Digital modeling response method, system, equipment and medium based on artificial intelligence

Abstract

The present invention discloses a digital modeling response method, system, device and medium based on artificial intelligence, wherein the method includes: S1: obtaining multi-level public security system business data; S2: performing pre-processing operations of data filtering, data recovery, data enrichment and data serialization on the public security system business data to form structured data; S3: the user inputs the text information of the digital modeling requirement through a visual interface, and uses a large language model to perform semantic recognition on the input text information to obtain text information features; S4: selecting corresponding digital modeling requirement information features according to the text information features, building a data analysis model in combination with standard structured data, and outputting digital modeling results according to the data analysis model. The present invention solves the problem that the digital modeling technology based on artificial intelligence has a high entry threshold and complex use process in the field of public security, and non-professional users are significantly restricted in participating in the digital modeling process.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and in particular relates to a digital modeling response method, system, device and medium based on artificial intelligence. Background Art

[0002] Digital modeling technology is widely used in risk prediction and assessment in the field of public security. By establishing mathematical models, historical data can be analyzed, potential risk factors can be identified, and possible public security incidents in the future can be predicted. This predictive ability helps public security organs take measures in advance to reduce risks. However, traditional digital modeling platforms have high requirements for personnel quality and require multi-faceted capabilities and qualities. Digital modeling involves complex mathematical operations and statistical analysis, requiring users to have a solid foundation in mathematics and statistics and to be able to understand and apply various mathematical models and algorithms. In addition, traditional digital modeling platforms often require users to write code or scripts to implement specific modeling functions, and users need to have certain programming skills. The above requirements limit more people from participating in digital modeling work.

[0003] The patent with publication number WO2022100357A1 discloses a method and storage device for constructing a deployment text entity relationship extraction model. The method for constructing a deployment text entity relationship extraction model includes the steps of: creating a project project and creating a data storage directory; collecting data and preprocessing the data; labeling the preprocessed data according to a fourth preset specification, extracting evaluation data, and performing data evaluation on the evaluation data according to a fifth preset specification; dividing the labeled data into different types of data; selecting a training model, and processing the different types of data to conform to the data format received by the training model; and performing model evaluation on the trained training model according to the processed test data and the processed evaluation data. If the evaluation is qualified, the qualified model is released. The method described in the invention is relatively conventional in the preprocessing operation of the data, and fails to explore the potential pattern, structure and integrity of the data. In addition, in the construction of the text entity relationship extraction model, the user must have relevant knowledge of data labeling and machine learning. During the labeling and evaluation stages, users usually need to use specific tools and techniques, which may pose an obstacle to non-professional users; and users need to perform manual operations in each step, such as data labeling and segmentation, selection of evaluation indicators, etc., which will reduce the efficiency of the overall process. Summary of the invention

[0004] The present invention provides an artificial intelligence-based digital modeling response method, system, device and medium, aiming to solve the problem that the existing artificial intelligence-based digital modeling technology has a high entry threshold and complex usage process in the public safety field, and non-professional users are significantly restricted in participating in the digital modeling process.

[0005] In order to solve the above technical problems, the present invention provides a digital modeling response method based on artificial intelligence, comprising the following steps:

[0006] S1: Obtain various forms of multi-level public security system business data through external data sources, and the public security system business data is used to query and warn of illegal behavior.

[0007] S2: Perform pre-processing operations such as data filtering, data recovery, data enrichment, and data serialization on the acquired public security system business data to form standard structured data.

[0008] S3: The user inputs the text information of the digital modeling requirements through the visual interface, and the large language model is used to perform semantic recognition on the text information input by the user to obtain the text information features.

[0009] S4: Select the corresponding digital modeling requirement information features according to the nature and task of the text information features, build a data analysis model based on standard structured data, and output the digital modeling results based on the data analysis model.

[0010] Preferably, the collected public security system business data includes, but is not limited to: public security system audit log data, public security system application data, public security organization data and public security personnel information data.

[0011] The public security system audit log data includes but is not limited to: security audit log, application audit log, and terminal behavior log.

[0012] The public security system application data includes, but is not limited to: application system inventory data, application access log data, application database access data, and application traffic data.

[0013] The public security organization data includes but is not limited to: organization name, organization code, department type, and position type.

[0014] The public security personnel information data includes but is not limited to: personnel name, ID number, affiliated organization, personnel rank, and personnel responsibilities.

[0015] Preferably, the pre-processing operation is specifically:

[0016] Data filtering is to remove duplicate records, null values ​​and abnormal values ​​from the business data of the public security system, clean it and format it uniformly, and filter the data according to the defined filtering conditions.

[0017] Data recovery is to construct data points for the business data of the public security system and form a data set, build a graph structure based on the data set, perform random path exploration on the constructed graph to generate several random paths, and calculate the recovery value of the missing data points through adaptive weighting based on the exploration results of the random paths, combined with the occurrence frequency and similarity of the random paths.

[0018] Data enrichment is to obtain the field names and field values ​​of the public security system business data for the fields in the public security system business data, obtain the field name and field value generation function of the target field according to the enrichment rules, and obtain the enriched public security system business data target field name and field value after calculation.

[0019] Data serialization is to generate uniformly serialized public security system business data according to the target public security system business data profile from the enriched public security system business data.

[0020] Preferably, the weight calculation in the adaptive weighted calculation is specifically as follows:

[0021]

[0022] In the formula, For the Random paths to missing data points The weight of is the weight adjustment factor, which controls the speed of distance attenuation; is a node in a random path and missing data points The Euclidean distance between them measures the spatial distance between nodes; is the weight adjustment factor, which controls the influence of path frequency; For Node The frequency of occurrence of nodes in a random path, i.e. The number of times it appears in multiple walks; is the number of walks, used to normalize the frequency.

[0023] Preferably, the semantic recognition of the text information input by the user using the large language model is performed to obtain the text information features as follows:

[0024] Entity recognition: identifying specific noun phrases in textual information through context.

[0025] Relation extraction: identifying and parsing semantic relationships between entities in textual information.

[0026] Semantic role labeling: Identifying the relationship between a predicate and its associated noun phrases in a textual sentence.

[0027] Dependency analysis: Study the structural relationship between words in text information sentences, find out the dependency relationship between words through syntactic analysis and semantic analysis, and thus understand the semantics of text information sentences.

[0028] Semantic template matching: A template-based semantic analysis method that compares text information with predefined templates to understand the semantics of the text information.

[0029] Deep learning semantic representation: Use deep learning technology to learn the semantic representation of text information, and automatically learn the semantic information of text information by training a large corpus.

[0030] Preferably, the data analysis model of the digital modeling specifically includes: an abnormal event analysis model, a statistical analysis model and a reverse tracing analysis model.

[0031] The abnormal event analysis model is specifically: using automated analysis algorithms to implement abnormal event analysis including abnormal accounts, unauthorized operations, high-frequency operations, sensitive data operations, large-scale export operations, abnormal time operations, and abnormal location operations, and supports the establishment of customized abnormal analysis models through the configuration of policy rules.

[0032] The statistical analysis model is specifically as follows: define the statistical analysis dimensions: "people, institutions, regions, applications, and operations", calculate the analysis indicators of each dimension through statistical methods, and conduct time series analysis and trend forecasting on the business data of the public security system.

[0033] The reverse tracing analysis model is specifically as follows: by defining known information and sensitive data in the business data of the public security system, constructing a personnel behavior graph, analyzing the behavioral associations and activity patterns between personnel, clustering the behaviors of personnel, mining potential associations, and performing reverse reasoning, the source event is traced and identified based on the graph structure.

[0034] Preferably, the method further includes the large language model generating modeling parameters and data analysis model performance indicators of the modeling results at the same time, and the user adjusts the data analysis model parameters of the digital modeling or triggers re-execution of the digital modeling operation through the large language model.

[0035] On the other hand, the present invention provides a digital modeling response system based on artificial intelligence, including a data acquisition module, a data preprocessing module, a user demand analysis module and a digital modeling module.

[0036] The data collection module is used to obtain various forms of multi-level public security system business data through external data sources. The public security system business data is used to query and warn of illegal behavior.

[0037] The data preprocessing module is used to perform preprocessing operations such as data filtering, data recovery, data enrichment and data serialization on the acquired public security system business data to form standard structured data.

[0038] The user demand analysis module is used for users to input text information of digital modeling requirements through a visual interface, and uses a large language model to perform semantic recognition on the text information input by users to obtain text information features.

[0039] The digital modeling module is used to select the corresponding digital modeling requirement information features according to the nature of the text information features and the task, build a data analysis model in combination with standard structured data, and output the digital modeling results according to the data analysis model.

[0040] On the other hand, the present invention also provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the digital modeling response method based on artificial intelligence as described in any embodiment of the present invention is implemented.

[0041] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the artificial intelligence-based digital modeling response method as described in any embodiment of the present invention is implemented.

[0042] Compared with the prior art, the present invention has the following technical effects:

[0043] The present invention provides a method for supporting regular personnel to use a custom digital modeling platform, which identifies and analyzes the digital modeling requirements input by users, and automatically outputs the results of digital modeling analysis and modeling analysis parameters for reference through semantic analysis and modeling scenario dialogue questions and answers. Through this method, the custom modeling platform lowers the user threshold and improves the friendliness of the modeling platform for non-professional users in the public safety field. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is an overall flow chart of the digital modeling response method based on artificial intelligence described in the present invention. DETAILED DESCRIPTION

[0045] In order to make the objectives, technical solutions and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in combination with specific embodiments of the present application and with reference to the accompanying drawings.

[0046] Embodiment 1

[0047] This embodiment provides a digital modeling response method based on artificial intelligence, which automatically obtains public security system business data from external data sources and pre-processes it to form standard structured data, and then uses a large language model to perform semantic recognition on the modeling requirements input by the user, thereby achieving accurate extraction of text information features and selection of corresponding modeling features. This process not only improves the efficiency of data processing, but also enhances the intelligence level of digital modeling, so that the model can better adapt to the specific needs of users, see Figure 1 As shown, the following steps are included:

[0048] S1: Obtain various forms of multi-level (including provincial, municipal, county and township) public security system business data through external data sources, which are used to query and warn of illegal behavior. Specifically, RESTful API and Web Scraper technology can be used to obtain data from external data sources, supporting file upload, API access and other public security system business data input methods to obtain the basic data required for digital modeling.

[0049] As a preferred implementation of this embodiment, the collected public security system business data includes but is not limited to: public security system audit log data, public security system application data, public security organization data and public security personnel information data.

[0050] The public security system audit log data includes but is not limited to: security audit logs, application audit logs, terminal behavior logs, etc.

[0051] The public security system application data includes but is not limited to: application system list data, application access log data, application database access data, application traffic data, etc.

[0052] The public security organization data includes but is not limited to: organization name, organization code, department type, position type, etc.

[0053] The public security personnel information data includes but is not limited to: personnel name, ID number, affiliated organization, personnel rank, personnel responsibilities, etc.

[0054] S2: Perform pre-processing operations such as data filtering, data recovery, data enrichment, and data serialization on the acquired public security system business data to form standard structured data.

[0055] As a preferred implementation of this embodiment, the multi-source data is processed through data management and control technology to make the data format unified, the classification and grading clear, and the identification clear after storage, so that it is suitable for subsequent modeling work. The preprocessing operation is specifically as follows:

[0056] Data filtering is to remove duplicate records, null values ​​and abnormal values ​​from the business data of the public security system, and to perform data cleaning and unified formatting. Data filtering is performed according to defined filtering conditions to ensure the quality of data use, and all business data of the public security system are adjusted to the same analysis level through unified formatting.

[0057] Data recovery is to construct data points for the business data of the public security system and form a data set, build a graph structure based on the data set, perform random path exploration on the constructed graph to generate several random paths, and calculate the recovery value of the missing data points through adaptive weighting based on the exploration results of the random paths, combined with the occurrence frequency and similarity of the random paths.

[0058] Specifically, the data recovery described in this embodiment is based on random path exploration in the data space, simulating the wandering process of missing data points, finding potential similarity relationships between data points by simulating multiple random paths, and generating the final recovery value through an adaptive dynamic weighting method.

[0059] Construct a data set based on the business data of the public security system ,in , As data points, the data set is regarded as a graph, the data points are nodes in the graph, and the similarity between nodes is the weight of the edge. Different similarity calculations are used for different types of data in the data set, and there is no restriction here. For example, dynamic time warping (DTW) or cosine similarity can be used for public security system audit log data; Jaccard similarity (application access log) or cosine similarity (application traffic data) can be used for public security system application data; category similarity measurement (such as Levenshtein distance) can be used for public security organization data; string similarity (such as Jaro-Winkler or Levenshtein distance) can be used for public security personnel information data.

[0060] Constructing the adjacency matrix of the graph To store the similarity information between all data points, the adjacency matrix elements The calculation is as follows:

[0061]

[0062] In the formula, For Node and nodes The similarity between .

[0063] Perform random path exploration on the constructed graph, starting from the missing data point, randomly select a neighboring node as the initial step, explore along the edge of the graph, and simulate the generation of the path; determine the probability of selecting the next node based on the similarity in the adjacency matrix, and use weighted random walks for exploration; repeat multiple random walks to generate several random paths. The probability calculation for selecting the next node is as follows:

[0064]

[0065] In the formula, For the node Select a node when starting The probability of For Node The neighbor set of node All connected nodes; ; For Node The sum of similarities of all neighbor nodes.

[0066] For each missing data point , set the number of walks to , calculate each neighbor node right The weight calculation is as follows:

[0067]

[0068] In the formula, For the The weights of the random paths for the missing data points; is the weight adjustment factor, which controls the speed of distance attenuation; is a node in a random path and The Euclidean distance between them measures the spatial distance between nodes; is the weight adjustment factor, which controls the influence of path frequency; For Node The frequency of occurrence of nodes in a random path, i.e. The number of times it appears in multiple walks; is the number of walks, used to normalize the frequency.

[0069] Missing data points The recovery value is its weighted average:

[0070]

[0071] In the formula, For missing data points The recovery value of is the total number of random paths; For the In a random path Similar nodes; For Node The weight of reflects the contribution of the node to the recovery value.

[0072] Data enrichment is to obtain the field names and field values ​​of the public security system business data from the fields in the public security system business data, obtain the field name and field value generation function of the target field according to the enrichment rules, and obtain the enriched public security system business data target field name and field value after calculation. Further, the enrichment rules extract other related field names and field value generation functions from the existing fields, calculate the target field name and field value according to the field name and field value generation function, and output the names and corresponding values ​​of these new fields in correspondence with the original data.

[0073] Specifically, taking the application audit log of the public security system as an example, the reference content after enrichment is as follows:

[0074]

[0075] Data serialization is to generate uniformly serialized public security system business data from the enriched public security system business data according to the target public security system business data profile, so as to complete the enrichment function of data rules.

[0076] S3: Users input text information of digital modeling requirements through a visual interface, and use a large language model to perform semantic recognition on the text information input by the user to obtain text information features. Specifically, an interactive user interface can be built through a front-end Web framework (such as Vue.js, React), allowing text information to be input through the user interface, and providing an API interface to allow third-party systems or applications to call the model through the API.

[0077] As a preferred implementation of this embodiment, a scenario question-answering method based on a large language model is used to select and reduce features, remove redundant features, and use a large language model to perform semantic recognition on the text information input by the user to obtain text information features, and extract key information in digital modeling requirements, such as modeling tasks, data features, target variables, and modeling goals, specifically:

[0078] Entity recognition: Identify specific noun phrases in text information through context, such as names of people, places, organizations, etc.

[0079] Relation extraction: identifying and parsing semantic relationships between entities in textual information.

[0080] Semantic role labeling: Identifying the relationship between a predicate and its associated noun phrases in a textual sentence.

[0081] Dependency analysis: Study the structural relationship between words in text information sentences, find out the dependency relationship between words through syntactic analysis and semantic analysis, and thus understand the semantics of text information sentences.

[0082] Semantic template matching: A template-based semantic analysis method that compares text information with predefined templates to understand the semantics of the text information.

[0083] Deep learning semantic representation: Use deep learning technology to learn the semantic representation of text information, and automatically learn the semantic information of text information by training a large corpus.

[0084] Specifically, the key links of the large language model in the method described in this embodiment corresponding to the large language model training and digital modeling are as follows:

[0085] The large language model is pre-trained based on a large amount of public security system business data to learn language patterns and rules, as well as the relevant knowledge of application system query and violation warning. Through learning, the large language model can understand the complexity and ambiguity of human language and provide accurate and relevant answers.

[0086] The large language model generates output based on the contextual information of the input, taking into account the context of the question and the question, improving the accuracy of question answering.

[0087] For certain specific tasks or problems, the preset data analysis models are fine-tuned to adapt to specific needs and environments.

[0088] S4: Select the corresponding digital modeling requirement information features according to the nature and task of the text information features, build a data analysis model in combination with standard structured data, and output the digital modeling results according to the data analysis model. Specifically, the large language model can be integrated with the data analysis platform (such as TensorFlow, PyTorch, Scikit-learn, etc.). The large language model can be responsible for understanding the user input, and pass the parsed modeling requirements to the data analysis platform through the API interface to perform specific modeling tasks, thereby being able to respond quickly to real-time streaming text data or new data, and perform incremental updates or real-time predictions of the digital analysis model. At the same time, questions are raised for modeling details that have not yet been clarified, and users improve the modeling requirements by answering. Based on the digital modeling learning features extracted and confirmed above, modeling parameters are automatically generated and model analysis is performed. After the digital analysis model analysis is completed, the digital modeling results are output, including digital analysis model performance indicators and prediction results.

[0089] As a preferred implementation of this embodiment, the data analysis model of the digital modeling specifically includes: an abnormal event analysis model, a statistical analysis model and a reverse tracing analysis model.

[0090] The abnormal event analysis model is specifically as follows: according to business needs, define different categories of abnormal event identification rules, use automated analysis algorithms to implement abnormal event analysis including abnormal accounts, unauthorized operations, high-frequency operations, sensitive data operations, large-scale export operations, abnormal time operations, and abnormal location operations, and support the establishment of customized abnormal analysis models through configuration policy rules. The automated analysis algorithms include baseline analysis, unsupervised clustering, and inter-domain migration.

[0091] The statistical analysis model is as follows: Define the statistical analysis dimensions: People: Count the behaviors of different people, analyze their operation frequency, permission usage, etc.; Organizations: Analyze the operation behaviors of different organizations and evaluate the security risks within the organization; Regions: Analyze the behavior based on the IP address and geographic location of the personnel to identify whether there are problems such as abnormal regional logins; Applications: Analyze the access volume, operation frequency, and potential risks of different applications; Operations: Count various types of operations (such as login, access, modification, deletion, etc.) to evaluate the security of operations. Through statistical methods (such as frequency statistics, trend analysis, mean and variance calculation, etc.), calculate the analysis indicators of each dimension (for example, calculate the operation frequency of a certain person, the frequency of use of a certain application, the security logs of different departments, etc.), conduct time series analysis and trend prediction on the business data of the public security system, and help identify potential problems.

[0092] The reverse tracing analysis model is specifically as follows: by defining known information and sensitive data in the business data of the public security system, constructing a personnel behavior graph, constructing elements such as personnel, equipment, and applications into a graph structure, and analyzing the behavioral associations and activity patterns between personnel. Clustering personnel behaviors (such as K-means clustering, etc.), mining potential associations, and finding key nodes in the impact chain. Using graph algorithms (such as Dijkstra algorithm, A* algorithm, etc.) for reverse reasoning, starting from known events or known information, trace back to potential source events based on the graph structure. Furthermore, the reverse tracing analysis model can also be combined with anomaly detection models to score behavioral links and identify potential abnormal events or violations.

[0093] Specifically, the elements such as personnel, equipment, and applications in the personnel behavior graph are defined as nodes. That is, each node may include: personnel node: identifying a specific person; equipment node: identifying the equipment used by the person; application node: identifying the application accessed by the person; sensitive data node: identifying the access to sensitive information, etc. The edges in the graph represent the relationship between nodes, and the attributes of the edges can represent the type, time, frequency, etc. of the interaction.

[0094] As a preferred implementation of this embodiment, the method also includes a large language model simultaneously generating modeling parameters and data analysis model performance indicators of modeling results, and the user adjusts the data analysis model parameters of digital modeling or triggers re-digital modeling operations through the large language model.

[0095] Here are some simple examples:

[0096] 1. Sample input text:

[0097] “Analyze the frequency of access to sensitive data by employees of all departments of a certain local public security system in the past three months, and list the departments and personnel with the most frequent access.”

[0098] 2. Extract text information features:

[0099] Task type: Statistical analysis of access frequency; Data source: sensitive data access records; Input features: department information, personnel information, access records, access time (past three months).

[0100] 3. Scenario Q&A and confirmation parameters:

[0101] The system asks: "Does the analysis data source include sensitive data access records of all departments in the X-Public Security System?"

[0102] User response: "Included."

[0103] 4. Generate the parameters required for digital modeling, start the data analysis task, and choose to establish a data analysis model.

[0104] 5. Output digital modeling results and issue warnings for violations.

[0105] Embodiment 2

[0106] Accordingly, this embodiment provides a digital modeling response system based on artificial intelligence, including a data acquisition module, a data preprocessing module, a user demand analysis module and a digital modeling module.

[0107] The data collection module is used to obtain various forms of multi-level public security system business data through external data sources. The public security system business data is used to query illegal behavior warnings. This module is used to implement the function of step S1 in Example 1 and will not be repeated here.

[0108] The data preprocessing module is used to perform preprocessing operations such as data filtering, data recovery, data enrichment and data serialization on the acquired public security system business data to form standard structured data. This module is used to implement the function of step S2 in Example 1 and will not be repeated here.

[0109] The user demand analysis module is used for users to input text information of digital modeling requirements through a visual interface, and uses a large language model to perform semantic recognition on the text information input by the user to obtain text information features. This module is used to implement the function of step S3 in Example 1 and will not be repeated here.

[0110] The digital modeling module is used to select corresponding digital modeling requirement information features according to the nature of the text information features and the task, build a data analysis model in combination with standard structured data, and output the digital modeling results according to the data analysis model. This module is used to implement the function of step S4 in Example 1 and will not be repeated here.

[0111] Embodiment 3

[0112] This embodiment provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the digital modeling response method based on artificial intelligence as described in any embodiment of the present invention is implemented.

[0113] Embodiment 4

[0114] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the digital modeling response method based on artificial intelligence as described in any embodiment of the present invention is implemented.

[0115] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0116] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0117] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0118] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0119] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A digital modeling response method based on artificial intelligence, characterized in that: The following steps are involved: S1: Obtain various forms of multi-level public security system business data through external data sources, and the public security system business data is used for early warning of illegal behaviors; S2: Perform pre-processing operations such as data filtering, data recovery, data enrichment, and data serialization on the acquired public security system business data to form standard structured data; S3: The user inputs the text information of the digital modeling requirements through the visual interface, and the large language model is used to perform semantic recognition on the text information input by the user to obtain the text information features; S4: Select the corresponding digital modeling requirement information features according to the nature of the text information features and the task, build a data analysis model based on the standard structured data, and output the digital modeling results according to the data analysis model; The data recovery is to construct data points for the business data of the public security system and form a data set, construct a graph structure based on the data set, regard the data set as a graph, the data points as nodes in the graph, and the similarity between the nodes as the weight of the edge; Construct an adjacency matrix of the graph to store the similarity information between all data points; perform random path exploration on the constructed graph, starting from the missing data point, randomly select a neighboring node as the initial step, explore along the edge of the graph, and simulate the generation of the path; determine the probability of selecting the next node based on the similarity in the adjacency matrix, and use weighted random walks for exploration; repeat multiple random walks to generate several random paths, and calculate the recovery value of the missing data point based on the exploration results of the random path. The recovery value is: In the formula, For missing data points The recovery value of is the total number of random paths; For the In a random path Similar nodes; For each missing data point , calculate each neighbor node right Weight: In the formula, For the Random paths to missing data points The weight of is the weight adjustment factor, which controls the speed of distance attenuation; is a node in a random path and The Euclidean distance between them measures the spatial distance between nodes; is the weight adjustment factor, which controls the influence of path frequency; For Node The frequency of occurrence of nodes in a random path, i.e. The number of times it appears in multiple walks; is the number of walks, used to normalize the frequency.

2. The digital modeling response method based on artificial intelligence according to claim 1 is characterized in that: The collected public security system business data includes but is not limited to: public security system audit log data, public security system application data, public security organization data and public security personnel information data; The public security system audit log data includes but is not limited to: security audit log, application audit log, terminal behavior log; The public security system application data includes but is not limited to: application system inventory data, application access log data, application database access data, and application traffic data; The public security organization data include but are not limited to: organization name, organization code, department type, position type; The public security personnel information data includes but is not limited to: personnel name, ID number, affiliated organization, personnel rank, and personnel responsibilities.

3. The digital modeling response method based on artificial intelligence according to claim 1 is characterized in that: The data filtering is to perform data cleaning and unified formatting of the public security system business data to remove duplicate records, null values ​​and abnormal values, and to filter the data according to the defined filtering conditions; The data enrichment is to obtain the field name and field value of the public security system business data from the fields in the public security system business data, obtain the field name generation function and field value generation function of the target field according to the enrichment rule, and obtain the enriched public security system business data target field name and target field value after calculation; The data serialization is to generate uniformly serialized public security system business data from the enriched public security system business data according to the target public security system business data profile.

4. The digital modeling response method based on artificial intelligence according to claim 1 is characterized in that: The semantic recognition of the text information input by the user is performed using a large language model to obtain the following text information features: Entity recognition: identifying specific noun phrases in text information through context; Relation extraction: identifying and parsing semantic relationships between entities in text information; Semantic role labeling: identifying the relationship between a predicate and its associated noun phrases in a text message sentence; Dependency analysis: Study the structural relationship between words in text information sentences, find out the dependency relationship between words through syntactic analysis and semantic analysis, and thus understand the semantics of text information sentences; Semantic template matching: A template-based semantic analysis method that compares text information with predefined templates to understand the semantics of text information; Deep learning semantic representation: Use deep learning technology to learn the semantic representation of text information, and automatically learn the semantic information of text information by training a large corpus.

5. The digital modeling response method based on artificial intelligence according to claim 1 is characterized in that: The data analysis model of the digital modeling specifically includes: an abnormal event analysis model, a statistical analysis model and a reverse tracing analysis model; The abnormal event analysis model is specifically: using automated analysis algorithms to implement abnormal event analysis including abnormal accounts, unauthorized operations, high-frequency operations, sensitive data operations, large-scale export operations, abnormal time operations, and abnormal location operations, and supports the establishment of customized abnormal analysis models through configuration policy rules; The statistical analysis model is as follows: define the statistical analysis dimensions: "people, organizations, regions, applications, and operations", calculate the analysis indicators of each dimension through statistical methods, and conduct time series analysis and trend forecasting on the business data of the public security system; The reverse tracing analysis model is specifically as follows: by defining known information and sensitive data in the business data of the public security system, constructing a personnel behavior graph, analyzing the behavioral associations and activity patterns between personnel, clustering the behaviors of personnel, mining potential associations, and performing reverse reasoning, the source events are traced and identified according to the structure of the personnel behavior graph; wherein the personnel behavior graph defines personnel, equipment, and applications as nodes; the edges in the personnel behavior graph represent the relationships between nodes, and the attributes of the edges include the type, time, and frequency of interactions.

6. The digital modeling response method based on artificial intelligence according to claim 1 is characterized in that: The method also includes the large language model simultaneously generating modeling parameters and data analysis model performance indicators of the modeling results, and the user adjusts the data analysis model parameters of the digital modeling or triggers the re-execution of the digital modeling operation through the large language model.

7. A digital modeling response system based on artificial intelligence, characterized in that: It includes data acquisition module, data preprocessing module, user demand analysis module and digital modeling module; A data collection module is used to obtain various forms of multi-level public security system business data through external data sources, and the public security system business data is used to issue early warnings for illegal behaviors; The data preprocessing module is used to perform preprocessing operations such as data filtering, data recovery, data enrichment and data serialization on the acquired public security system business data to form standard structured data; User demand analysis, which is used for users to input text information of digital modeling requirements through a visual interface, and use a large language model to perform semantic recognition on the text information input by users to obtain text information features; The digital modeling module is used to select the corresponding digital modeling requirement information features according to the nature of the text information features and the task, build a data analysis model in combination with standard structured data, and output the digital modeling results according to the data analysis model; The data recovery is to construct data points for the business data of the public security system and form a data set, construct a graph structure based on the data set, regard the data set as a graph, the data points as nodes in the graph, and the similarity between the nodes as the weight of the edge; Construct an adjacency matrix of the graph to store the similarity information between all data points; perform random path exploration on the constructed graph, starting from the missing data point, randomly select a neighboring node as the initial step, explore along the edge of the graph, and simulate the generation of the path; determine the probability of selecting the next node based on the similarity in the adjacency matrix, and use weighted random walks for exploration; repeat multiple random walks to generate several random paths, and calculate the recovery value of the missing data point based on the exploration results of the random path. The recovery value is: In the formula, For missing data points The recovery value of is the total number of random paths; For the In a random path Similar nodes; For each missing data point , calculate each neighbor node right Weight: In the formula, For the Random paths to missing data points The weight of is the weight adjustment factor, which controls the speed of distance attenuation; is a node in a random path and The Euclidean distance between them measures the spatial distance between nodes; is the weight adjustment factor, which controls the influence of path frequency; For Node The frequency of occurrence of nodes in a random path, i.e. The number of times it appears in multiple walks; is the number of walks, used to normalize the frequency.

8. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the artificial intelligence-based digital modeling response method as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the artificial intelligence-based digital modeling response method according to any one of claims 1 to 6 is implemented.

Images