CN113361855A - Short, medium and long-term risk warning method and device - Google Patents

Short, medium and long-term risk warning method and device Download PDF

Info

Publication number
CN113361855A
CN113361855A CN202110496412.9A CN202110496412A CN113361855A CN 113361855 A CN113361855 A CN 113361855A CN 202110496412 A CN202110496412 A CN 202110496412A CN 113361855 A CN113361855 A CN 113361855A
Authority
CN
China
Prior art keywords
risk
target
feature vector
medium
short
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
CN202110496412.9A
Other languages
Chinese (zh)
Inventor
孙培梁
林枫
王军
余韦
钟云伟
斯越薪
张志明
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.)
Hangzhou Judicial Bureau
Hangzhou Xijiao Prison
ZHEJIANG POLICE VOCATIONAL ACADEMY
Tianjin University
Zhejiang University of Technology ZJUT
Zhejiang Sci Tech University ZSTU
Zhejiang University of Science and Technology ZUST
Original Assignee
Hangzhou Judicial Bureau
Hangzhou Xijiao Prison
ZHEJIANG POLICE VOCATIONAL ACADEMY
Tianjin University
Zhejiang University of Technology ZJUT
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 Hangzhou Judicial Bureau, Hangzhou Xijiao Prison, ZHEJIANG POLICE VOCATIONAL ACADEMY, Tianjin University, Zhejiang University of Technology ZJUT filed Critical Hangzhou Judicial Bureau
Priority to CN202110496412.9A priority Critical patent/CN113361855A/en
Publication of CN113361855A publication Critical patent/CN113361855A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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; CALCULATING OR 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
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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

Landscapes

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

Abstract

The invention discloses a short, medium and long term risk warning method and a device, wherein a characteristic vector of a risk warning target is obtained through word embedding, the characteristic vector of the risk warning target is input into a convolutional neural network, a target portrait characteristic vector of the risk warning target is extracted, a time series characteristic vector is obtained according to a monitoring data sample corresponding to an object with multiple risk accidents when each risk accident happens, a recursion neural network model is adopted to carry out recursion operation on the time series characteristic vector, a risk tendency occurrence characteristic vector is extracted, finally, the target portrait characteristic vector and the risk tendency occurrence characteristic vector are fused and input into a full-connection neural network model, a risk value corresponding to the risk warning target is predicted, and when the risk value is larger than a preset threshold value, warning is sent. The method effectively improves the prediction precision of the reoccurrence risk, innovating a short-medium-long term risk assessment method system, and promoting the short-medium-long term risk assessment efficiency and the intelligent level.

Description

Short, medium and long-term risk warning method and device
Technical Field
The application belongs to the technical field of safety production, and particularly relates to a short-medium-long-term risk warning method and device.
Background
Safety is always an important link for production and social stability, safety is not emphasized, and huge cost is brought. Safety is not only reflected in safe production, but also in aspects of social life, such as fire safety, traffic safety, dangerous goods handling, public security, trample accidents and the like.
The first and the first prevention is the basic guideline of safety production in China, how to early warn and eliminate potential safety hazards is always the most important, and the safety early warning and supervision and inspection work are the main means and measures of the first prevention.
At present, the safety early warning is usually carried out according to each danger source in advance, the influence of the accident which occurs once in the past on the secondary accident in the future is rarely considered when the early warning is carried out, and the early warning is carried out on the condition that the accident which occurs once occurs again in a new environment is rarely considered.
Disclosure of Invention
The application aims to provide a short-medium-long-term risk warning method and device, and the problem that the reoccurrence risk judgment in the prior art is inaccurate is solved.
In order to achieve the purpose, the technical scheme of the application is as follows:
a short, medium and long term risk warning method comprises the following steps:
obtaining a feature vector X of a risk warning target through word embedding0The feature vector X of the risk warning target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1；
Acquiring a monitoring data sample corresponding to a risk accident object occurring for multiple times when a risk accident occurs each time, and acquiring a time series characteristic vector H through word embedding0Using a recurrent neural network model RNN to time-series feature vector H0Performing recursion operation to extract N times of risk tendency characteristic vectors HN；
The feature vector X of the target image1And N occurrence risk tendency feature vectors HNPerforming fusion to fuse the feature X2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and sending an alarm when the risk value is greater than a preset threshold value.
Further, the characteristic vector X of the target for alarming risk0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1The mathematical expression is as follows:
X1＝CNN(X0)。
further, the time series characteristic vector H is subjected to the RNN (recurrent neural network) model0Performing recursion operation to extract N times of risk tendency characteristic vectors HNThe mathematical expression is as follows:
HN＝RNN(H0)。
further, the target portrait feature vector X1And N occurrence risk tendency feature vectors HNAnd performing fusion, wherein the mathematical expression is as follows:
X2＝concat(X1,HN)。
further, the feature X is fused2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and comprising the following steps:
first, the feature X is fused2Inputting the data into a full connection layer to obtain a three-dimensional comprehensive characteristic vector X3：
X3＝FC(X2)；
Then, predicting a risk value P corresponding to a risk alarm target by adopting softmax operationi：
Figure BDA0003054589120000021
Wherein the content of the first and second substances,
Figure BDA0003054589120000022
representing a synthetic feature vector X3The corresponding i-component of (a) to (b),
Figure BDA0003054589120000023
representing a synthetic feature vector X3The corresponding j components in (1), i, j, represent short, medium, and long term, PiRepresenting the probability that the i component corresponds to.
The application also provides an alarm device for short, medium and long-term risks, which comprises:
a target portrait feature extraction module for obtaining a feature vector X of a risk alarm target through word embedding0Inform the riskFeature vector X of alert target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1；
A risk tendency characteristic vector extraction module for acquiring corresponding monitoring data samples of multiple risk accident occurrence objects when risk accidents occur each time, and obtaining a time sequence characteristic vector H through word embedding0Using a recurrent neural network model RNN to time-series feature vector H0Performing recursion operation to extract N times of risk tendency characteristic vectors HN；
A fusion early warning module for converting the target portrait feature vector X1And N occurrence risk tendency feature vectors HNPerforming fusion to fuse the feature X2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and sending an alarm when the risk value is greater than a preset threshold value.
Further, the characteristic vector X of the target for alarming risk0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1The mathematical expression is as follows:
X1＝CNN(X0)。
further, the time series characteristic vector H is subjected to the RNN (recurrent neural network) model0Performing recursion operation to extract N times of risk tendency characteristic vectors HNThe mathematical expression is as follows:
HN＝RNN(H0)。
further, the target portrait feature vector X1And N occurrence risk tendency feature vectors HNAnd performing fusion, wherein the mathematical expression is as follows:
X2＝concat(X1,HN)。
further, the feature X is fused2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and comprising the following steps:
first, the feature X is fused2Inputting the data into the full connection layer to obtain a three-dimensional comprehensive characteristic directionQuantity X3：
X3＝FC(X2)；
Then, predicting a risk value P corresponding to a risk alarm target by adopting softmax operationi：
Figure BDA0003054589120000031
Wherein the content of the first and second substances,
Figure BDA0003054589120000032
representing a synthetic feature vector X3The corresponding i-component of (a) to (b),
Figure BDA0003054589120000033
representing a synthetic feature vector X3The corresponding j components in (1), i, j, represent short, medium, and long term, PiRepresenting the probability that the i component corresponds to.
According to the short-medium-long-term risk warning method and device, subjective difference and contingency exist in a secondary risk assessment method based on a traditional manual scoring table, the limitation of the traditional method is broken through, the neural network method of the leading edge is introduced for the first time to extract portrait features, the driving features of reoccurrence risks are extracted based on priori knowledge, portrait features and priori knowledge are efficiently fused, the prediction accuracy of reoccurrence risks is effectively improved, a short-medium-long-term risk assessment method system is innovated, and the short-medium-long-term risk assessment efficiency and the intelligent level are promoted.
Drawings
Fig. 1 is a flowchart of a short-medium-long term risk warning method according to the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Whether food safety, public health safety, school safety and the like, a certain safety early warning mechanism is required. In terms of food safety, food produced by a certain company does not meet standards, and a lot of safety problems are easily caused. For example, the quality of milk powder in the current year does not reach the standard, so that a large-head doll phenomenon appears in many infants, and the infants are seriously malnourished. Therefore, safety supervision and early warning are carried out on the factories, and the safety accidents can be effectively prevented from happening again. At present, the prediction and the alarm of the reoccurrence of the safety accident can be only scored once based on a risk degree scale after the factory is rectified, and various risk factors of the factory are scored manually to form a prediction result. The specific effect of the method is difficult to evaluate, and particularly, the development of prediction on secondary risks in the medium and long periods cannot be really realized due to the fact that new risk factors and the like are difficult to predict.
In one embodiment, as shown in fig. 1, there is provided a short, medium and long term risk warning method, including:
step S1, obtaining the characteristic vector X of the risk warning target through word embedding0The feature vector X of the risk warning target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1。
According to the method and the device, the risk warning target is subjected to prediction warning, and can be a certain processing factory to perform early warning on the safety risk of the processing factory. And the safety risk of a dangerous goods warehouse can be pre-warned. Or aiming at the illegal personnel, carrying out safety risk early warning on the short, medium and long term illegal behaviors.
For the risk warning target, various risk data of the target are firstly obtained through general survey, and by taking a processing factory as an example, the type of the produced product, the area of the factory, the number of staff, the culture degree of the staff, the disinfection times, the quality guarantee period of raw materials, the storage environment temperature and the like are obtained.
Obtaining the characteristic vector of the risk alarm target through word embedding, and setting
Figure BDA0003054589120000051
Is a word embedding matrix, where n is the number of words in the modeled field, d2Is the word embedding dimension.Thus, each content of the modeling field is embedded as a mathematical vector
Figure BDA0003054589120000052
And the model is a trainable parameter, and the embedding matrix can be gradually optimized by the model along with the training of the whole model, so that the optimized word embedding is more beneficial to short, medium and long-term risk prediction tasks.
After word embedding, the obtained characteristic vector is X0In order to improve the abstract capability of the target image feature vector, a Convolution Neural Network (CNN) is used to extract the final target image feature vector, namely X1：
X1＝CNN(X0)。
Step S2, acquiring corresponding monitoring data samples of multiple risk accident occurrence objects when risk accidents occur each time, and obtaining time series characteristic vectors H through word embedding0Using a recurrent neural network model RNN to time-series feature vector H0Performing recursion operation to extract N times of risk tendency characteristic vectors HN。
Recurrent neural network model RNN vs time series eigenvector H0A recursive operation is performed, represented as follows:
HN＝RNN(H0)。
for a plurality of risk accident objects, each risk accident object has multiple risk accidents, corresponding monitoring data when each risk accident occurs is obtained, the feature vectors corresponding to each risk accident are obtained through word embedding, and the feature vectors corresponding to each risk accident form a time sequence feature vector H0。
For example, for a plurality of processing plants in which a security event occurs many times, word embedding is performed on monitoring data corresponding to each occurrence of the security event to obtain corresponding feature vectors. Forming a time sequence feature vector H by the feature vector sequence corresponding to the multiple occurrence risk accidents according to the time sequence0Labeling corresponding labels as sample data, inputting the sample data into the recurrent neural network model, and extracting N times of windRisk trend feature vectors.
The method and the device can extract the tendency characteristics of the N risks, which are in different times, and fuse time characteristic information, so that the characteristics which are in long time can be invisibly weighted and weakened by the model.
Step S3, target image feature vector X1And N occurrence risk tendency feature vectors HNPerforming fusion to fuse the feature X2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and sending an alarm when the risk value is greater than a preset threshold value.
Fusing the feature vector of the target portrait and the feature vector of the risk tendency of occurrence for N times, wherein the fusion process is expressed as follows:
X2＝concat(X1,HN)
wherein the concat operation is a join operation of two vectors, X1Is a feature vector of the target image, HNRepresenting the N occurrence risk tendency feature vectors.
Inputting the fusion characteristics into a full-connection layer of a full-connection neural network model to obtain a three-dimensional comprehensive characteristic vector X3I.e. by
X3＝FC(X2)。
Finally, performing softmax operation, wherein the calculation formula is as follows:
Figure BDA0003054589120000061
wherein, among others,
Figure BDA0003054589120000062
representing a synthetic feature vector X3The corresponding i-component of (a) to (b),
Figure BDA0003054589120000063
representing a synthetic feature vector X3The corresponding j components in (1), i, j, represent short, medium, and long term, PiRepresenting the probability that the i component corresponds to. i. j belongs to {1,2,3}, and corresponds to short term, medium term, and long term, respectively.
Therefore, the probability of short-term, medium-term and long-term occurrence risks can be predicted, and when the probability is larger than a preset threshold (for example, 80%), an alarm is given.
The short, medium and long term risks are related to the characteristics of the target portrait and are also highly related to the nearest time series of the target, and the method models the space-time environment of the target, which is not achieved by a simple neural network model in the past. Given that it is difficult for a model to adequately assess the target short, medium and long term risk through one data type, multiple data types (multimodal), particularly text data types, are modeled.
This application has better risk prediction ability of reporting an emergency and asking for help or increased vigilance to the target that takes place secondary risk easily, and the target can be mill, hazardous articles warehouse etc. also can be to the community correction personnel or go into the risk prediction that prison personnel carry out the rescission and report an emergency and ask for help or increased vigilance, can in time prevent various safety risks, take precautions against in the bud.
In one embodiment, the present application further provides a short, medium and long term risk warning device, including:
a target portrait feature extraction module for obtaining a feature vector X of a risk alarm target through word embedding0The feature vector X of the risk warning target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1；
A risk tendency characteristic vector extraction module for acquiring corresponding monitoring data samples of multiple risk accident occurrence objects when risk accidents occur each time, and obtaining a time sequence characteristic vector H through word embedding0Using a recurrent neural network model RNN to time-series feature vector H0Performing recursion operation to extract N times of risk tendency characteristic vectors HN；
A fusion early warning module for converting the target portrait feature vector X1And N occurrence risk tendency feature vectors HNPerforming fusion to fuse the feature X2Inputting the data into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and calculating the risk valueAnd when the threshold value is larger than the preset threshold value, sending an alarm.
For specific limitations of the short, medium and long term risk warning device, reference may be made to the above limitations of the short, medium and long term risk warning method, which are not described herein again. The modules in the short, medium and long term risk warning device can be wholly or partially implemented by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
The memory and the processor are electrically connected, directly or indirectly, to enable transmission or interaction of data. For example, the components may be electrically connected to each other via one or more communication buses or signal lines. The memory stores a computer program that can be executed on the processor, and the processor executes the computer program stored in the memory, thereby implementing the network topology layout method in the embodiment of the present invention.
The Memory may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like. The memory is used for storing programs, and the processor executes the programs after receiving the execution instructions.
The processor may be an integrated circuit chip having data processing capabilities. The Processor may be a general-purpose Processor including a Central Processing Unit (CPU), a Network Processor (NP), and the like. The various methods, steps and logic blocks disclosed in embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (10)

1. A short, medium and long term risk warning method is characterized by comprising the following steps:
obtaining a feature vector X of a risk warning target through word embedding0The feature vector X of the risk warning target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1；
Acquiring a monitoring data sample corresponding to a risk accident object occurring for multiple times when a risk accident occurs each time, and acquiring a time series characteristic vector H through word embedding0Using a recurrent neural network model RNN to time-series feature vector H0Performing recursion operation to extract N times of risk tendency characteristic vectors HN；
The feature vector X of the target image1And N occurrence risk tendency feature vectors HNPerforming fusion to fuse the feature X2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and sending an alarm when the risk value is greater than a preset threshold value.
2. The short, medium and long term risk alerting method of claim 1, wherein the risk alerting target's feature vector X0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1The mathematical expression is as follows:
X1＝CNN(X0)。
3. a short-medium-long term risk alerting method as defined in claim 1, whichCharacterized in that the characteristic vector H of the time sequence is subjected to RNN (recurrent neural network) by adopting a recurrent neural network model0Performing recursion operation to extract N times of risk tendency characteristic vectors HNThe mathematical expression is as follows:
HN＝RNN(H0)。
4. a short-medium-long term risk alerting method as claimed in claim 1, wherein said target image feature vector X is used to alert target image1And N occurrence risk tendency feature vectors HNAnd performing fusion, wherein the mathematical expression is as follows:
X2＝concat(X1,HN)。
5. a short-medium-long term risk alerting method as claimed in claim 1, wherein said feature X is fused2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and comprising the following steps:
first, the feature X is fused2Inputting the data into a full connection layer to obtain a three-dimensional comprehensive characteristic vector X3：
X3＝FC(x2)；
Then, predicting a risk value P corresponding to a risk alarm target by adopting softmax operationi：
Figure FDA0003054589110000021
Wherein the content of the first and second substances,
Figure FDA0003054589110000022
representing a synthetic feature vector X3The corresponding i-component of (a) to (b),
Figure FDA0003054589110000023
representing a synthetic feature vector X3The corresponding j components in (1), i, j, represent short, medium, and long term, PiRepresenting the probability that the i component corresponds to.
6. A short-medium-long term risk alerting device, characterized in that the short-medium-long term risk alerting device comprises:
a target portrait feature extraction module for obtaining a feature vector X of a risk alarm target through word embedding0The feature vector X of the risk warning target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1；
A risk tendency characteristic vector extraction module for acquiring corresponding monitoring data samples of multiple risk accident occurrence objects when risk accidents occur each time, and obtaining a time sequence characteristic vector H through word embedding0Using a recurrent neural network model RNN to time-series feature vector H0Performing recursion operation to extract N times of risk tendency characteristic vectors HN；
A fusion early warning module for converting the target portrait feature vector X1And N occurrence risk tendency feature vectors HNPerforming fusion to fuse the feature X2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and sending an alarm when the risk value is greater than a preset threshold value.
7. A short-medium-long term risk alerting device as claimed in claim 6, wherein the feature vector X of the risk alerting target0Inputting the convolution neural network to extract the target portrait feature vector X of the risk alarm target1The mathematical expression is as follows:
X1＝CNN(X0)。
8. the short-medium-term risk warning apparatus according to claim 5, wherein the time-series feature vector H is subjected to a recurrent neural network model RNN0Performing recursion operation to extract N times of risk tendency characteristic vectors HNThe mathematical expression is as follows:
HN＝RNN(H0)。
9. a short-medium-long term risk warning device as claimed in claim 5, wherein the target image feature vector X1And N occurrence risk tendency feature vectors HNAnd performing fusion, wherein the mathematical expression is as follows:
X2＝concat(X1,HN)。
10. a short-medium-long term risk alerting device as claimed in claim 5, wherein the feature X is fused2Inputting the risk value into a full-connection neural network model, predicting a risk value corresponding to a risk alarm target, and comprising the following steps:
first, the feature X is fused2Inputting the data into a full connection layer to obtain a three-dimensional comprehensive characteristic vector X3：
X3＝FC(X2)；
Then, predicting a risk value P corresponding to a risk alarm target by adopting softmax operationi：
Figure FDA0003054589110000031
Wherein the content of the first and second substances,
Figure FDA0003054589110000032
representing a synthetic feature vector X3The corresponding i-component of (a) to (b),
Figure FDA0003054589110000033
representing a synthetic feature vector X3The corresponding j components in (1), i, j, represent short, medium, and long term, PiRepresenting the probability that the i component corresponds to.
CN202110496412.9A 2021-05-07 2021-05-07 Short, medium and long-term risk warning method and device Pending CN113361855A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110496412.9A CN113361855A (en) 2021-05-07 2021-05-07 Short, medium and long-term risk warning method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110496412.9A CN113361855A (en) 2021-05-07 2021-05-07 Short, medium and long-term risk warning method and device

Publications (1)

Publication Number Publication Date
CN113361855A true CN113361855A (en) 2021-09-07

Family

ID=77525899

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110496412.9A Pending CN113361855A (en) 2021-05-07 2021-05-07 Short, medium and long-term risk warning method and device

Country Status (1)

Country Link
CN (1) CN113361855A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117709732A (en) * 2024-02-06 2024-03-15 北京天译科技有限公司 Agricultural disaster report generation method and system combined with meteorological monitoring data

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107104978A (en) * 2017-05-24 2017-08-29 赖洪昌 A kind of network risks method for early warning based on deep learning
CN110191113A (en) * 2019-05-24 2019-08-30 新华三信息安全技术有限公司 A kind of user behavior methods of risk assessment and device
US20200036611A1 (en) * 2018-07-24 2020-01-30 Baidu Online Network Technology (Beijing) Co., Ltd. Method and apparatus for warning
CN111898836A (en) * 2020-08-18 2020-11-06 石拓 Crime space-time prediction method and system
CN111967565A (en) * 2020-10-23 2020-11-20 支付宝(杭州)信息技术有限公司 Neural network system, method and device for risk assessment
CN112149413A (en) * 2020-09-07 2020-12-29 国家计算机网络与信息安全管理中心 Method and device for identifying state of internet website based on neural network and computer readable storage medium
US20210035437A1 (en) * 2019-08-01 2021-02-04 Fuji Xerox Co., Ltd. System and method for event prevention and prediction
WO2021042843A1 (en) * 2019-09-06 2021-03-11 平安科技（深圳）有限公司 Alert information decision method and apparatus, computer device and storage medium

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107104978A (en) * 2017-05-24 2017-08-29 赖洪昌 A kind of network risks method for early warning based on deep learning
US20200036611A1 (en) * 2018-07-24 2020-01-30 Baidu Online Network Technology (Beijing) Co., Ltd. Method and apparatus for warning
CN110191113A (en) * 2019-05-24 2019-08-30 新华三信息安全技术有限公司 A kind of user behavior methods of risk assessment and device
US20210035437A1 (en) * 2019-08-01 2021-02-04 Fuji Xerox Co., Ltd. System and method for event prevention and prediction
WO2021042843A1 (en) * 2019-09-06 2021-03-11 平安科技（深圳）有限公司 Alert information decision method and apparatus, computer device and storage medium
CN111898836A (en) * 2020-08-18 2020-11-06 石拓 Crime space-time prediction method and system
CN112149413A (en) * 2020-09-07 2020-12-29 国家计算机网络与信息安全管理中心 Method and device for identifying state of internet website based on neural network and computer readable storage medium
CN111967565A (en) * 2020-10-23 2020-11-20 支付宝(杭州)信息技术有限公司 Neural network system, method and device for risk assessment

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117709732A (en) * 2024-02-06 2024-03-15 北京天译科技有限公司 Agricultural disaster report generation method and system combined with meteorological monitoring data
CN117709732B (en) * 2024-02-06 2024-04-26 北京天译科技有限公司 Agricultural disaster report generation method and system combined with meteorological monitoring data

Similar Documents

Publication Publication Date Title
Kong et al. Deep‐stacking network approach by multisource data mining for hazardous risk identification in IoT‐based intelligent food management systems
CN109686036A (en) A kind of fire monitoring method, device and edge calculations device
CN107346388A (en) Web attack detection methods and device
Sultan Mahmud et al. Smart fire detection system with early notifications using machine learning
CN115358155A (en) Power big data abnormity early warning method, device, equipment and readable storage medium
CN110009126B (en) Online alarm analysis method based on fusion of PLS model and PCA contribution degree
CN115828145B (en) Online monitoring method, system and medium for electronic equipment
CN116010886B (en) Security monitoring method, device, electronic equipment and storage medium
CN113361855A (en) Short, medium and long-term risk warning method and device
Liu et al. Detection and location of unsafe behaviour in digital images: A visual grounding approach
CN116759053A (en) Medical system prevention and control method and system based on Internet of things system
CN115169779A (en) Enterprise safety risk early warning prediction system and method
CN117595504A (en) Intelligent monitoring and early warning method for power grid running state
CN116311751B (en) Underground coal mine use fire prevention and extinguishment electric control system
CN116385031A (en) Data tracing method based on big data and multiple data sources
Chang et al. Safety risk assessment of electric power operation site based on variable precision rough set
CN116701866A (en) Park event linkage processing method based on Internet of things equipment
CN116665401A (en) Accident prevention alarm system for chemical production
CN115409100A (en) Method, device, medium and equipment for generating power accident emergency scheme
CN111260173A (en) Production monitoring method and device and computer readable storage medium
CN115187026A (en) Industrial risk monitoring method and system and readable storage medium
Adi et al. Automatic false-alarm labeling for sensor data
CN113393017B (en) Individual complex relationship-based supervision place safety early warning method and device
CN117725478A (en) Cable protection prediction management system and method based on fire-fighting linkage
CN112307090B (en) Detection method and system based on tunnel security data

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20210907