the desensitized data, determine a desensitization execution gradient, and set the filtering constraints based on the desensitization execution gradient.

[0079] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0080] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0081] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. An intelligent traffic asset data management method, characterized in that: include: Perform multi-dimensional analysis on the traffic asset data to determine a data classification target label, wherein the data classification target label is used to characterize the multi-dimensional data type, and each traffic asset data has a label for each dimension; Based on the data classification target label, a machine learning model is constructed, and a training sample set is constructed to perform deep learning convergence on the machine learning model to obtain an intelligent data classification model; Connecting to a traffic asset database, identifying and classifying asset data through the intelligent data classification model, generating data labels, and using the data labels to label data in the traffic asset database; Establish responsive associations between traffic asset data and data application scenarios; Perform data label dependency and transparency analysis according to the response association relationship to obtain data call constraints; The data tags in the data call constraints are searched and filtered in the traffic asset database to output scenario application data.

2. The intelligent traffic asset data management method according to claim 1, characterized in that: Perform multi-dimensional analysis on traffic asset data to determine data classification target labels, including: Obtain all sources of traffic asset data, integrate the collected data from all sources, and annotate the data sources; Setting data clustering features, wherein the data clustering features include collection equipment, time features, spatial features, geographic location, data structure, data content features, and environmental features; Based on the data clustering characteristics, setting a data distance metric; The traffic asset data is hierarchically clustered according to the data clustering characteristics and the data distance metric, and the data classification target label is determined according to the clustering result.

3. The intelligent traffic asset data management method according to claim 2 is characterized in that: Hierarchical clustering of the traffic asset data is performed according to the data clustering characteristics and the data distance metric, and the data classification target label is determined according to the clustering result, including: According to the data clustering feature, a clustering feature level is set, wherein the clustering feature level corresponds to a feature range relationship of the data clustering feature; According to the clustering hierarchical relationship of the clustering results, hierarchical labels are set according to the cluster center characteristics of the clustering results; The hierarchical labels of the clustering feature levels are fused with the hierarchical labels of the corresponding clustering results to determine the data classification target labels, and the data classification target labels have a hierarchical relationship.

4. The intelligent traffic asset data management method according to claim 3 is characterized in that: Based on the data classification target label, a machine learning model is constructed, and a training sample set is constructed to perform deep learning convergence on the machine learning model to obtain an intelligent data classification model, including: According to the hierarchical relationship of the data classification target labels, a multi-layer machine learning model is set, wherein the multi-layer machine learning model has multiple label processing layers; Constructing training data sets and test data sets at each level according to the data classification target labels; The training data sets and test data sets of each level are used to train and converge each label processing layer to obtain the intelligent data classification model.

5. The intelligent traffic asset data management method according to claim 1, characterized in that: The establishing of a responsive association relationship between traffic asset data and data application scenarios includes: Mining association rules between the traffic asset data and data application scenarios through an association rule learning algorithm; According to the association rule, identifying the association relationship; The data tags of the traffic asset data are extracted according to the association relationship, and based on the data tags and the association relationship, a response association relationship between the data application scenario and the data tags is established.

6. The intelligent traffic asset data management method according to claim 5, characterized in that: According to the response association relationship, data label dependency and transparency analysis are performed to obtain data call constraints, including: According to the response association relationship, identifying the dependency relationship between the data tag and the data application scenario; Based on dependencies, content transparency identification is performed on the traffic asset data of data tags to determine sensitive data; The sensitive data is used as a filtering constraint, and the dependency relationship is used as an extraction constraint to generate the data call constraint.

7. The intelligent traffic asset data management method according to claim 6, characterized in that: The sensitive data is used as a filtering constraint, including: Evaluate the sensitivity of the sensitive data; According to the sensitivity, setting a content desensitization gradient for the sensitive data, and performing desensitization processing on the sensitive data according to the desensitization gradient to obtain desensitized data; An impact assessment is performed on the data application scenario according to the desensitized data, a desensitization execution gradient is determined, and the filtering constraints are set based on the desensitization execution gradient.

8. An intelligent traffic asset data management system, characterized in that: The system comprises: A parsing module, used to perform multi-dimensional parsing on the traffic asset data and determine a data classification target label, wherein the data classification target label is used to characterize the multi-dimensional data type, and each traffic asset data has a label for each dimension; A model building module is used to build a machine learning model based on the data classification target label, and to perform deep learning convergence on the machine learning model by building a training sample set to obtain an intelligent data classification model; An identification and classification module, which is used to connect to a traffic asset database, identify and classify asset data through the intelligent data classification model, generate data labels, and use the data labels to annotate data in the traffic asset database; The association relationship establishment module is used to establish the response association relationship between traffic asset data and data application scenarios; A data call constraint acquisition module is used to perform data label dependency and transparency analysis according to the response association relationship to obtain data call constraints; The scenario application data acquisition module is used to search and filter in the traffic asset database according to the data tags in the data call constraints, and output the scenario application data.

Images