A Predictive Analysis Method for Fire and Rescue Technologies Based on Personalized Centrality Calculation

Abstract

This invention proposes a predictive analysis method for fire rescue technologies based on personalized centrality calculation, comprising: S1, performing multi-source heterogeneous data processing on the source data of fire rescue information; S2, constructing a knowledge graph of fire rescue technologies and their related technologies based on the processed data obtained in step S1; S3, calculating the prediction score of each node in the knowledge graph of fire rescue technologies and their related technologies using a personalized centrality calculation method; and S4, judging the influence of fire rescue technologies and predicting their technological trends based on the centrality scores, thereby solving the problem of predictive analysis of fire rescue technologies.

Description

Technical Field

[0001] This invention belongs to the field of fire protection technology, and in particular relates to a fire rescue technology prediction and analysis method based on personalized centrality calculation. Background Technology

[0002] Firefighting and rescue technology has a wide range of applications in disaster and accident rescue. In recent years, with the rapid development of disciplines such as materials science, information technology, and automation control, disruptive technologies have emerged, ushering in a new wave of rapid iterative innovation in firefighting and rescue technology. Scientifically assessing the role of firefighting and rescue technology in disaster and accident rescue, and predicting technological development trends and directions, is of great significance for formulating scientific and reasonable development plans for firefighting and rescue technology and for carrying out innovative research and development in firefighting and rescue technology. Since numerous factors influence and determine the direction of firefighting and rescue technology development, effectively identifying and judging these factors from massive amounts of data on disaster and accident rescue needs, driving forces of technological development, and the influence of related disciplines, analyzing and assessing the main factors restricting technological innovation, and determining the impact of firefighting and rescue technology innovation on the development of each type of disaster and accident, is key to predicting the innovative development of firefighting and rescue technology.

[0003] Current forecasts for innovation and development in fire and rescue technologies primarily rely on specific variables. For example, for different types of disasters, these variables record the required fire and rescue forces and resources, the financial costs, the economic losses, the time required to resolve the disaster, the equipment used in rescue techniques, and the practical needs and problems encountered at the rescue site. However, even for disasters of the same nature, the recorded data can vary significantly. While some data describes the application of technical equipment, this data cannot be effectively utilized to identify which rescue technologies and equipment are most effective or what the core of these technologies is. Therefore, the true value of this data cannot be fully realized in forecasting innovation and development in fire and rescue technologies. Furthermore, technological development intelligence data from related disciplines in this field are currently also not effectively utilized to enrich forecasts for innovation and development in this technology. Summary of the Invention

[0004] This invention proposes a method and system for predictive analysis of fire rescue technology based on personalized centrality calculation. It uses knowledge graph technology to solve the problem of predictive analysis of fire rescue technology through centrality calculation.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0006] A predictive analysis method for fire rescue technology based on personalized centrality calculation, comprising:

[0007] S1. Perform multi-source heterogeneous data processing on the source data of fire and rescue information;

[0008] S2. Construct a knowledge graph of fire rescue technology and its related technologies based on the processed data obtained in step S1;

[0009] S3. Using the centrality calculation method, calculate the proximity centrality score of each node in the knowledge graph of fire rescue technology and its related technologies;

[0010] S4. Based on the proximity centrality score, determine the influence of fire rescue technology and predict its technological trends.

[0011] Furthermore, the multi-source heterogeneous data processing described in step S1 includes:

[0012] S101. Data Preprocessing: Cleaning and feature filtering of structured data in terms of data integrity, data quality, data redundancy, and data authenticity; evaluating whether unstructured data conforms to the corresponding fire and rescue technologies.

[0013] S102. Data Fusion: First, data fusion before preprocessing is carried out using a mesh structure for integrated processing. The core of the mesh structure is the name or number of the fire rescue technical solution. For unstructured data, its name is used as the node of the mesh structure, and the name corresponds to the storage location of the unstructured data. Second, data fusion after preprocessing is carried out by compressing the preprocessed data and statistically classifying the compressed data of the same fire rescue technical solution.

[0014] Furthermore, the method for constructing a knowledge graph of fire rescue technology and its related technologies as described in step S2 includes:

[0015] S201. Extract the triples from the name of the fire rescue technical solution as the lowest level structure. The triples include entity-relationship-entity or entity-attribute-entity.

[0016] S202. From the source data processed in step S1, extract keywords for entities and relationships using the keyword extraction method and connect them to the lowest-level triples; complete the design of the knowledge graph of fire rescue technology and its related technologies.

[0017] S203. Use a graph database to verify the rationality and performance of the fire rescue technology and its related technology knowledge graph.

[0018] Furthermore, the calculation process for the proximity centrality score in step S3 includes:

[0019] S301. Calculate the maximum and minimum degree of the knowledge graph of fire rescue technology and its related technologies, as well as the degree of each node; and standardize the degree of each node.

[0020] S302. Using a personalized centrality calculation method, the importance of each node in the knowledge graph of fire rescue technology and its related technologies is obtained;

[0021]

[0022] Where: j represents node j; α represents the probability of moving to the next node; in(j) represents the set of nodes pointing to j, and i represents a member of the set in(j); out(j) represents the set of nodes pointed to by j, |out(i)| represents the number of nodes pointed to by i, u represents the initial node, i.e., the type of disaster for which fire rescue technology needs to be recommended; PR(j) represents the importance of node j to the initial node u, i.e., the proximity centrality score of j to u.

[0023] Furthermore, step S4 specifically includes:

[0024] Taking the node of a specific disaster type as the central node, the node with the greatest importance from the central node is taken as the core technology for future fire rescue of that disaster type.

[0025] In another aspect, this invention proposes a fire rescue technology prediction and analysis system based on personalized centrality calculation, comprising:

[0026] Data processing module: performs multi-source heterogeneous data processing on the source data of fire and rescue information;

[0027] Graph Construction Module: Constructs a knowledge graph of fire rescue technology and its related technologies based on the processed data obtained from the data processing module;

[0028] Centrality calculation module: Using the centrality calculation method, calculate the proximity centrality score of each node in the knowledge graph of fire rescue technology and related technologies;

[0029] Discrimination and prediction module: Based on the proximity centrality score, it performs influence discrimination and technology trend prediction of fire rescue technology and its related technologies.

[0030] Furthermore, the data processing module includes:

[0031] The preprocessing unit is used for data preprocessing: cleaning and feature filtering of structured data in terms of data integrity, data quality, data redundancy, and data authenticity; and evaluating whether unstructured data conforms to the corresponding fire and rescue technologies.

[0032] The fusion unit is used for data fusion: First, fusion before data preprocessing is carried out using a mesh structure for integrated processing. The core of the mesh structure is the name or number of the fire rescue technical solution. For unstructured data, its name is used as the node of the mesh structure, and the name corresponds to the storage location of the unstructured data. Second, fusion after data preprocessing is carried out by compressing the preprocessed data and statistically classifying the compressed data of the same fire rescue technical solution.

[0033] Furthermore, the map construction module includes:

[0034] Lower-level unit: Extract the triples from the name of the fire rescue technical solution as the lowest-level structure. The triples include entity-relationship-entity or entity-attribute-entity.

[0035] Design Unit: Extract keywords for entities and relationships from the source data processed in step S1 using keyword extraction method, and connect them to the lowest-level triples; complete the design of the knowledge graph of fire rescue technology and its related technologies;

[0036] Verification Unit: Uses a graph database to verify the rationality and performance of the fire rescue technology and its related technology knowledge graph.

[0037] Furthermore, the centrality calculation module includes:

[0038] The degree calculation unit calculates the maximum and minimum degrees of the knowledge graph of fire rescue technology and its related technologies, as well as the degree of each node; and standardizes the degree of each node.

[0039] Centrality calculation unit: The importance of each node in the knowledge graph of fire rescue technology and its related technologies is obtained by using a personalized centrality calculation method;

[0040]

[0041] Where: j represents node j; α represents the probability of moving to the next node; in(j) represents the set of nodes pointing to j, and i represents a member of the set in(j); out(j) represents the set of nodes pointed to by j, |out(i)| represents the number of nodes pointed to by i, u represents the initial node, i.e., the type of disaster for which fire rescue technology needs to be recommended; PR(j) represents the importance of node j to the initial node u, i.e., the proximity centrality score of j to u.

[0042] Furthermore, the discrimination and prediction module includes:

[0043] Taking the node of a specific disaster type as the central node, the node with the greatest importance from the central node is taken as the core technology for future fire rescue of that disaster type.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) This invention uses knowledge graph technology to process knowledge and extract entities from fire and rescue information to construct a knowledge graph. It can integrate structured and unstructured data, extract entities and relationships using keyword extraction, and adopt a completely new data storage method in the industry.

[0046] (2) This invention uses centrality calculation in the field of fire rescue technology. Relying on the feature form of knowledge graph, it describes the influence and importance of fire rescue technology from the perspective of degree. Through centrality calculation, the data information of fire rescue technology can be flexibly utilized.

[0047] (3) This invention indirectly predicts the future development trend of fire rescue technology for different disaster types by calculating the proximity centrality of nodes, thereby accelerating technological innovation in the fire rescue industry;

[0048] (4) This invention can be widely applied to strategic research on fire and rescue technology, preparation of science and technology development plans, forecasting of fire and rescue technology development, formulation of technology roadmaps, and research and development of fire and rescue technology innovation. It can also serve the technological innovation and application of new technologies of enterprises and play an important guiding role in improving the independent innovation capability of fire and rescue technology. Attached Figure Description

[0049] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the knowledge graph of fire rescue technology and related technologies according to embodiments of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0052] The design concept of this invention is to predict and analyze fire rescue technology based on knowledge graphs.

[0053] Proposed in 2012, knowledge graphs represent the connections between knowledge elements in a knowledge-based way. Nodes in a knowledge graph represent entities in the real world, and the connections between nodes represent the relationships between entities in the real world. Essentially, it is a large semantic network, commonly used for knowledge representation, knowledge acquisition, knowledge fusion, knowledge reasoning, knowledge analysis, and knowledge retrieval. With the development of artificial intelligence, the application of knowledge graphs in the Internet is becoming increasingly widespread, with search and recommendation being the two most widely used fields.

[0054] Centrality is a commonly used concept in graph (network) analysis, used to express the degree to which a node in a graph (network) is located in the center of the entire network, also known as centrality. Depending on the method used to measure centrality, it can be divided into degree centrality, and based on the direction, it can be further divided into in-degree centrality, out-degree centrality, proximity centrality, and between-degree centrality.

[0055] Based on the above design concept, the fire rescue technology prediction and analysis method based on personalized centrality calculation proposed in this invention is as follows: Figure 1 As shown, it includes:

[0056] 1. Processing multi-source heterogeneous data:

[0057] Historical information on fire and rescue operations includes a wealth of structured and unstructured data. Structured data includes rescue time data, casualty data, and property damage data, while unstructured data includes fire and rescue technical documents, on-site images, and videos. This invention employs the following method to process multi-source heterogeneous data. First, data preprocessing involves cleaning and feature filtering of existing data to assess data integrity, quality, redundancy, and authenticity. Data integrity is determined by checking whether the collected technical solutions contain important information, such as the name of the rescue technology and the inventing organization. Incomplete basic information indicates data incompleteness. Data quality is assessed by checking whether certain characteristic values ​​of fire and rescue technologies are within reasonable ranges. For example, for small-scale rescue scenarios, the number of people rescued should be less than 100; if it exceeds 10,000, the data quality is considered unqualified. Data redundancy is determined by comparing new rescue technology solutions with existing ones; duplication indicates redundancy. Data authenticity is assessed by verifying the organization from which the rescue technology solutions originate. If the fire and rescue technology solutions originate from a legally authorized fire and rescue agency, they are considered qualified. If they originate from other organizations or individuals, authenticity needs to be verified by comparing the technology solutions with the application scenario. Unstructured data, such as images and videos, undergoes manual processing. Humans evaluate whether the images or videos conform to the corresponding fire and rescue technology solutions. If they do, the images and videos are stored in binary form in the computer. Second, data fusion. The types and sources of data related to fire and rescue are diverse, and the data are highly independent, making it impossible for them to communicate and interact. Data fusion can reduce the complexity of the data and enable communication between them. This fusion mainly includes two levels: pre-processing fusion, which involves fusing the raw data. The classification of disaster types is based on the classification method of the Fire and Rescue Bureau of the Ministry of Emergency Management, which divides them into six categories: fire, earthquake, geological disaster, flood, meteorological disaster, and other disasters. The integrated processing adopts a mesh structure, with the core of the mesh structure being the name (or number) of the fire rescue technical solution. Basic data, casualty information, and property loss information of the same disaster type are integrated and fused. The fusion process involves extracting entity semantics from its sub-data. For example, the entity "video" is extracted from video data and inserted into the core position of the original solution to expand the mesh structure. For instance, for a new mountain rescue technical solution, the solution technology is used as the core of the mesh structure. Its sub-data is classified according to the corresponding category. Since the solution belongs to geological disasters, geological disasters are used as new mesh nodes connected to the previous node. For its unstructured data, such as the corresponding demonstration video, the corresponding video name is used as a node to connect to the core node, and the name corresponds to the video storage location.After data preprocessing, the information is compressed. The essence of the network structure is cross-tabulation; different cross-tabulations correspond to different technical solutions. There is a lot of duplicate data in the cross-tabulations, such as the name of each rescue technical solution in each table. Tables with the same rescue technical solution can be merged. The network structure is represented in one attribute of the table, such as recording the node order in the corresponding table. Compressed data of the same type are statistically categorized. The purpose of categorization is to enable preliminary analysis of the collected data. The categorization is based on the fact that rescue technical methods belonging to different disaster types are classified as different categories. After multi-source heterogeneous data processing, it is convenient to perform operations such as panel presentation, graph transformation, and data analysis.

[0058] II. Atlas Structure Design and Construction:

[0059] Given the rapid iteration of information in rescue technology and the ever-changing rescue scenarios, existing entities and relationships are easily disrupted, and new entities and relationships are readily generated. This invention adopts a bottom-up approach to building a knowledge graph. The logical structure uses an entity-relationship-entity structure, with each triple constituting a semantic knowledge link. Entity-relationship-entity and entity-attribute-entity structures are extracted from each type of data source. The triples extracted from the solution name are used as the lowest-level structure, and triples are sequentially connected to the lowest-level triples. Keyword extraction is used to extract entity and relationship keywords from the original fire and rescue technology data. Document keywords are extracted using statistical information between documents; for example, keywords extracted from the most frequently repeated phrases or proper nouns can be used as entities or relationships. Existing automatic keyword extraction technologies may use synonyms to replace keywords, and these synonyms may not be the desired descriptive keywords. The extracted keywords may have different names than the original data. For such keywords, a manual screening method is used, with verbs designated as relationship keywords and nouns as entity keywords. For example, for a mountain rescue technical solution, the triple extracted from the name can be: Mountain rescue solution - includes - Mountain rescue solution for rain and snow weather.

[0060] For unstructured data, the deep learning algorithm CRF can be used to extract triples.

[0061] After the graph design was completed, it was deployed and debugged in the open-source graph database Neo4j to verify its rationality and performance. The main focus was on data ingestion efficiency, query accuracy, data integrity, query efficiency, and analysis performance. Graph performance was tested using database read / write performance benchmarks. In each test, CRUD operations were repeated five times, and the average of the five runs was taken. The database was restarted before each experiment to avoid the influence of previous experiments on the results. The read / write efficiency was compared with other storage structures. The fire rescue technology and related technology knowledge graph constructed in this embodiment is as follows: Figure 2 As shown.

[0062] III. Centrality Calculation Method: Centrality is used to identify the node in the knowledge graph that is most frequently associated with other nodes. First, the maximum and minimum degrees of the entire graph are calculated. Degree refers to the number of lines connecting the node to other nodes. Degree is divided into out-degree and in-degree. A simple search can obtain the maximum, minimum, and average degrees of the entire knowledge graph. Next, the centrality of each node is standardized by dividing its degree by the maximum degree in the graph. After standardization, centrality calculation is required. Centrality is divided into betweenness centrality and proximity centrality. This application aims to measure the influence and future technological trends of fire and rescue technologies, and adopts a personalized centrality calculation method. The specific explanation of personalized centrality calculation is as follows:

[0063] First, start from a node A and perform a random walk, moving to the next node with a probability of α, and returning to the initial node A with a probability of 1-α. After a sufficiently long time, the probability distribution of the entire graph will tend to a stable value, which is the importance of node A to each entity.

[0064] Assuming we start from A, initially PR(A) = 1, and the rest are 0;

[0065] In the first step, A moves to nodes a and c with a probability of 1 / 2 × α respectively:

[0066]

[0067] In the second step, A moves to a and c with a probability of 1 / 2 × α, a moves to A and B with a probability of 1 / 2 × α, and c moves to A, B, and C with a probability of 1 / 3 × α.

[0068]

[0069]

[0070]

[0071]

[0072] Where PR(A'), PR(a'), PR(c'), PR(B'), and PR(C') represent the importance of each node after the second step update;

[0073] Following this logic, we can derive the formula for the importance of each update step:

[0074]

[0075] Where: j represents node j; α represents the probability of moving to the next node; in(j) represents the set of nodes pointing to j, and i represents a member of the set in(j); out(j) represents the set of nodes pointed to by j, |out(i)| represents the number of nodes pointed to by i, u represents the initial node, i.e., the type of disaster for which fire rescue technology needs to be recommended; PR(j) represents the importance of node j to the initial node u, i.e., the proximity centrality score of j to u.

[0076] IV. Impact Assessment and Technological Trend Prediction:

[0077] After obtaining the influence of each node, we can obtain the specific information of the node closest to the central node. The central node of the application proposal is the specific disaster type. The fire rescue technology node closest to the specific central node is called the node with the greatest influence, that is, the fire rescue technology has a greater influence. The node closest to the central node and the node with the fastest progress in the constantly changing knowledge graph are called the future core technology of the disaster type.

[0078] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predictive analysis of fire rescue technology based on personalized centrality calculation, characterized in that, include: S1. Perform multi-source heterogeneous data processing on the source data of fire and rescue information; S2. Construct a knowledge graph of fire rescue technology and its related technologies based on the processed data obtained in step S1; S3. Using the centrality calculation method, calculate the proximity centrality score of each node in the knowledge graph of fire rescue technology and its related technologies; The calculation process for the proximity centrality score includes: S301. Calculate the maximum and minimum degree of the knowledge graph of fire rescue technology and its related technologies, as well as the degree of each node; and standardize the degree of each node. S302. Using a personalized centrality calculation method, the importance of each node in the knowledge graph of fire rescue technology and its related technologies is obtained; Where: j represents node j; α represents the probability of moving to the next node; in(j) represents the set of nodes pointing to j, and i represents a member of the set in(j); out(j) represents the set of nodes pointed to by j, |out(i)| represents the number of nodes pointed to by i, u represents the initial node, i.e., the type of disaster for which fire rescue technology needs to be recommended; PR(j) represents the importance of node j to the initial node u, i.e., the proximity centrality score of j to u; S4. Based on the proximity centrality score, determine the influence of fire rescue technology and predict its technological trends.

2. The fire rescue technology prediction and analysis method based on personalized centrality calculation according to claim 1, characterized in that, The multi-source heterogeneous data processing in step S1 includes: S101. Data Preprocessing: Cleaning and feature filtering of structured data in terms of data integrity, data quality, data redundancy, and data authenticity; evaluating whether unstructured data conforms to the corresponding fire and rescue technologies. S102. Data Fusion: First, data fusion before preprocessing is carried out using a mesh structure for integrated processing. The core of the mesh structure is the name or number of the fire rescue technical solution. For unstructured data, its name is used as the node of the mesh structure, and the name corresponds to the storage location of the unstructured data. Second, data fusion after preprocessing is carried out by compressing the preprocessed data and statistically classifying the compressed data of the same fire rescue technical solution.

3. The fire rescue technology prediction and analysis method based on personalized centrality calculation according to claim 1, characterized in that, The method for constructing a knowledge graph of fire rescue technology and its related technologies as described in step S2 includes: S201. Extract the triples from the name of the fire rescue technical solution as the lowest level structure. The triples include entity-relationship-entity or entity-attribute-entity. S202. From the source data processed in step S1, extract keywords for entities and relationships using the keyword extraction method and connect them to the lowest-level triples; complete the design of the knowledge graph of fire rescue technology and its related technologies. S203. Use a graph database to verify the rationality and performance of the fire rescue technology and its related technology knowledge graph.

4. The fire rescue technology prediction and analysis method based on personalized centrality calculation according to claim 1, characterized in that, Step S4 specifically includes: Taking the node of a specific disaster type as the central node, the node with the greatest importance from the central node is taken as the core technology for future fire rescue of that disaster type.

5. A fire rescue technology prediction and analysis system based on personalized centrality calculation, characterized in that, include: Data processing module: performs multi-source heterogeneous data processing on the source data of fire and rescue information; Graph Construction Module: Constructs a knowledge graph of fire rescue technology and its related technologies based on the processed data obtained from the data processing module; Centrality Calculation Module: This module uses a centrality calculation method to calculate the proximity centrality score of each node in the knowledge graph of fire rescue technology and its related technologies; including: The degree calculation unit calculates the maximum and minimum degrees of the knowledge graph of fire rescue technology and its related technologies, as well as the degree of each node; and standardizes the degree of each node. Centrality calculation unit: The importance of each node in the knowledge graph of fire rescue technology and its related technologies is obtained by using a personalized centrality calculation method; Where: j represents node j; α represents the probability of moving to the next node; in(j) represents the set of nodes pointing to j, and i represents a member of the set in(j); out(j) represents the set of nodes pointed to by j, |out(i)| represents the number of nodes pointed to by i, u represents the initial node, i.e., the type of disaster for which fire rescue technology needs to be recommended; PR(j) represents the importance of node j to the initial node u, i.e., the proximity centrality score of j to u; Discrimination and prediction module: Based on the proximity centrality score, the influence of fire rescue technology is discriminated and the technology trend is predicted.

6. The fire rescue technology prediction and analysis system based on personalized centrality calculation according to claim 5, characterized in that, The data processing module includes: The preprocessing unit is used for data preprocessing: cleaning and feature filtering of structured data in terms of data integrity, data quality, data redundancy, and data authenticity; and evaluating whether unstructured data conforms to the corresponding fire and rescue technologies. The fusion unit is used for data fusion: First, fusion before data preprocessing is carried out using a mesh structure for integrated processing. The core of the mesh structure is the name or number of the fire rescue technical solution. For unstructured data, its name is used as the node of the mesh structure, and the name corresponds to the storage location of the unstructured data. Second, fusion after data preprocessing is carried out by compressing the preprocessed data and statistically classifying the compressed data of the same fire rescue technical solution.

7. The fire rescue technology prediction and analysis system based on personalized centrality calculation according to claim 5, characterized in that, The map construction module includes: Lower-level unit: Extract the triples from the name of the fire rescue technical solution as the lowest-level structure. The triples include entity-relationship-entity or entity-attribute-entity. Design Unit: Extract keywords for entities and relationships from the source data processed in step S1 using keyword extraction method, and connect them to the lowest-level triples; complete the design of the knowledge graph of fire rescue technology and its related technologies; Verification Unit: Uses a graph database to verify the rationality and performance of the fire rescue technology and its related technology knowledge graph.

8. The fire rescue technology prediction and analysis system based on personalized centrality calculation according to claim 5, characterized in that, The discrimination and prediction module includes: Taking the node of a specific disaster type as the central node, the node with the greatest importance from the central node is taken as the core technology for future fire rescue of that disaster type.

Images