CN119398525A - Urban safety risk assessment method and system based on digital standards - Google Patents

Urban safety risk assessment method and system based on digital standards Download PDF

Info

Publication number
CN119398525A
CN119398525A CN202411910534.8A CN202411910534A CN119398525A CN 119398525 A CN119398525 A CN 119398525A CN 202411910534 A CN202411910534 A CN 202411910534A CN 119398525 A CN119398525 A CN 119398525A
Authority
CN
China
Prior art keywords
risk
event
data
events
monitoring data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202411910534.8A
Other languages
Chinese (zh)
Inventor
张超
秦挺鑫
王皖
黄帅
徐凤娇
孟祥程
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China National Institute of Standardization
Original Assignee
China National Institute of Standardization
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China National Institute of Standardization filed Critical China National Institute of Standardization
Priority to CN202411910534.8A priority Critical patent/CN119398525A/en
Publication of CN119398525A publication Critical patent/CN119398525A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2337—Non-hierarchical techniques using fuzzy logic, i.e. fuzzy clustering
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/245—Classification techniques relating to the decision surface
    • G06F18/2451—Classification techniques relating to the decision surface linear, e.g. hyperplane
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/259—Fusion by voting
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/26—Discovering frequent patterns
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0499—Feedforward networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G06N5/022—Knowledge engineering; Knowledge acquisition
    • G06N5/025—Extracting rules from data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G06Q50/265—Personal security, identity or safety
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A30/00—Adapting or protecting infrastructure or their operation
    • Y02A30/60—Planning or developing urban green infrastructure

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Software Systems (AREA)
  • Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Tourism & Hospitality (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Computer Security & Cryptography (AREA)
  • Probability & Statistics with Applications (AREA)
  • Fuzzy Systems (AREA)
  • Primary Health Care (AREA)
  • Automation & Control Theory (AREA)
  • Physiology (AREA)
  • Medical Informatics (AREA)
  • Game Theory and Decision Science (AREA)

Abstract

本发明公开了基于数字化标准的城市安全风险评估方法及系统，包括获取城市安全风险监测数据和历史数据，根据所述历史数据构建风险指标事件库，对所述监测数据进行异常筛选和分类获得风险指标，匹配所述风险指标与所述风险指标事件库确定第一风险事件，根据所述第一风险事件预测不确定风险事件，根据所述第一风险事件、所述不确定风险事件和所述风险事件权重构建城市安全风险评估模型，将待评估城市安全风险监测数据和历史数据输入所述城市安全风险评估模型获得城市安全风险评估结果。该方法不仅可以提高城市安全风险评估的效率和准确性，同时具有较好的可解释性，可以直接应用于城市安全风险评估系统中。

The present invention discloses a city safety risk assessment method and system based on digital standards, including obtaining city safety risk monitoring data and historical data, constructing a risk indicator event library according to the historical data, performing abnormal screening and classification on the monitoring data to obtain risk indicators, matching the risk indicators with the risk indicator event library to determine a first risk event, predicting an uncertain risk event according to the first risk event, constructing a city safety risk assessment model according to the first risk event, the uncertain risk event and the risk event weight, and inputting the city safety risk monitoring data and historical data to be assessed into the city safety risk assessment model to obtain a city safety risk assessment result. The method can not only improve the efficiency and accuracy of city safety risk assessment, but also has good interpretability, and can be directly applied to the city safety risk assessment system.

Description

Urban security risk assessment method and system based on digital standard
Technical Field
The invention relates to the field of quality evaluation, in particular to an urban security risk evaluation method and system based on a digital standard.
Background
The continuous acceleration of the global urban process is accompanied by a series of challenges of safety risks in the development of scale and complexity, so that the management and evaluation of urban safety risks are of great significance for guaranteeing the life and property safety of urban residents, maintaining social stability and promoting sustainable development. The rapid development of information technology, in particular to the application of technologies such as big data, cloud computing, artificial intelligence and the like, provides a new opportunity for urban security risk assessment.
The traditional evaluation mode often depends on manual experience judgment, and the lack of data support and scientific analysis leads to inaccurate evaluation results and difficult real-time updating, and meanwhile, urban security risks relate to multiple fields and layers, and the factors are mutually interwoven and mutually influenced, so that risk evaluation work is more complex. Although some urban security risk assessment methods based on digital technology exist, most of these methods stay in the primary stage of data collection and analysis, and lack a systematic and comprehensive assessment model. Therefore, a systematic standard database is constructed through historical risk data, risk events are effectively preprocessed and identified on monitoring data, uncertain events caused by the risk events are predicted, an urban safety risk assessment model is constructed on the basis of a machine learning technology to score urban safety risks and predict risks and occurrence probability, and the urban safety risk assessment method and system based on the digital standard are designed to overcome the defects of the conventional assessment model, so that urban safety risk management is promoted to develop towards the digital and intelligent directions, and the method and system have important significance in constructing harmonious and safe urban environments.
Disclosure of Invention
The invention aims to provide a city security risk assessment method and system based on a digital standard.
In order to achieve the above purpose, the invention is implemented according to the following technical scheme:
the invention comprises the following steps:
Acquiring urban safety risk monitoring data and historical data, and preprocessing the monitoring data and the historical data;
constructing a risk index event library according to the historical data, and determining risk event weights;
Performing exception screening and classification on the monitoring data to obtain risk indexes, and matching the risk indexes with the risk index event library to determine a first risk event, and predicting an uncertain risk event according to the first risk event, wherein the uncertain risk event comprises a concurrent risk event and a probability risk event;
And constructing an urban security risk assessment model according to the first risk event, the uncertain risk event and the risk event weight, and inputting urban security risk monitoring data to be assessed and historical data into the urban security risk assessment model to obtain an urban security risk assessment result.
Further, the method for constructing a risk index event library according to the historical data and determining the risk event weight comprises the following steps:
determining risk events of different categories in the historical data by adopting a clustering method, extracting features of the risk events to obtain corresponding risk indexes, and constructing a risk index library according to the risk events and the corresponding risk indexes;
Dividing security risk levels according to risk categories, risk events and risk indexes, and determining the weight of each level evaluation unit by combining a level analysis method and an expert consultation method.
Further, the method for obtaining the risk index by carrying out anomaly screening and classification on the monitoring data comprises the following steps:
determining a safety risk item according to a regional risk assessment standard, acquiring safety risk item monitoring data, carrying out standardized processing on the monitoring data, substituting the monitoring data into a fuzzy clustering algorithm, dividing n sample space clusters according to the similarity among the monitoring data, and determining a clustering objective function, wherein the expression is as follows:
,
Wherein the method comprises the steps of For the clustering objective function, m is the class dimension of the samples within the cluster,The data samples are monitored for the j clusters,The data mean value is monitored for the j clusters,For the center of the cluster of samples in the i dimension,To monitor data samplesMembership belonging to dimension i;
Calculating the density value of a data object in a monitored data sample, selecting a data point with the maximum density value as a first clustering center point, and calculating the density values of other data points, wherein the expression is as follows:
,
,
Wherein the method comprises the steps of For the density value of the cluster center points,For density values of data points outside the cluster center point,In order to be a radius of the field,For the field radius to update the value,The density value of the initial maximum clustering center is obtained;
updating the cluster center and monitoring the membership degree of the data sample, and stopping iteration until the maximum iteration number or the objective function is met, so as to obtain K clusters with different data categories;
performing multi-hypersphere learning, performing macroscopic hypersphere learning according to all the monitoring data, and performing mesoscopic hypersphere learning on each cluster of monitoring data according to the cluster clustering result of the monitoring data, wherein the expression is as follows:
,
,
,
Wherein the method comprises the steps of For the macroscopic hypersphere learning loss function,For mesoscopic hypersphere learning loss functions,In order to compare the learning loss function,In order to be a node of the training set,In order to train the set of nodes,In order to train the number of nodes,Is a nodeIn the representation of the space in which the space is to be defined,Is the center of the macro-super-sphere,Is thatThe norm of the sample is calculated,In order for the parameters to be regularized,Is a nodeThe super sphere belongs to the center of the super sphere,Is a nodeThe index of the cluster to which it belongs,As a function of the cosine similarity,For the query vector of the kth sample,For the positive sample enhancement vector of the kth,Enhancing the vector set for all positive samples;
Determining a multi-sphere learning target according to the macroscopic hypersphere learning loss function, the mesoscopic hypersphere learning loss function and the contrast learning loss function, and performing anomaly judgment to obtain a risk index, wherein the expression is as follows:
,
,
wherein L is a multi-sphere learning object, In order to mesoscopic hypersphere learning parameters,In contrast to the parameters of the learning,Is a nodeIs an anomaly score for (2);
Labeling according to the data category and the data acquisition positioning as the risk index.
Further, a method of determining the first risk event includes:
Defining a plurality of risk indexes with probability to trigger all or a plurality of risk events, carrying out principal component analysis on the risk indexes to obtain principal component characteristics, wherein one principal component characteristic corresponds to one group of risk indexes, and determining the risk index group corresponding to the principal component characteristics by calculating the correlation coefficient of the principal component characteristics and the risk indexes, wherein the expression is as follows:
,
,
Wherein the method comprises the steps of Is the characteristic value of the ith principal componentAnd the j-th risk index valueIs used for the correlation coefficient of (a),For the mean value of the absolute values of the correlation coefficients of the ith principal component eigenvalue and all risk index values,For the mean value of all principal component eigenvalues,For all risk indicator value averages,As the characteristic quantity of the main component,As the number of risk indicators to be used,As a weight of the non-linear term,Is a nonlinear function for capturingAnd (3) withThe non-linear relationship is used to determine,For the regularization coefficient(s),Is regularized intensity;
Selection of The risk index of (a) forms the risk index group of the ith principal component feature, and defines the principal component feature vector set as,Set of risk event vectors as feature vector elements,The method comprises the steps of calculating the correlation between a plurality of risk index combinations and risk events in a risk index event library to determine a first risk event, wherein the first risk event is formed by a risk index threshold corresponding to an h-th risk event, s is the number of the risk events in the risk index event library, and the expression is as follows:
,
,
,
Wherein the method comprises the steps of For the distance of principal component feature i from risk event h,For the mean vector of the ith principal component feature sample,For the mean vector of the h-th risk event sample,For the covariance matrix of the ith principal component feature sample,A covariance matrix for the h-th risk event sample threshold,Is a matrix determinant;
setting a distance threshold When (when)Judging that the risk index group corresponding to the principal component characteristic i possibly causes a risk event h, otherwise, the risk index group corresponding to the principal component characteristic i does not cause a risk event;
labeling the first risk event according to the data category and the data collection positioning.
Further, a method for predicting a concurrent risk event according to the first risk event includes:
Constructing an FP-growth algorithm module, mining the relevance among risk events, determining concurrent risk events through known risk events, and constructing the FP-growth algorithm module by the following steps:
The frequency of the historical data statistics items is scanned, the minimum support degree is set to select frequent items, and an item header table is constructed by the frequent items and the frequency;
According to the item head list, items in the transaction are screened and ordered, and an FP-tree is constructed according to the screening result;
Determining a condition mode base and a transaction item set of frequent items according to the item head table, establishing a condition FP tree according to the condition mode base, training an FP-tree condition example by adopting historical data, exploring a common item set by adopting a deep excavation and recursion method, and outputting the FP-tree;
And inputting the first risk event into an FP-growth algorithm module to obtain a concurrent risk event, and labeling the concurrent risk event according to the data type and the data acquisition positioning.
Further, the method for predicting the probability risk event according to the first risk event comprises the following steps:
defining the feature vector of the risk index corresponding to the first risk event as Defining the feature vector of the risk index threshold corresponding to the risk index of the risk index event library target risk event asAnd calculating the comprehensive similarity of the first risk event feature vector and the target risk event feature vector, wherein the expression is as follows:
,
Wherein the method comprises the steps of For the f first risk event feature vectorAnd g target risk event feature vectorIs used for the combination of the similarity of the two,As the weight coefficient of the light-emitting diode,Is a feature vectorIs a set of risk indicators for a person,Feature vectorT is the number of risk indexes of all projects,For the maximum number of item risk indicators,Is a feature vectorIs used for the number of risk indicators of (a),Is a feature vectorIs set for the risk indicator threshold number of (a),、Is a feature vectorThe mean value and the variance of the risk index,、Is a feature vectorThe risk index threshold mean and variance, K is the vector dimension,Is a feature vectorIs used to determine the value of the element,Is a feature vectorIs used to determine the value of the element,Is a resolution coefficient;
setting a comprehensive similarity threshold, determining a probability risk event corresponding to the first risk event according to the comprehensive similarity, and labeling the probability risk event according to the data type and the data acquisition positioning.
Further, the method for obtaining the urban security risk assessment result comprises the following steps:
The method comprises the steps of forming a safety risk data set by a first risk event, a concurrent risk event, a probability risk event, a risk index and a risk event weight, and dividing the safety risk set into a training set and a testing set according to a ratio of 7:3 by adopting a random forest algorithm;
Constructing an urban security risk assessment model, wherein the urban security risk assessment model comprises a Bayesian network, a multi-mode prediction model and a strategy layer;
The Bayesian network estimates the probability relation of risk events according to the known first risk event, concurrent risk event and the historical data of probability risk event occurrence, and calculates the occurrence probability of each risk event given other risk events;
The multi-mode prediction model is used for learning training set data and risk event occurrence probability to carry out security risk assessment, and comprises a genetic-BP neural network base model, a support vector machine regression base model and a random forest base model;
The genetic-BP neural network is used for predicting risk scores according to input data, evaluating the difference between a predicted value and an actual value by adopting a mean square error loss function, optimizing network structures and parameters by adopting a genetic algorithm, and adjusting a learning rate by adopting an Adam optimizer;
The support vector machine regression finds out hyperplane through input data to distinguish events with different risk grades and conduct risk scoring prediction, and the support vector machine regression is adopted Insensitive loss function atIn-range prediction scoring error, adopting SMO algorithm to solve optimal hyperplane;
the random forest votes by constructing a decision tree to improve the accuracy and stability of risk scoring, the difference between a predicted value and an actual value is estimated by adopting a mean square error loss function, and the optimal splitting point is selected according to the information gain;
The strategy layer adopts a weighted voting method to integrate the base model to predict risk scores, the output layer extracts the category and the occurrence position of risk events, finally outputs a safety risk assessment result, and adopts a test set to assess the urban safety risk assessment model;
and inputting the urban security risk monitoring data to be evaluated and the historical data into an urban security risk evaluation model to obtain a security risk evaluation result.
In a second aspect, a digital standard-based urban security risk assessment system, comprising:
The data acquisition module is used for acquiring urban safety risk monitoring data and historical data and preprocessing the monitoring data and the historical data;
The data processing module is used for determining the weight of the risk event, carrying out abnormal screening and classification on the monitoring data to obtain a risk index, matching the risk index with a risk index event library and predicting an uncertain risk event;
The evaluation model module is used for constructing an urban safety risk evaluation model according to the first risk event, the uncertain risk event and the risk event weight, and inputting urban safety risk monitoring data to be evaluated and historical data into the urban safety risk evaluation model to obtain an urban safety risk evaluation result;
The intelligent supervision module is used for storing, checking and managing the monitoring data, the historical data and the urban security risk assessment result, determining security risk level according to the urban security risk assessment result, carrying out risk early warning and suggestion to a user, and marking risk events and categories, security risk level and occurrence probability thereof on a security risk map according to the position information.
The beneficial effects of the invention are as follows:
Compared with the prior art, the urban security risk assessment method and system based on the digital standard have the following technical effects:
According to the method, through the steps of determining the risk event weight, carrying out abnormal screening and classification on monitored data, carrying out data matching to determine a first risk event, predicting an uncertain risk event and constructing a model, the capability of data preprocessing and the model adaptability are improved in urban safety risk assessment, so that the efficiency and the accuracy of urban safety risk assessment are improved, the urban safety risk assessment technology is optimized, resources can be greatly saved, the working efficiency is improved, the assessment of urban safety risk can be realized, the safety risk condition of urban areas can be comprehensively, objectively and accurately obtained, accurate scientific guidance is provided for urban safety risk management and control, the method can adapt to the terminal assessment requirements of different urban safety risk assessment systems based on digital standards and urban safety risk assessment systems based on digital standards of different users, and has certain universality.
Drawings
FIG. 1 is a flowchart showing steps of a method for evaluating urban security risk based on digital standards according to the present invention.
Detailed Description
The invention is further described by the following specific examples, which are presented to illustrate, but not to limit, the invention.
The invention relates to urban safety based on digital standard the risk assessment method and system comprise the following steps:
as shown in fig. 1, in this embodiment, the steps include:
Acquiring urban safety risk monitoring data and historical data, and preprocessing the monitoring data and the historical data;
constructing a risk index event library according to the historical data, and determining risk event weights;
Performing exception screening and classification on the monitoring data to obtain risk indexes, and matching the risk indexes with the risk index event library to determine a first risk event, and predicting an uncertain risk event according to the first risk event, wherein the uncertain risk event comprises a concurrent risk event and a probability risk event;
And constructing an urban security risk assessment model according to the first risk event, the uncertain risk event and the risk event weight, and inputting urban security risk monitoring data to be assessed and historical data into the urban security risk assessment model to obtain an urban security risk assessment result.
In this embodiment, the method for constructing a risk indicator event library according to the historical data and determining the risk event weight includes:
determining risk events of different categories in the historical data by adopting a clustering method, extracting features of the risk events to obtain corresponding risk indexes, and constructing a risk index library according to the risk events and the corresponding risk indexes;
Dividing security risk levels according to risk categories, risk events and risk indexes, and determining weights of all levels of evaluation units by combining a level analysis method and an expert consultation method;
In actual evaluation, the security risk level is divided according to the risk category, the risk event and the risk index to obtain a first-level index risk category and a corresponding weight, namely natural disasters of-0.3 (earthquake, storm, typhoon, high Wen Tianqi), accidents of-0.25 (chemical accidents, firework and firecracker accidents, civil explosive accidents, industrial and trade accidents, transportation accidents and building construction accidents), public health of-0.2 (infectious disease outbreaks, food safety events and occupational disease hazards), public safety of-0.15 (crime activities and other safety events), infrastructure of-0.1 (old house collapse, large passenger flows, building fires, urban elevator operation, transportation, electric power operation, gas leakage explosion, bridge operation and water supply and drainage pipe network faults).
In this embodiment, the method for obtaining risk indexes by performing anomaly screening and classification on the monitored data includes:
determining a safety risk item according to a regional risk assessment standard, acquiring safety risk item monitoring data, carrying out standardized processing on the monitoring data, substituting the monitoring data into a fuzzy clustering algorithm, dividing n sample space clusters according to the similarity among the monitoring data, and determining a clustering objective function, wherein the expression is as follows:
,
Wherein the method comprises the steps of For the clustering objective function, m is the class dimension of the samples within the cluster,The data samples are monitored for the j clusters,The data mean value is monitored for the j clusters,For the center of the cluster of samples in the i dimension,To monitor data samplesMembership belonging to dimension i;
Calculating the density value of a data object in a monitored data sample, selecting a data point with the maximum density value as a first clustering center point, and calculating the density values of other data points, wherein the expression is as follows:
,
,
Wherein the method comprises the steps of For the density value of the cluster center points,For density values of data points outside the cluster center point,In order to be a radius of the field,For the field radius to update the value,The density value of the initial maximum clustering center is obtained;
updating the cluster center and monitoring the membership degree of the data sample, and stopping iteration until the maximum iteration number or the objective function is met, so as to obtain K clusters with different data categories;
performing multi-hypersphere learning, performing macroscopic hypersphere learning according to all the monitoring data, and performing mesoscopic hypersphere learning on each cluster of monitoring data according to the cluster clustering result of the monitoring data, wherein the expression is as follows:
,
,
,
Wherein the method comprises the steps of For the macroscopic hypersphere learning loss function,For mesoscopic hypersphere learning loss functions,In order to compare the learning loss function,In order to be a node of the training set,In order to train the set of nodes,In order to train the number of nodes,Is a nodeIn the representation of the space in which the space is to be defined,Is the center of the macro-super-sphere,Is thatThe norm of the sample is calculated,In order for the parameters to be regularized,Is a nodeThe super sphere belongs to the center of the super sphere,Is a nodeThe index of the cluster to which it belongs,As a function of the cosine similarity,For the query vector of the kth sample,For the positive sample enhancement vector of the kth,Enhancing the vector set for all positive samples;
Determining a multi-sphere learning target according to the macroscopic hypersphere learning loss function, the mesoscopic hypersphere learning loss function and the contrast learning loss function, and performing anomaly judgment to obtain a risk index, wherein the expression is as follows:
,
,
wherein L is a multi-sphere learning object, In order to mesoscopic hypersphere learning parameters,In contrast to the parameters of the learning,Is a nodeIs an anomaly score for (2);
labeling according to the data category and the data acquisition positioning as risk indexes;
In actual evaluation, anomaly screening and classification are carried out on safety risk monitoring data of a certain city to obtain risk indexes, namely natural disasters (daily rainfall of 250mm, riverbed height of +6 meters, 15 waterlogged areas, waterlogged water depth of 0.3m and social media of 10000 times per hour about the discussion of storm), accidents (industrial area chemical leakage of 1), public An Quanlei (public transportation faults of 2, theft events of 5 and fight events of 1), infrastructures (traffic congestion index of 8.5, emergency service telephone number of 1500 times, power failure report of 40, increase of mobile network flow rate by 50 times at ordinary times, increase of network attack rate by 30 percent at ordinary times and public facility damage of 1).
In this embodiment, the method for determining the first risk event includes:
Defining a plurality of risk indexes with probability to trigger all or a plurality of risk events, carrying out principal component analysis on the risk indexes to obtain principal component characteristics, wherein one principal component characteristic corresponds to one group of risk indexes, and determining the risk index group corresponding to the principal component characteristics by calculating the correlation coefficient of the principal component characteristics and the risk indexes, wherein the expression is as follows:
,
,
Wherein the method comprises the steps of Is the characteristic value of the ith principal componentAnd the j-th risk index valueIs used for the correlation coefficient of (a),For the mean value of the absolute values of the correlation coefficients of the ith principal component eigenvalue and all risk index values,For the mean value of all principal component eigenvalues,For all risk indicator value averages,As the characteristic quantity of the main component,As the number of risk indicators to be used,As a weight of the non-linear term,Is a nonlinear function for capturingAnd (3) withThe non-linear relationship is used to determine,For the regularization coefficient(s),Is regularized intensity;
Selection of The risk index of (a) forms the risk index group of the ith principal component feature, and defines the principal component feature vector set as,Set of risk event vectors as feature vector elements,The method comprises the steps of calculating the correlation between a plurality of risk index combinations and risk events in a risk index event library to determine a first risk event, wherein the first risk event is formed by a risk index threshold corresponding to an h-th risk event, s is the number of the risk events in the risk index event library, and the expression is as follows:
,
,
,
Wherein the method comprises the steps of For the distance of principal component feature i from risk event h,For the mean vector of the ith principal component feature sample,For the mean vector of the h-th risk event sample,For the covariance matrix of the ith principal component feature sample,A covariance matrix for the h-th risk event sample threshold,Is a matrix determinant;
setting a distance threshold When (when)Judging that the risk index group corresponding to the principal component characteristic i possibly causes a risk event h, otherwise, the risk index group corresponding to the principal component characteristic i does not cause a risk event;
Labeling a first risk event according to the data type and the data acquisition positioning;
In actual evaluation, two risk index groups (daily rainfall, riverbed height, the number of waterlogged areas, waterlogged water depth and the discussion number of social media about storm) are determined by calculating the correlation of the risk indexes, wherein the correlation coefficient is 0.88> and the absolute value average value of the correlation coefficient is 0.62), (the number of waterlogged areas, public transportation faults, traffic congestion indexes, power fault reports and the average increase of the mobile network flow ratio are 0.75 >;
And determining that the first risk event is heavy rain extreme weather and traffic jam according to the distance (0.08, 0.15 and the distance threshold value of 0.2) between the correlation principal component characteristics and the risk event.
In this embodiment, the method for predicting a concurrent risk event according to the first risk event includes:
Constructing an FP-growth algorithm module, mining the relevance among risk events, determining concurrent risk events through known risk events, and constructing the FP-growth algorithm module by the following steps:
The frequency of the historical data statistics items is scanned, the minimum support degree is set to select frequent items, and an item header table is constructed by the frequent items and the frequency;
According to the item head list, items in the transaction are screened and ordered, and an FP-tree is constructed according to the screening result;
Determining a condition mode base and a transaction item set of frequent items according to the item head table, establishing a condition FP tree according to the condition mode base, training an FP-tree condition example by adopting historical data, exploring a common item set by adopting a deep excavation and recursion method, and outputting the FP-tree;
Inputting the first risk event into an FP-growth algorithm module to obtain a concurrent risk event, and labeling the concurrent risk event according to the data type and data acquisition positioning;
in actual evaluation, inputting risk indexes corresponding to heavy rain extreme weather and traffic jam into an FP-growth algorithm module, and determining that the concurrent risk event is public traffic delay and drainage system obstacle.
In this embodiment, the method for predicting a probability risk event according to the first risk event includes:
defining the feature vector of the risk index corresponding to the first risk event as Defining the feature vector of the risk index threshold corresponding to the risk index of the risk index event library target risk event asAnd calculating the comprehensive similarity of the first risk event feature vector and the target risk event feature vector, wherein the expression is as follows:
,
Wherein the method comprises the steps of For the f first risk event feature vectorAnd g target risk event feature vectorIs used for the combination of the similarity of the two,As the weight coefficient of the light-emitting diode,Is a feature vectorIs a set of risk indicators for a person,Feature vectorT is the number of risk indexes of all projects,For the maximum number of item risk indicators,Is a feature vectorIs used for the number of risk indicators of (a),Is a feature vectorIs set for the risk indicator threshold number of (a),、Is a feature vectorThe mean value and the variance of the risk index,、Is a feature vectorThe risk index threshold mean and variance, K is the vector dimension,Is a feature vectorIs used to determine the value of the element,Is a feature vectorIs used to determine the value of the element,Is a resolution coefficient;
setting a comprehensive similarity threshold, determining a probability risk event corresponding to the first risk event according to the comprehensive similarity, and labeling the probability risk event according to the data type and the data acquisition positioning;
In actual evaluation, setting the comprehensive similarity threshold to be 0.6, calculating the comprehensive similarity of the feature vector of the risk index corresponding to the first risk event and the feature vector of the risk index threshold corresponding to the risk event of the risk index event library target, and determining the probability risk event such as public equipment damage, house damage, underground facility water accumulation, public health problem, public activity cancellation, public service interruption, network communication damage, agricultural economic loss and supply chain interruption according to the comprehensive similarity threshold.
In this embodiment, the method for obtaining the urban security risk assessment result includes:
The method comprises the steps of forming a safety risk data set by a first risk event, a concurrent risk event, a probability risk event, a risk index and a risk event weight, and dividing the safety risk set into a training set and a testing set according to a ratio of 7:3 by adopting a random forest algorithm;
Constructing an urban security risk assessment model, wherein the urban security risk assessment model comprises a Bayesian network, a multi-mode prediction model and a strategy layer;
The Bayesian network estimates the probability relation of risk events according to the known first risk event, concurrent risk event and the historical data of probability risk event occurrence, and calculates the occurrence probability of each risk event given other risk events;
The multi-mode prediction model is used for learning training set data and risk event occurrence probability to carry out security risk assessment, and comprises a genetic-BP neural network base model, a support vector machine regression base model and a random forest base model;
The genetic-BP neural network is used for predicting risk scores according to input data, evaluating the difference between a predicted value and an actual value by adopting a mean square error loss function, optimizing network structures and parameters by adopting a genetic algorithm, and adjusting a learning rate by adopting an Adam optimizer;
The support vector machine regression finds out hyperplane through input data to distinguish events with different risk grades and conduct risk scoring prediction, and the support vector machine regression is adopted Insensitive loss function atIn-range prediction scoring error, adopting SMO algorithm to solve optimal hyperplane;
the random forest votes by constructing a decision tree to improve the accuracy and stability of risk scoring, the difference between a predicted value and an actual value is estimated by adopting a mean square error loss function, and the optimal splitting point is selected according to the information gain;
The method comprises the steps that a policy layer adopts a weighted voting method to integrate a base model to predict risk scores, an output layer extracts categories and occurrence positions of risk events, and finally outputs a safety risk assessment result, and a test set is adopted to assess a city safety risk assessment model;
inputting urban security risk monitoring data and historical data to be evaluated into an urban security risk evaluation model to obtain a security risk evaluation result;
In the actual evaluation, calculating the occurrence probability of the risk event according to the risk event, the risk index and the corresponding category weight by adopting a Bayesian network, wherein the risk event is 90% of extreme weather of heavy rain and 80% of traffic jam, (concurrent risk event is 70% of public transportation delay and 60% of drainage system obstruction), and the risk event is (uncertain risk event is 30% of public equipment damage, 20% of house damage, 25% of underground facility ponding, 15% of public health problem, 10% of public activity cancellation, 5% of public service interruption, 10% of network communication damage, 20% of agricultural economic loss and 15% of supply chain interruption;
The urban safety risk monitoring data and the historical data to be evaluated are input into an urban safety risk evaluation model to obtain an urban safety risk evaluation result, wherein the urban safety risk score is 47 points, and the urban safety risk is a public traffic jam/delay event caused by an extreme storm weather event, belongs to natural disaster type and accident disaster type risk events, and is accompanied by accident disaster type, public health type and infrastructure type risk events.
In a second aspect, a digital standard-based urban security risk assessment system, comprising:
The data acquisition module is used for acquiring urban safety risk monitoring data and historical data and preprocessing the monitoring data and the historical data;
The data processing module is used for determining the weight of the risk event, carrying out abnormal screening and classification on the monitoring data to obtain a risk index, matching the risk index with a risk index event library and predicting an uncertain risk event;
The evaluation model module is used for constructing an urban safety risk evaluation model according to the first risk event, the uncertain risk event and the risk event weight, and inputting urban safety risk monitoring data to be evaluated and historical data into the urban safety risk evaluation model to obtain an urban safety risk evaluation result;
The intelligent supervision module is used for storing, checking and managing the monitoring data, the historical data and the urban security risk assessment result, determining a security risk level according to the urban security risk assessment result, carrying out risk early warning and suggestion on a user, and marking a risk event and category, the security risk level and occurrence probability thereof on a security risk map according to the position information;
In actual evaluation, importing an urban security risk evaluation result through a supervision module, marking the occurrence position of a risk index in a security risk map, marking a corresponding risk event, a risk event category, a risk event probability and an urban security risk score, judging that the urban security risk score is 47 and is classified into medium risk through a risk threshold, and sending early warning information and an answer suggestion to a user side.
The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims (8)

1.基于数字化标准的城市安全风险评估方法，其特征在于，包括以下步骤：1. A method for assessing urban safety risks based on digital standards, characterized in that it comprises the following steps: S1、获取城市安全风险监测数据和历史数据，对所述监测数据和所述历史数据进行预处理；S1. Acquire urban safety risk monitoring data and historical data, and pre-process the monitoring data and historical data; S2、根据所述历史数据构建风险指标事件库，确定风险事件权重；S2. Build a risk indicator event library based on the historical data and determine the risk event weights; S3、对所述监测数据进行异常筛选和分类获得风险指标，将所述风险指标匹配所述风险指标事件库确定第一风险事件，根据所述第一风险事件预测不确定风险事件；所述不确定风险事件包括并发风险事件和概率风险事件；S3, performing abnormal screening and classification on the monitoring data to obtain risk indicators, matching the risk indicators with the risk indicator event library to determine a first risk event, and predicting an uncertain risk event based on the first risk event; the uncertain risk event includes a concurrent risk event and a probabilistic risk event; S4、根据所述第一风险事件、所述不确定风险事件和所述风险事件权重构建城市安全风险评估模型，将待评估城市安全风险监测数据和历史数据输入所述城市安全风险评估模型获得城市安全风险评估结果。S4. Construct a city safety risk assessment model based on the first risk event, the uncertain risk event and the risk event weight, and input the city safety risk monitoring data and historical data to be assessed into the city safety risk assessment model to obtain a city safety risk assessment result. 2.根据权利要求1所述基于数字化标准的城市安全风险评估方法，其特征在于，根据所述历史数据构建风险指标事件库并确定所述风险事件权重的方法，包括：2. The urban safety risk assessment method based on digital standards according to claim 1 is characterized in that the method of constructing a risk indicator event library based on the historical data and determining the risk event weights includes: 采用聚类法确定历史数据中不同类别的风险事件，对风险事件进行特征提取获得对应风险指标，根据风险事件和对应风险指标构建风险指标库；Use clustering method to identify risk events of different categories in historical data, extract features of risk events to obtain corresponding risk indicators, and build a risk indicator library based on risk events and corresponding risk indicators; 按照风险类别、风险事件和风险指标划分安全风险层次，结合层次分析法和专家咨询法确定各层次评估单元的权重。Divide the safety risk levels according to risk categories, risk events and risk indicators, and determine the weights of the assessment units at each level by combining the analytic hierarchy process and expert consultation method. 3.根据权利要求1所述基于数字化标准的城市安全风险评估方法，其特征在于，对所述监测数据进行异常筛选和分类获得风险指标的方法，包括：3. The urban safety risk assessment method based on digital standards according to claim 1 is characterized in that the method of obtaining risk indicators by performing abnormal screening and classification on the monitoring data comprises: 按照区域风险评估标准确定安全风险项目，获取安全风险项目监测数据，标准化处理监测数据并代入模糊分簇聚类算法中，根据监测数据间相似度划分n个样本空间分簇，并确定分簇目标函数，表达式为：Determine the safety risk projects according to the regional risk assessment standards, obtain the monitoring data of the safety risk projects, standardize the monitoring data and substitute it into the fuzzy clustering algorithm, divide the n sample space into clusters according to the similarity between the monitoring data, and determine the clustering objective function, which is expressed as: ， , 其中为分簇目标函数，m为簇内样本的类别维度，为j簇监测数据样本，为j簇监测数据均值，为样本在i维度的聚类中心，为监测数据样本属于i维度的隶属度；in is the clustering objective function, m is the category dimension of the samples in the cluster, is the monitoring data sample of cluster j , is the mean value of the monitoring data of cluster j , is the cluster center of the sample in dimension i , For monitoring data samples The degree of membership to dimension i ; 计算监测数据样本中数据对象的密度值，选取密度值最大的数据点作为第一个簇聚类中心点，并计算其余数据点的密度值，表达式为：Calculate the density value of the data object in the monitoring data sample, select the data point with the largest density value as the first cluster center point, and calculate the density value of the remaining data points. The expression is: ， , ， , 其中为聚类中心点的密度值，为聚类中心点以外数据点的密度值，为领域半径，为领域半径更新值，为初始最大聚类中心的密度值；in is the density value of the cluster center point, is the density value of data points outside the cluster center point, is the area radius, Update the value for the area radius, is the density value of the initial maximum cluster center; 更新聚类中心和监测数据样本的隶属度，一直迭代直到最大迭代次数或满足目标函数时停止迭代，获得分为数据类别不同的K个簇；Update the cluster center and the membership of the monitoring data sample, and iterate until the maximum number of iterations is reached or the objective function is satisfied, and then obtain K clusters with different data categories. 进行多超球面学习，根据所有监测数据进行宏观超球面学习，根据监测数据分簇聚类结果对每个簇监测数据进行介观超球面学习，表达式为：Perform multi-hypersphere learning, perform macroscopic hypersphere learning based on all monitoring data, and perform mesoscopic hypersphere learning on each cluster monitoring data based on the monitoring data clustering results. The expression is: ， , ， , ， , 其中为宏观超球面学习损失函数，为介观超球面学习损失函数，为对比学习损失函数，为训练集的节点，为训练节点集合，为训练节点数量，为节点在空间的表征，为宏观超球体中心，为范数，为正则化参数，为节点所属超球体中心，为节点所属簇索引，为余弦相似度函数，为第k个样本的查询向量，为第k个正样本增强向量，为所有正样本增强向量集合；in is the loss function for learning the macroscopic hypersphere, is the mesoscopic hypersphere learning loss function, is the contrastive learning loss function, is the node of the training set, is the set of training nodes, is the number of training nodes, For Node Representation in space, is the center of the macroscopic hypersphere, for norm, is the regularization parameter, For Node The center of the hypersphere, For Node The cluster index, is the cosine similarity function, is the query vector of the kth sample, is the kth positive sample enhancement vector, Enhance the vector set for all positive samples; 根据宏观超球面学习损失函数、介观超球面学习损失函数和对比学习损失函数确定多球体学习目标，进行异常判定获得风险指标，表达式为：The multi-sphere learning target is determined according to the macroscopic hypersphere learning loss function, the mesoscopic hypersphere learning loss function and the contrastive learning loss function, and the risk index is obtained by abnormal judgment. The expression is: ， , ， , 其中L为多球体学习目标，为介观超球面学习参数，对比学习参数，为节点的异常评分；Where L is the multi-sphere learning target, Learn the parameters for the mesoscopic hypersphere, Compare the learning parameters, For Node Abnormal score of 按数据类别和数据采集定位为风险指标贴标签。Label risk indicators by data category and data collection location. 4.根据权利要求1所述基于数字化标准的城市安全风险评估方法，其特征在于，确定所述第一风险事件的方法，包括：4. The urban safety risk assessment method based on digital standards according to claim 1, characterized in that the method for determining the first risk event comprises: 定义多个风险指标有概率引发一起或多起风险事件，对风险指标进行主成分分析获得主成分特征，一个主成分特征对应一组风险指标，通过计算主成分特征与风险指标相关系数确定该主成分特征对应的风险指标组，表达式为：It is defined that multiple risk indicators have the probability of causing one or more risk events. The principal component analysis is performed on the risk indicators to obtain the principal component characteristics. One principal component characteristic corresponds to a group of risk indicators. The risk indicator group corresponding to the principal component characteristic is determined by calculating the correlation coefficient between the principal component characteristic and the risk indicator. The expression is: ， , ， , 其中为第i个主成分特征值与第j个风险指标值的相关系数，为第i个主成分特征值与所有风险指标值的相关系数绝对值均值，为所有主成分特征值均值，为所有风险指标值均值，为主成分特征数量，为风险指标数量，为非线性项权重，为非线性函数用于捕捉与非线性关系，为正则化系数，为正则化强度；in is the eigenvalue of the i- th principal component and the j -th risk indicator value The correlation coefficient of is the absolute value mean of the correlation coefficient between the eigenvalue of the i -th principal component and all risk index values, is the mean of all principal component eigenvalues, is the mean value of all risk indicators, is the number of principal component features, is the number of risk indicators, is the nonlinear term weight, is a nonlinear function used to capture and Non-linear relationships, is the regularization coefficient, is the regularization strength; 选择的风险指标组成第i个主成分特征的风险指标组，定义主成分特征向量集合为，为特征向量元素，风险事件向量集合，为风险事件向量元素，由第h个风险事件对应的风险指标阈值构成，s为风险指标事件库中风险事件的数量，计算多个风险指标组合与风险指标事件库中风险事件的相关性确定第一风险事件，表达式为：choose The risk indicators of constitute the risk indicator group of the i - th principal component characteristics, and the principal component characteristic vector set is defined as , is the feature vector element, the risk event vector set , is the risk event vector element, which is composed of the risk indicator threshold corresponding to the hth risk event. s is the number of risk events in the risk indicator event library. The correlation between multiple risk indicator combinations and risk events in the risk indicator event library is calculated to determine the first risk event. The expression is: ， , ， , ， , 其中为主成分特征i与风险事件h的距离，为第i个主成分特征样本的均值向量，为第h个风险事件样本的均值向量，为第i个主成分特征样本的协方差矩阵，为第h个风险事件样本阈值的协方差矩阵，为矩阵行列式；in is the distance between the principal component feature i and the risk event h , is the mean vector of the i -th principal component feature sample, is the mean vector of the h -th risk event sample, is the covariance matrix of the i- th principal component feature sample, is the covariance matrix of the h -th risk event sample threshold, is the matrix determinant; 设定距离阈值，当时判定主成分特征i对应的风险指标组可能引发风险事件h，否则主成分特征i对应的风险指标组不引发风险事件；Setting distance threshold ,when When , it is determined that the risk indicator group corresponding to the principal component feature i may cause risk event h , otherwise, the risk indicator group corresponding to the principal component feature i does not cause risk event; 按数据类别和数据采集定位为第一风险事件贴标签。Label the first risk event by data category and data collection location. 5.根据权利要求1所述基于数字化标准的城市安全风险评估方法，其特征在于，根据所述第一风险事件预测并发风险事件的方法，包括：5. The urban safety risk assessment method based on digital standards according to claim 1 is characterized in that the method of predicting concurrent risk events according to the first risk event comprises: 构建FP-growth算法模块，挖掘风险事件间的关联性，通过已知风险事件确定并发风险事件，FP-growth算法模块的构建步骤为：Construct the FP-growth algorithm module, explore the correlation between risk events, and determine concurrent risk events through known risk events. The construction steps of the FP-growth algorithm module are as follows: 扫描历史数据统计项的频数，设定最小支持度选出频繁项，由频繁项和频数构建项头表；Scan the frequency of statistical items in historical data, set the minimum support to select frequent items, and construct the item header table based on frequent items and frequencies; 根据项头表筛选和排序事务中的项，由筛选结果构建FP-tree；Filter and sort the items in the transaction according to the item header table, and build the FP-tree based on the filtering results; 根据项头表确定频繁项的条件模式基和事务项集合，根据条件模式基建立条件FP树，采用历史数据训练FP-tree条件实例，采用深挖和递归方法探索常见项集合，输出FP-tree；Determine the conditional pattern base and transaction item set of frequent items according to the item header table, establish a conditional FP-tree according to the conditional pattern base, use historical data to train FP-tree conditional instances, use deep mining and recursive methods to explore common item sets, and output FP-tree; 将第一风险事件输入FP-growth算法模块获得其并发风险事件，按数据类别和数据采集定位为并发风险事件贴标签。The first risk event is input into the FP-growth algorithm module to obtain its concurrent risk events, and the concurrent risk events are labeled according to the data category and data collection location. 6.根据权利要求1所述基于数字化标准的城市安全风险评估方法，其特征在于，据所述第一风险事件预测概率风险事件的方法，包括：6. The urban safety risk assessment method based on digital standards according to claim 1 is characterized in that the method of predicting a probabilistic risk event based on the first risk event comprises: 定义第一风险事件对应风险指标的特征向量为，定义风险指标事件库目标风险事件对应风险指标阈值的特征向量为，计算第一风险事件特征向量和目标风险事件特征向量的综合相似度，表达式为：Define the characteristic vector of the risk indicator corresponding to the first risk event as , define the characteristic vector of the risk indicator threshold corresponding to the target risk event in the risk indicator event library as , calculate the comprehensive similarity between the first risk event feature vector and the target risk event feature vector, the expression is: ， , 其中为第f个第一风险事件特征向量与第g个目标风险事件特征向量的综合相似度，为权重系数，为特征向量的风险指标集合，特征向量的风险指标阈值集合，t为所有项目风险指标数量，为最大项目风险指标数量，为特征向量的风险指标数量，为特征向量的风险指标阈值数量，、为特征向量风险指标均值和方差，、为特征向量的风险指标阈值均值和方差，K为向量维度，为特征向量的元素值，为特征向量的元素值，为分辨系数；in is the characteristic vector of the first risk event of the fth and the g -th target risk event feature vector The comprehensive similarity of is the weight coefficient, is the feature vector A set of risk indicators, Eigenvector The risk indicator threshold set of , t is the number of risk indicators of all projects, is the maximum number of project risk indicators, is the feature vector The number of risk indicators, is the feature vector The number of risk indicator thresholds, , is the feature vector Risk indicator mean and variance, , is the feature vector The mean and variance of the risk indicator threshold, K is the vector dimension, is the feature vector The element value of is the feature vector The element value of is the resolution coefficient; 设定综合相似度阈值，根据综合相似度确定第一风险事件对应的概率风险事件，按数据类别和数据采集定位为概率风险事件贴标签。A comprehensive similarity threshold is set, and the probabilistic risk event corresponding to the first risk event is determined based on the comprehensive similarity, and the probabilistic risk event is labeled according to the data category and data collection location. 7.根据权利要求1所述基于数字化标准的城市安全风险评估方法，其特征在于，获得所述城市安全风险评估结果的方法，包括：7. The urban safety risk assessment method based on digital standards according to claim 1 is characterized in that the method for obtaining the urban safety risk assessment result comprises: 将第一风险事件、并发风险事件、概率风险事件、风险指标和风险事件权重组成安全风险数据集，采用随机森林算法将安全风险集按照7：3划分成训练集和测试集；The first risk event, concurrent risk event, probability risk event, risk index and risk event weight are combined into a safety risk data set, and the random forest algorithm is used to divide the safety risk set into a training set and a test set in a ratio of 7:3; 构建城市安全风险评估模型，所述城市安全风险评估模型包括贝叶斯网络、多模态预测模型和策略层；Constructing a city safety risk assessment model, wherein the city safety risk assessment model includes a Bayesian network, a multimodal prediction model, and a strategy layer; 贝叶斯网络根据已知的第一风险事件、并发风险事件和概率风险事件发生的历史数据来估计风险事件的概率关系，计算每个风险事件在给定其它风险事件下的发生概率；The Bayesian network estimates the probability relationship of risk events based on the historical data of the known first risk event, concurrent risk events, and probabilistic risk events, and calculates the probability of each risk event given other risk events; 多模态预测模型用于学习训练集数据和风险事件发生概率进行安全风险评估，包括遗传-BP神经网络基模型、支持向量机回归基模型和随机森林基模型；The multimodal prediction model is used to learn the training set data and the probability of risk events for safety risk assessment, including the genetic-BP neural network base model, the support vector machine regression base model and the random forest base model; 遗传-BP神经网络用于根据输入数据预测风险评分，采用均方误差损失函数评估预测值与实际值的差异，采用遗传算法优化网络结构和参数，采用Adam优化器调整学习率；The genetic-BP neural network is used to predict the risk score based on the input data, the mean square error loss function is used to evaluate the difference between the predicted value and the actual value, the genetic algorithm is used to optimize the network structure and parameters, and the Adam optimizer is used to adjust the learning rate; 支持向量机回归通过输入数据找到超平面来区分不同风险等级的事件并进行风险评分预测，采用不敏感损失函数在范围内预测评分误差，采用SMO算法求解最优超平面；Support vector machine regression finds a hyperplane by inputting data to distinguish events of different risk levels and make risk score predictions. Insensitive loss function The score error is predicted within the range, and the SMO algorithm is used to solve the optimal hyperplane; 随机森林通过构建决策树进行投票提高风险评分的准确性和稳定性，采用均方误差损失函数评估预测值与实际值差异，根据信息增益选择最佳分裂点；Random forest improves the accuracy and stability of risk scoring by constructing decision trees for voting, uses mean square error loss function to evaluate the difference between predicted values and actual values, and selects the best split point based on information gain; 策略层采用加权投票法整合基模型预测风险评分，输出层提取风险事件的类别和发生位置，最终输出安全风险评估结果，采用测试集评估城市安全风险评估模型；The strategy layer uses a weighted voting method to integrate the base model prediction risk score, and the output layer extracts the category and location of risk events, and finally outputs the security risk assessment results. The test set is used to evaluate the urban security risk assessment model; 将待评估城市安全风险监测数据和历史数据输入城市安全风险评估模型获得安全风险评估结果。The safety risk monitoring data and historical data of the city to be assessed are input into the urban safety risk assessment model to obtain the safety risk assessment results. 8.基于数字化标准的城市安全风险评估系统，用以执行权利要求1-7任一项所述的方法，其特征在于，包括：8. A city safety risk assessment system based on digital standards, used to execute the method according to any one of claims 1 to 7, characterized in that it comprises: 数据采集模块：用于获取城市安全风险监测数据和历史数据，对所述监测数据和所述历史数据进行预处理；Data acquisition module: used to obtain urban safety risk monitoring data and historical data, and pre-process the monitoring data and historical data; 数据处理模块：用于确定风险事件权重，用于对所述监测数据进行异常筛选和分类获得风险指标，用于匹配所述风险指标与风险指标事件库，用于预测不确定风险事件；Data processing module: used to determine the risk event weight, to perform abnormal screening and classification on the monitoring data to obtain risk indicators, to match the risk indicators with the risk indicator event library, and to predict uncertain risk events; 评估模型模块：根据所述第一风险事件、所述不确定风险事件和所述风险事件权重构建城市安全风险评估模型，将待评估城市安全风险监测数据和历史数据输入所述城市安全风险评估模型获得城市安全风险评估结果；Assessment model module: constructing a city safety risk assessment model according to the first risk event, the uncertain risk event and the risk event weight, and inputting the city safety risk monitoring data and historical data to be assessed into the city safety risk assessment model to obtain the city safety risk assessment result; 智能监管模块：用于储存、查看和管理所述监测数据和所述历史数据和所述城市安全风险评估结果，用于根据所述城市安全风险评估结果确定安全风险等级和向用户进行风险预警和建议，用于根据位置信息在安全风险地图上标注风险事件及其类别、安全风险等级和发生概率。Intelligent supervision module: used to store, view and manage the monitoring data, the historical data and the city safety risk assessment results, used to determine the safety risk level and provide risk warnings and suggestions to users based on the city safety risk assessment results, and used to mark risk events and their categories, safety risk levels and probability of occurrence on the safety risk map based on location information.
CN202411910534.8A 2024-12-24 2024-12-24 Urban safety risk assessment method and system based on digital standards Pending CN119398525A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411910534.8A CN119398525A (en) 2024-12-24 2024-12-24 Urban safety risk assessment method and system based on digital standards

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411910534.8A CN119398525A (en) 2024-12-24 2024-12-24 Urban safety risk assessment method and system based on digital standards

Publications (1)

Publication Number Publication Date
CN119398525A true CN119398525A (en) 2025-02-07

Family

ID=94426484

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411910534.8A Pending CN119398525A (en) 2024-12-24 2024-12-24 Urban safety risk assessment method and system based on digital standards

Country Status (1)

Country Link
CN (1) CN119398525A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120197942A (en) * 2025-03-17 2025-06-24 中国标准化研究院 A method and system for constructing an early warning model for urban safety risk assessment benchmark
CN120354249A (en) * 2025-06-24 2025-07-22 中国矿业大学 Dual-index clustering-based worker unsafe behavior trend prediction method and system

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101808127B1 (en) * 2016-12-05 2018-01-19 한국지질자원연구원 Method for evaluating subsidence risk in urban area
WO2023178961A1 (en) * 2022-05-16 2023-09-28 中建三局绿色产业投资有限公司 Silting risk assessment, prediction and early warning method for urban deep water-drainage tunnel
CN118246744A (en) * 2024-04-12 2024-06-25 中国水利水电第四工程局有限公司 Risk assessment method and system for construction site of extra-long tunnel
CN118966511A (en) * 2024-07-02 2024-11-15 天津大学 A human-computer interactive method for the whole process assessment of bioterrorism risk in public places
CN119026917A (en) * 2024-09-09 2024-11-26 华南理工大学 A method for assessing urban power safety and stability risk based on Bayesian network

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101808127B1 (en) * 2016-12-05 2018-01-19 한국지질자원연구원 Method for evaluating subsidence risk in urban area
WO2023178961A1 (en) * 2022-05-16 2023-09-28 中建三局绿色产业投资有限公司 Silting risk assessment, prediction and early warning method for urban deep water-drainage tunnel
CN118246744A (en) * 2024-04-12 2024-06-25 中国水利水电第四工程局有限公司 Risk assessment method and system for construction site of extra-long tunnel
CN118966511A (en) * 2024-07-02 2024-11-15 天津大学 A human-computer interactive method for the whole process assessment of bioterrorism risk in public places
CN119026917A (en) * 2024-09-09 2024-11-26 华南理工大学 A method for assessing urban power safety and stability risk based on Bayesian network

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120197942A (en) * 2025-03-17 2025-06-24 中国标准化研究院 A method and system for constructing an early warning model for urban safety risk assessment benchmark
CN120197942B (en) * 2025-03-17 2025-09-05 中国标准化研究院 A method and system for constructing an early warning model for urban safety risk assessment benchmark
CN120354249A (en) * 2025-06-24 2025-07-22 中国矿业大学 Dual-index clustering-based worker unsafe behavior trend prediction method and system

Similar Documents

Publication Publication Date Title
CN118037047B (en) AI-based mine safety monitoring system
CN117726181B (en) Collaborative fusion and hierarchical prediction method for typical disaster risk heterogeneous information of coal mine
CN118094379B (en) Oil well fault diagnosis prediction method and system for multi-factor fusion analysis
CN110555617B (en) Real-time dynamic quantitative assessment method for building fire risk based on Internet of things
CN118246744B (en) A risk assessment method and system for extra-long tunnel construction site
CN118548114B (en) A tunnel monitoring data emergency response warning method and system
CN119398525A (en) Urban safety risk assessment method and system based on digital standards
CN111460167A (en) Method and related equipment for locating sewage objects based on knowledge graph
CN115063020A (en) Cascade hydropower station multidimensional safety scheduling device and method based on risk monitoring fusion
CN120069233B (en) An environmental emergency decision-making assistance method and system based on artificial intelligence
CN112949202A (en) Bayesian network-based rockburst probability prediction method
CN117238126B (en) A traffic accident risk assessment method in continuous flow road scenarios
CN117787726A (en) A road natural disaster risk assessment method based on supervised learning
CN119047855A (en) Multi-dimensional operation risk assessment method based on big data analysis
CN118297414A (en) Mine water hazard forecasting electronic equipment and computer program product
JP3674707B1 (en) Disaster prevention business plan support system and method
CN120450411A (en) Methods, systems, equipment and media for risk identification and treatment of underground powerhouse construction
CN119916130A (en) Real-time analysis and precise positioning of distribution network faults and their treatment methods and systems
CN117057601B (en) Non-coal mine safety monitoring and early warning system based on Internet of Things
CN119130425A (en) A method and device for determining the optimal maintenance position for power grid fault prediction
CN119417224A (en) A method and system for predicting regional fire risk in catering places based on XGBoost model
CN119647641A (en) Debris flow hazard level prediction method and system based on GRA-PCA-RBF neural network model
CN118297375A (en) A method and system for assessing underground pipe network operation risk based on machine learning
CN117933716A (en) A road natural disaster risk assessment method based on fuzzy neural network
CN120197942B (en) A method and system for constructing an early warning model for urban safety risk assessment benchmark

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination