Methods for risk personnel assessment and early warning

Abstract

This invention belongs to the field of computer application technology, and specifically relates to a method for risk assessment and early warning. It achieves parallel data processing of multiple electronic data entries through Map and Reduce phases, enabling high-speed data cleaning and improving data processing efficiency. A personnel risk assessment and early warning analysis model based on knowledge graphs and NLP is constructed, introducing multimodal approaches into the field of personnel risk assessment and early warning, improving prediction accuracy. It achieves feature fusion of text and image data in electronic data, resulting in broad feature coverage and strong dynamic adaptability. The loss function of the proposed ViLBERT network considers both information integrity and temporal smoothness.

Description

Technical Field

[0001] This invention belongs to the field of computer application technology, and in particular relates to a method for risk assessment and early warning of personnel. Background Technology

[0002] With the widespread application of high-tech technologies such as the Internet and the Internet of Things in daily life and work, new types of high-risk individuals are emerging, and traditional risks are being given new meaning. A series of telecommunications risk cases are increasing, with increasingly sophisticated and covert methods, and longer durations, posing new challenges to public security. It is necessary to strengthen the prediction, early warning, and prevention of these frequent incidents, turning a passive approach into a proactive one.

[0003] By analyzing and mining big data, high-risk individuals can be predicted and their activity trends can be identified. This provides crucial decision-making support for police officers, enabling them to take timely and targeted measures to eliminate or reduce potential harm, thereby improving prevention and crackdown capabilities and maintaining social stability. Predicting and issuing early warnings about personnel risks is a key and core task in proactive policing. However, existing high-risk personnel management platforms and static integration models are no longer sufficient for practical needs due to their limited feature coverage and poor dynamic adaptability. Therefore, it is necessary to explore new early warning models. Summary of the Invention

[0004] This invention addresses the technical problems existing in personnel risk assessment and early warning by proposing a method for risk personnel assessment and early warning that is rationally designed, simple in method, theoretically sound, and capable of achieving broad feature coverage and strong dynamic adaptability.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: a method for risk personnel assessment and early warning, comprising the following steps:

[0006] S1. Obtain multiple pieces of electronic data information for evidence collection, including text and images, and upload the text and images in batches to the Hadoop Distributed File System for storage;

[0007] S2. Parallel data processing is achieved by processing multiple electronic data information through the Map stage and the Reduce stage.

[0008] S3. Construct a personnel risk assessment and early warning analysis model based on knowledge graphs and NLP. First, extract the regions of interest (ROIs) of the image using a pre-trained Mask R-CNN model. Encode the ROIs of both text and image. Then, use the ViLBERT network from NLP to model the cross-modal correlation of the ROIs of text and image, learn accurate multimodal representations, and achieve multimodal information fusion. The loss function of the ViLBERT network is defined as:

[0009]

[0010] Where w is the weight, α, β, γ are the balance parameters, n represents the number, and D i Represents the predicted score, ||w|| F It is the Frobenius norm of w. The time smoothness constraint is represented. Next, the neighbor node information of knowledge graph entities is aggregated through the Neo4j graph database, and then entity relationship reasoning is performed. The reasoning relationship is constructed using a scoring function of triples (h,r,t)∈G, where the loss function of the scoring function is:

[0011]

[0012] in, (h,r,t ′ ) is an invalid triplet obtained by randomly replacing one of the valid triplets, g(h,r,t) is the scoring function of the triplet, and σ() is the sigmoid function;

[0013] S4. Use ElasticSearch for key data retrieval and analysis;

[0014] S5. Implement personnel risk level assessment and early warning.

[0015] Preferably, the text and images in the electronic data information in step S1 include at least contact information, call log information, social text information, and social picture information.

[0016] Preferably, the data processing in step S2 includes the Map stage processing verifying multiple electronic data information, and the Reduce stage processing using the same reduce function to clean up duplicate data.

[0017] Preferably, the lower the fraction of the prime number g(h,r,t) in step S3, the more accurate the vector representation of the triple.

[0018] Preferably, the personnel risk level mentioned in step S5 includes Level 1, Level 2, Level 3, and General Risk.

[0019] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0020] This invention proposes a method for risk assessment and early warning of personnel. Multiple electronic data entries are processed in parallel through Map and Reduce phases, enabling high-speed data cleaning and improving processing efficiency. A personnel risk assessment and early warning analysis model based on knowledge graphs and NLP is constructed, introducing multimodal approaches to the field and improving prediction accuracy. It achieves feature fusion of text and images from electronic data entries, resulting in broad feature coverage and strong dynamic adaptability. The loss function of the proposed ViLBERT network considers both information integrity and temporal smoothness. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of a risk personnel assessment and early warning method provided in an embodiment of the present invention. Detailed Implementation

[0023] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0024] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0025] Additionally, it should be noted that all data acquisition actions in this application were carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with authorization from the owner of the relevant device.

[0026] In this example, considering the widespread application of high-tech technologies such as the Internet and the Internet of Things in daily life and work, new types of high-risk individuals are emerging, giving traditional risks new meaning. A series of telecommunications risk cases are increasing, with increasingly sophisticated and covert methods, and longer durations, posing new challenges to public security. It is necessary to strengthen the prediction, early warning, and prevention of these frequent incidents, turning a passive approach into a proactive one.

[0027] By analyzing and mining big data, high-risk individuals can be predicted and their activity trends identified. This provides crucial decision-making support for police officers, enabling timely and targeted measures to mitigate or reduce potential harm, thereby improving prevention and crackdown capabilities and maintaining social stability. Predicting and issuing early warnings about personnel risks is a key and core task in proactive policing. However, existing high-risk personnel management platforms and static integration models are no longer sufficient for practical needs due to their limited feature coverage and poor dynamic adaptability. Therefore, it is necessary to explore new early warning models. For example, Figure 1 As shown, this invention proposes a method for risk assessment and early warning. First, multiple pieces of electronic data information are obtained for evidence collection. The multiple pieces of electronic data information include text and images. The text and images include at least contact information, call record information, social text information, and social image information.

[0028] Given that Hadoop is a recognized industry standard open-source software for big data, a software framework capable of running in a distributed manner and providing massive data processing capabilities, and being developed based on the Java language, it possesses good cross-platform compatibility. Hadoop is designed to run on ordinary commercial computers, treating compute node failures as normal. The distributed system can automatically handle hardware and software anomalies, ensuring normal system operation through redundancy and backups. Data transfer is transparent to users, and the system provides transparent and reliable services to users at the upper layer. Therefore, batch uploading of text and images to the Hadoop Distributed File System storage is appropriate.

[0029] Considering the parallel data processing capabilities of Map / Reduce to preprocess raw data uploaded from various data source applications, extract commonalities, distinguish differences, filter and clean the data to generate platform-wide logs in a standard format, and then dump the data, this invention, compared to existing mechanisms, can both guarantee the personalized needs of each application in collecting user behavior and meet the needs of platform-wide global log analysis, while also enabling high-speed data cleaning and improving data processing efficiency. Therefore, multiple electronic data entries are processed in parallel through Map and Reduce phases. The Map phase verifies multiple electronic data entries, while the Reduce phase uses the same reduce function to clean up duplicate data.

[0030] Considering that knowledge graphs describe concepts, entities, and their relationships in the objective world in a structured form, expressing internet information in a form closer to human cognition, they provide a better ability to organize, manage, and understand the massive amounts of information on the internet. Natural Language Processing (NLP) is an important research direction in the field of artificial intelligence in recent years; it is a science integrating linguistics, computer science, and mathematics. Simply put, NLP technology mainly enables machines to understand naturally expressed human language. Multiple electronic data messages, including images and text, can be fused using their multimodal information to enrich the information of other entities. Therefore, before modeling the reasoning relationships between entities, a direct interactive method is needed to explicitly fuse multimodal information into the corresponding entities. To this end, this invention constructs a personnel risk assessment and early warning analysis model based on knowledge graphs and NLP. First, it extracts the regions of interest (ROIs) of images using a pre-trained Mask R-CNN model, encodes the ROIs of text and images, and uses the ViLBERT network in NLP to model the cross-modal correlation of the ROIs of text and images, learning accurate multimodal representations to achieve multimodal information fusion. Considering both information integrity and temporal smoothness, the loss function of the ViLBERT network is defined as:

[0031]

[0032] Where w is the weight, α, β, γ are the balance parameters, n represents the number, and D i Represents the predicted score, ||w|| F It is the Frobenius norm of w. The time smoothness constraint is represented. Next, the neighbor node information of knowledge graph entities is aggregated through the Neo4j graph database, and then entity relationship reasoning is performed. The reasoning relationship is constructed using a scoring function of triples (h,r,t)∈G. To represent entities and relationships as vectors, the loss function of the scoring function is:

[0033]

[0034] in, (h,r,t ′ ) represents an invalid triple obtained by randomly replacing one of the valid triples. g(h,r,t) is the scoring function for the triple; the lower the score of g(h,r,t), the more accurate the vector representation of the triple. σ() is the sigmoid function. Then, ElasticSearch is used for key data retrieval and string-parallel analysis. Finally, a personnel risk level assessment and early warning system is implemented, including Level 1, Level 2, Level 3, and General Risk.

[0035] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for risk assessment and early warning, characterized in that, Includes the following steps: S1. Obtain multiple pieces of electronic data information for evidence collection, including text and images, and upload the text and images in batches to the Hadoop Distributed File System for storage; S2. Parallel data processing is achieved by processing multiple electronic data information through the Map stage and the Reduce stage. S3. Construct a personnel risk assessment and early warning analysis model based on knowledge graphs and NLP. First, extract the regions of interest (ROIs) of the image using a pre-trained Mask R-CNN model. Encode the ROIs of both text and image. Then, use the ViLBERT network from NLP to model the cross-modal correlation of the ROIs of text and image, learn accurate multimodal representations, and achieve multimodal information fusion. The loss function of the ViLBERT network is defined as: Where w is the weight, α, β, γ are the balance parameters, n represents the number, and D i Represents the predicted score, ||w|| F It is the Frobenius norm of w. This represents a time smoothness constraint; next, the neighbor node information of knowledge graph entities is aggregated through the Neo4j graph database, and then entity relationship reasoning is performed. The reasoning relationship is constructed using a scoring function of triples (h, r, t) ∈ G, and the loss function of the scoring function is: in, It is an invalid triple obtained by randomly replacing one of the valid triples, g(h,r,t) is the scoring function of the triple, and σ() is the sigmoid function; S4. Use ElasticSearch for key data retrieval and analysis; S5. Implement personnel risk level assessment and early warning.

2. The method for risk assessment and early warning of personnel according to claim 1, characterized in that, The text and images in the electronic data information in step S1 include at least contact information, call log information, social text information, and social picture information.

3. The method for risk assessment and early warning according to claim 1, characterized in that, The data processing in step S2 includes the following: the Map phase verifies multiple electronic data entries, and the Reduce phase uses the same reduce function to clean up duplicate data.

4. The method for risk assessment and early warning according to claim 1, characterized in that, The lower the score of the prime number g(h, r, t) in step S3, the more accurate the vector representation of the triplet.

5. The method for risk assessment and early warning of personnel according to claim 1, characterized in that, The personnel risk levels described in step S5 include Level 1, Level 2, Level 3, and General Risk.

Images