CN109657880A - A kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network - Google Patents
A kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network Download PDFInfo
- Publication number
- CN109657880A CN109657880A CN201910021019.7A CN201910021019A CN109657880A CN 109657880 A CN109657880 A CN 109657880A CN 201910021019 A CN201910021019 A CN 201910021019A CN 109657880 A CN109657880 A CN 109657880A
- Authority
- CN
- China
- Prior art keywords
- bayesian network
- ships
- collision accidents
- feature
- 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
Links
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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
-
- 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
- G06F18/24155—Bayesian classification
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Computational Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Mathematical Analysis (AREA)
- Databases & Information Systems (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Software Systems (AREA)
- Algebra (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network, this method passes through building Collision Accidents of Ships prediction model, realize the grade forecast to Collision Accidents of Ships, from the angle of data, the connection between accident factor is excavated by depth, the rule between accident factor is found, instead of the too strong expertise method of subjectivity, realizes the foundation of Collision Accidents of Ships Bayesian network.From the angle of method, Collision Accidents of Ships prediction model is constructed based on Bayesian network, solves the forecasting problem of Collision Accidents of Ships grade, pushes the development of maritime affairs Risk Theory.From the angle of application, research achievement can provide scientific basis for that can meet ship's navigation commander and Emergency decision.
Description
Technical field
The present invention relates to traffic safety electric powder prediction, in particular to a kind of ship collision thing based on Bayesian network
Therefore grade prediction technique and system.
Background technique
China is world shipping big country, and shipping has important strategic importance to country.With the prosperity of marine transportation industry,
Course line is increasingly busy, and concentration of vessel continues to increase, and ship is continuously improved to enlargement, professional development, ship speed, ring of opening the navigation or air flight
Border more becomes to complicating, and all kinds of Maritime Traffic Accidents is caused frequently to occur.In the case of at sea traffic safety situation is increasingly serious,
How safe and efficient sea transport is guaranteed, it has also become countries in the world, each maritime sector focus of attention.
Maritime accidents forecast statistics shows that Collision Accidents of Ships incidence is in always the forefront of all kinds of casualties.This
Ship collision refers to that two or two or more ships navigable waters logical at sea or with marine facies is in contact and causes in invention
The accident of damage.Consequence brought by Collision Accidents of Ships is often catastrophic, seriously threatens life, the property safety of people,
And the serious pollution of marine environment of the meeting such as ship spill, danger leakage, consequence are not easy to control.Meanwhile collision at sea thing
Therefore there is very big randomness and uncertainty.Therefore, various countries' maritime field pays much attention to always the research of Collision Accidents of Ships.
Research about safety of traffic on water is all for many years the research hotspot and emphasis of field of traffic, causes the country
The highest attention of outer related researcher, and produce plentiful and substantial research achievement.The risk of Collision Accidents of Ships generally uses thing
Therefore the product of a possibility that occurring (probability) and loss caused by accident is measured.On the whole, domestic and foreign scholars are for ship
The research of collision accident is concentrated mainly in the analysis of the probability and accident occurrence cause of accident generation, to collision causality loss
Forecasting research it is less.
Collision Accidents of Ships probabilistic forecasting, in terms of, a large amount of in-depth studies, mould has been carried out in foreign countries
Type and Theory comparison are mature.The Collision Accidents of Ships reason that Friis-Hansen P and Simonsen B C is built in summary forefathers
On the basis of chain, the influence factor that Collision Accidents of Ships occurs further is analyzed, and establish model and carry out to contingency occurrence probability
Inference Forecast.Ylitalo J proposes ship collision probability assessment model criteria, successively uses for the ship of specific bodies of water navigation
Three model Ship ' collision probabilities, and ship collision wind between the waters 2006 to 2015 is predicted according to assessment result
The tendency of danger.S U etc. by analysis straits geographical feature and traffic condition, in conjunction with VTS (Ship Traffic Service,
Vessel Traffic Service) data, risk analysis mathematical model is established, analyzes different types of factor to straits
Cause the influence of risk.
Compared to external research, China carries out relatively a little later the correlative study of Collision Accidents of Ships, but passes through
The joint efforts of numerous scholars, expert are crossed, plentiful and substantial research achievement is also obtained.Fan Yaotian will be a certain on the basis of main channel axis
Waters carries out gridding division, comprehensively considers the encounter rate that various factors calculates each cell, and the wind based on geographical distribution
Dangerous appraisal procedure establishes ship collision risk forecast model by taking PORT OF TIANJIN as an example, water transportation is investigated, construct prediction model,
Model verifying and future anticipation combine, and form the waters navigation environment risk assessment new method of closed loop.Yang Tianxue
The ship collision risk analysis frame based on probability distributive function is established, the risk analysis mould of artificial neural network is established
Type analyzes the ship collision risk of 30 harbour water areas of domestic coastal.
Currently, the research of maritime accidents forecast consequence is still in the junior stage, up for carrying out deep probe into.However,
In terms of road traffic accident consequence, situation, many researchs have been done both at home and abroad, and common method includes BP neural network, ash
The multivariate analytical techniques of the structurings such as color association, decision tree, Bayesian network.Due to the particularity and correspondence of water transportation situation
The high requirement suddenly managed, Collision Accidents of Ships research cannot indiscriminately imitate the analytical model of land traffic accident, but the two has perhaps
More similarities.Therefore, it can refer to what field of road traffic was used when carrying out Collision Accidents of Ships consequences analysis and prediction
Method, but scientific and reasonable research method should be selected in conjunction with the characteristics of Collision Accidents of Ships and Evolution Mechanism when concrete application.
Research of the domestic and foreign scholars in recent years in terms of Collision Accidents of Ships is made a general survey of it is found that research direction is mainly accident hair
Raw probabilistic forecasting, causation analysis etc., and research method graduallys mature.However, loss caused by Collision Accidents of Ships is accident
The important component of risk can provide decision-making foundation to the prediction of collision damage sequence for the formulation of emergency disposal scheme,
It is had very important effect during emergency disposal.At this stage in accident emergency administrative decision, often correlation is led,
The experience dependence of expert is stronger, causes the efficiency of decision-making relatively low.Research in terms of field of road traffic, damage sequence is more,
Correlative study method is applied to water transportation field not yet widely.Generally speaking, collision at sea damage sequence is (as sternly
Weight degree, casualty situations etc.) research also in the budding period, has very big research potential and space, need to be further strengthened and
In-depth.
Summary of the invention
The present invention provides a kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network, can be realized
To the grade forecast of Collision Accidents of Ships.
According to an aspect of the invention, there is provided a kind of Collision Accidents of Ships grade forecast side based on Bayesian network
Method, comprising the following steps:
Maritime accidents forecast survey report is obtained, accident factor data are therefrom extracted；
The accident factor data are subjected to preliminary sliding-model control according to Joint Distribution feature, by each influence factor and thing
Therefore consequence element carries out sensitivity analysis, merges some discrete set of each influence factor, obtain discretization set as a result,
As Bayesian network node；
Based between the correlation analysis variable, the row of the empirically determined Bayesian network node is modeled in conjunction with expert
Sequence；Machine language is write, bayesian network structure learning is carried out；Bayesian network structure is obtained according to K2 structure learning algorithm；
According to the bayesian network structure, corresponding prior information is inputted, with Bayes's Parameter Learning Algorithm, study is obtained
Obtain the conditional probability distribution of each network node；Gained conditional probability table is assigned to determining Bayesian Network Topology Structures, is established
The Bayesian network of Collision Accidents of Ships；
According to the method for joint tree reasoning, Bayesian network is constructed according to the Bayesian network of the Collision Accidents of Ships
Inference pattern obtains Collision Accidents of Ships consequence prediction model；
Using marine incident grade as query interface, after calculating that accidents happened according to the Collision Accidents of Ships consequence prediction model
The probability distribution of fruit.
The accident factor data, comprising:
Accidents Disasters consequence factor: incident classification；
Environmental factor: time of casualty, wind-force, visibility, water surface situation, temperature and/or water temperature；
Ship factor: Description of Ship, captain, the beam, the age of a ship, gross tonnage, loading, crewman's number and/or all properties of ship.
The method also includes:
Study precision, the prediction accuracy level of Bayesian network model from Bayesian network, touch the ship of building
It hits accident consequence prediction model and carries out validation verification.
It further include data cleansing step after the acquisition accident factor data, specific as follows:
Collision Accidents of Ships essential information data are arranged, for error logging number present in the accident analysis report
According to judging the whether accurate confidence level of a certain data according to evidence theory, and then clean to the data of mistake；
Ship basic static information is proofreaded and supplemented using ship static data.
During the acquisition bayesian network structure, further includes:
Using the method based on scoring search in the study of the bayesian network structure.
The Bayesian network node, further includes following treatment process:
Using the method for cluster, the stronger feature of correlation in the Bayesian network node is polymerized to cluster one by one, often
Elemental characteristic in a cluster is all similar characteristic, and cluster is considered as abstract node；
Each cluster that feature clustering is obtained, complete class in multiple features to an abstract characteristics codomain mapping.It will not
The feature of cluster merges with abstract characteristics, generates new data set, as the Bayesian network node after clustering processing.
Further include:
Before carrying out the search study of bayesian network structure according to K2 searching algorithm, the sequence of network node is first given；
It is ranked up using the BD score function based on Bayesian statistics.
Further include:
Based between the correlation analysis variable, experience is modeled in conjunction with expert to determine the sequence of Bayesian network node.
According to another aspect of the present invention, a kind of Collision Accidents of Ships grade forecast based on Bayesian network is provided
System, including data input cell, endpoint processing unit, topological determination unit, probability distribution determination unit, model construction unit
And predicting unit, wherein
The data input cell therefrom obtains accident factor data for obtaining maritime accidents forecast survey report；
The endpoint processing unit, for the accident factor data to be carried out preliminary discretization according to Joint Distribution feature
Each influence factor and damage sequence element are carried out sensitivity analysis, merge some discrete set of each influence factor by processing,
Obtain discretization set as a result, as Bayesian network node；
The topology determination unit, for modeling empirically determined institute in conjunction with expert based between the correlation analysis variable
State the sequence of Bayesian network node；Machine language is write, bayesian network structure learning is carried out；According to K2 structure learning algorithm
Obtain bayesian network structure；
The probability distribution determination unit is used for inputting corresponding prior information according to the bayesian network structure
Bayes's Parameter Learning Algorithm, study obtain the conditional probability distribution of each network node；Gained conditional probability table is assigned and is determined
Bayesian Network Topology Structures, establish the Bayesian network of Collision Accidents of Ships；
The model construction unit, for the method according to joint tree reasoning, according to the pattra leaves of the Collision Accidents of Ships
This network struction goes out Bayesian Network Inference model, obtains Collision Accidents of Ships consequence prediction model；
The predicting unit is used for using marine incident grade as query interface, pre- according to the Collision Accidents of Ships consequence
Survey the probability distribution that model extrapolates damage sequence.
Data cleansing unit, for the method using cluster, by the stronger spy of correlation in the Bayesian network node
Sign is polymerized to cluster one by one, and the elemental characteristic in each cluster is similar characteristic, and cluster is considered as abstract node；
Each cluster that feature clustering is obtained, complete class in multiple features to an abstract characteristics codomain mapping.It will not
The feature of cluster merges with abstract characteristics, generates new data set, as the Bayesian network node after clustering processing.
It is realized by constructing Collision Accidents of Ships prediction model to Collision Accidents of Ships using technical solution of the present invention
Grade forecast the connection between accident factor is excavated by depth, is found between accident factor from the angle of data
Rule realizes the foundation of Collision Accidents of Ships Bayesian network instead of the too strong expertise method of subjectivity.From the angle of method
It sets out, Collision Accidents of Ships prediction model is constructed based on Bayesian network, the forecasting problem of Collision Accidents of Ships grade is solved, pushes away
The development of dynamic maritime affairs Risk Theory.From the angle of application, research achievement can be that can meet ship's navigation commander and Emergency decision
Scientific basis is provided.
Below by drawings and examples, technical scheme of the present invention will be described in further detail.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with reality of the invention
It applies example to be used to explain the present invention together, not be construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the Collision Accidents of Ships grade prediction technique principle process based on Bayesian network in the embodiment of the present invention one
Figure；
Fig. 2 is the Collision Accidents of Ships grade forecast technology path signal in the embodiment of the present invention one based on Bayesian network
Figure；
Fig. 3 is the Collision Accidents of Ships grade forecast system structure signal in the embodiment of the present invention two based on Bayesian network
Figure.
Specific embodiment
Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings, it should be understood that preferred reality described herein
Apply example only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention.
Bayesian network method is the ideal and effective tool of processing uncertain problem developed in recent years, and
It is widely used in multiple fields such as data mining, intelligent decision, pattern-recognition, medical diagnosis.Bayesian network
It is a kind of graphical model of probability dependency between intuitively expressing variable, by network structure and conditional probability distribution two parts group
At indicating the structure attribute between research object stochastic variable with directed acyclic graph, node corresponds to the stochastic variable in model, section
Directed edge between point represents the condition dependence of variable, and generates the conditional probability value between node.Bayesian network being capable of shape
As the causal correlation between expression stochastic variable, and carry out Uncertain knowledge reasoning.
In recent years, Bayesian network was also applied in traffic accident analysis field on the water.Ren etc. is for marine boat
People and organizational factor in row safety accident establish Bayesian network cause and effect in evaluation frame foundation by expert judgement method
Analysis model.Hnninen etc. passes through Research of Bayesian Network Model human factor pair using Finland's gulf ship collision as research object
The influence of ship collision probability detects the variation feelings of the probability of cause by the variation of human factor node state in observation network
Condition, and show that the avoiding operation unreasonable when two ships can be met is that initiation ship collision is most important by sensitivity analysis
Reason.Chen Yadong etc. constructs inland navigation craft by the domain knowledge of expert group and collides Bayesian network analysis model, and combines and examine
Break reasoning and causal reasoning, the cause and effect dependence between quantitative analysis human factor.
These are a kind of analysis Accident-causing factors research shows that Bayesian network model possesses very high precision of prediction
Effective means.But these bayesian network structures are often based on expertise foundation, as a result subjective.Such as
What establishes more objective, reasonable bayesian network structure, to obtain more effectively as a result, being Bayesian network in accident point
Analyse a critical issue in application study.
The embodiment of the present invention aims at the research to Collision Accidents of Ships grade prediction technique, solves maritime affairs risk investigation
The forecasting problem of middle damage sequence, auxiliary maritime control personnel formulate the Decision of Collision Avoidance that can meet ship, reach reduction ship collision
Accident quantity, the purpose for reducing Collision Accidents of Ships disaster.Historical data based on accident establishes Collision Accidents of Ships Bayes
Network.Based on Bayesian Network Inference algorithm, constructs Collision Accidents of Ships grade forecast model and verified.
Fig. 1 is the Collision Accidents of Ships grade prediction technique principle process based on Bayesian network in the embodiment of the present invention one
Figure.As shown in Figure 1, should Collision Accidents of Ships grade forecast process based on Bayesian network the following steps are included:
Step 101 obtains maritime accidents forecast survey report, therefrom obtains accident factor data.
Technology path of the invention is as shown in Figure 2.By maritime accidents forecast survey report Database Systems, maritime affairs are collected
Office's casualty data.Maritime accidents forecast survey report generally comprises: ship or floating facility overview, affiliated company's situation, accident
Occur, search and rescue, damaed cordition, accident process, the analysis of causes, party's confirmation of responsibility, the contents such as safety management suggestion.Pass through thing
Therefore survey report, the essential information of available Collision Accidents of Ships.But due to many accident analysis reports record it is endless
Whole, the shortage of data of some factors is serious, is not used to statistical law.In addition, accident can not be obtained from accident report
When driver behavior and psychology, physiological characteristic, therefore the human factor that can not be used in research accident.The present invention mainly selects
It is for statistical analysis with following accident factor:
Accidents Disasters consequence factor: dangerous situation rank (incident classification).
Environmental factor: time of casualty, wind-force, visibility, water surface situation, temperature, water temperature.
Ship factor: Description of Ship, captain, the beam, the age of a ship, gross tonnage, loading, crewman's number, all properties of ship.
Further, Collision Accidents of Ships essential information data are arranged, for error logging present in report database
Data judge the whether accurate confidence level of a certain data using evidence theory, and then clean to the data of mistake, guarantee number
According to accuracy.And ship static database is utilized, ship basic static information is proofreaded and supplemented.
The accident factor data are carried out preliminary sliding-model control according to Joint Distribution feature by step 102, by each influence
Factor and damage sequence element carry out sensitivity analysis, merge some discrete set of each influence factor, obtain discretization collection
Close as a result, as Bayesian network node.
Because the value of node should be discrete value, if node variable be continuous variable or attribute variable, need to variable into
Row sliding-model control.The continuous data of accident factor is subjected to preliminary sliding-model control according to the Joint Distribution feature of data,
Then each influence factor and damage sequence element are subjected to sensitivity analysis, merge some discrete set of each influence factor,
Obtain the result of final discretization set.
Further, since Collision Accidents of Ships factor type is numerous, network structure is excessively complicated.Using the side of cluster
The stronger feature of correlation is polymerized to cluster one by one by method, and the elemental characteristic in each cluster is similar characteristic, cluster is considered as abstract
Node.Each cluster that feature clustering is obtained, complete class in multiple features to an abstract characteristics codomain mapping.It will not cluster
Feature merge with abstract characteristics, generate new data set.To reduce the node of Bayesian network, to reduce the complexity of modeling
Degree.
Step 103, based between the correlation analysis variable, model the empirically determined Bayesian network section in conjunction with expert
The sequence of point；Machine language is write, bayesian network structure learning is carried out；Bayesian network is obtained according to K2 structure learning algorithm
Structure.
Study the determination method of Bayesian Network Topology Structures.Since data sample information is complete in the present embodiment, model
Interstitial content is more, thus more particularly suitable using the method based on scoring search in the study of model structure.In score function
Selection on, choice accuracy and the fitting effect more preferably BD score function based on Bayesian statistics.In search strategy selection, choosing
It selects and commonly uses efficient K2 searching algorithm to find optimal model structure.Firstly, it is necessary to the sequence of given network node.Network section
What the sequence of point was reacted is a kind of measurement sequence of relation of interdependence between node variable.The present embodiment is based between variable
Correlation analysis models experience in conjunction with expert to determine the sequence of Bayesian network node.Secondly, being write according to relevant algorithm
Machine language carries out the study of network structure by machine tools.By the Bayesian Networks Toolbox of MATLAB software, call
The bayesian network structure of K2 structure learning algorithm acquisition data set.
Step 104, according to the bayesian network structure, input corresponding prior information, with Bayes's parameter learning calculate
Method, study obtain the conditional probability distribution of each network node；Gained conditional probability table is assigned to determining Bayesian network topology
Structure establishes the Bayesian network of Collision Accidents of Ships.
Since the present embodiment data sample information is complete, chooses Bayes's parametric learning method and carry out the node parameter to network
Learnt, wherein the value of node variable is parameter prior distribution.By the relevant machine language of MATLAB software programming.It will
The data set put in order imports, and then the structure composition according to the Bayesian network obtained, inputs corresponding prior information such as parameter
Prior distribution, with Bayes's Parameter Learning Algorithm, study obtains the conditional probability distribution of each network node.By gained condition
Probability tables assign determining Bayesian Network Topology Structures, just complete building for the Bayesian network of entire Collision Accidents of Ships
It is vertical.
Step 105, the method that reasoning is set according to joint, by the joint tree inference engine in the tool box MATLAB, according to institute
The Bayesian network for stating Collision Accidents of Ships constructs Bayesian Network Inference model, show that Collision Accidents of Ships consequence predicts mould
Type.
Step 106, using marine incident grade as query interface, according to the Collision Accidents of Ships consequence prediction model calculate
The probability distribution for consequence that accidents happened.
Present invention contemplates that query interface is set as incident classification, it, can foundation under the conditions of giving the value of its dependent variable
Inference pattern extrapolates the probability distribution of damage sequence, that is, predicts the consequence of accident.And from the study precision of Bayesian network, shellfish
The levels such as the prediction accuracy of this network model of leaf to carry out validation verification to the accident consequence prediction model of building.
Bayesian network is the probability mould of a kind of relationship described between variable using directed acyclic graph and condition degree of dependence
Type.One complete Bayesian network of building is mainly made of three aspects: determining each node in network；Pass through Structure learning
It determines dependence between each node, network node is formed into a directed acyclic graph and determines network topology structure；And pass through
Parameter learning quantifies the dependence incidence relation node to determine each node probability distribution.Then, it is based on Bayesian network
Network could finally establish prediction model by suitable reasoning algorithm.Therefore, technology path of the invention is as shown in Figure 2.
Collision Accidents of Ships is a complex process, as a result subjective if only establishing network structure by hand with expertise
Property is too strong.If it is single with historical data come construct network structure can be relatively inflexible, calculation amount is huge, hardly results in effective result.
Therefore, the Baysian network structure construction method that the present invention uses expertise to combine with machine learning, builds in machine learning
Expertise is added during mould, jointly come complete model Bayesian network building.It is carried out in application K2 searching algorithm
Before the search study of bayesian network structure, the sequence of network node need to be given in advance.Based between the correlation analysis variable,
Experience is modeled in conjunction with expert to determine the sequence of Bayesian network node.This method can not only accelerate Bayesian network in machine
Efficiency in study, moreover it is possible to reduce there is a situation where wrong in modeling process, to improve the accuracy of model.As shown in figure 3,
For a kind of Collision Accidents of Ships grade forecast system based on Bayesian network provided in an embodiment of the present invention, which is characterized in that
Including data input cell 201, endpoint processing unit 202, topological determination unit 203, probability distribution determination unit 204, model
Construction unit 205 and predicting unit 206, wherein
The data input cell 201 therefrom obtains accident factor number for obtaining maritime accidents forecast survey report
According to；
The endpoint processing unit 202, for by the accident factor data according to Joint Distribution feature carry out tentatively from
Each influence factor and damage sequence element are carried out sensitivity analysis, merge some discrete of each influence factor by dispersion processing
Set, obtain discretization set as a result, as Bayesian network node；
The topology determination unit 203, for being modeled in conjunction with expert empirically determined based between the correlation analysis variable
The sequence of the Bayesian network node；Machine language is write, bayesian network structure learning is carried out；It is calculated according to K2 Structure learning
Method obtains bayesian network structure；
The probability distribution determination unit 204, for inputting corresponding prior information according to the bayesian network structure,
With Bayes's Parameter Learning Algorithm, study obtains the conditional probability distribution of each network node；Gained conditional probability table is assigned
Determining Bayesian Network Topology Structures establish the Bayesian network of Collision Accidents of Ships；
The model construction unit 205, for the method according to joint tree reasoning, according to the shellfish of the Collision Accidents of Ships
This network struction of leaf goes out Bayesian Network Inference model, obtains Collision Accidents of Ships consequence prediction model；
The predicting unit 206 is used for using marine incident grade as query interface, according to the Collision Accidents of Ships consequence
Prediction model extrapolates the probability distribution of damage sequence.
Further, above-described embodiment further include:
Data cleansing unit 207, it is for the method using cluster, correlation in the Bayesian network node is stronger
Feature is polymerized to cluster one by one, and the elemental characteristic in each cluster is similar characteristic, and cluster is considered as abstract node；
Each cluster that feature clustering is obtained, complete class in multiple features to an abstract characteristics codomain mapping.It will not
The feature of cluster merges with abstract characteristics, generates new data set, as the Bayesian network node after clustering processing.
The each embodiment of the present invention, by constructing Collision Accidents of Ships prediction model, realize to Collision Accidents of Ships etc.
Grade prediction excavates the connection between accident factor by depth, finds the rule between accident factor from the angle of data
Rule realizes the foundation of Collision Accidents of Ships Bayesian network instead of the too strong expertise method of subjectivity.Go out from the angle of method
Hair constructs Collision Accidents of Ships prediction model based on Bayesian network, solves the forecasting problem of Collision Accidents of Ships grade, pushes
The development of maritime affairs Risk Theory.From the angle of application, research achievement can be mentioned for that can meet ship's navigation commander with Emergency decision
For scientific basis.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the present invention
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the present invention, which can be used in one or more,
The shape for the computer program product implemented in usable storage medium (including but not limited to magnetic disk storage and optical memory etc.)
Formula.
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (10)
1. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network, which comprises the following steps:
Maritime accidents forecast survey report is obtained, accident factor data are therefrom extracted；
The accident factor data are subjected to preliminary sliding-model control according to Joint Distribution feature, after each influence factor and accident
Fruit element carry out sensitivity analysis, merge some discrete set of each influence factor, obtain discretization set as a result, conduct
Bayesian network node；
Based between the correlation analysis variable, the sequence of the empirically determined Bayesian network node is modeled in conjunction with expert；It compiles
Machine language is write, bayesian network structure learning is carried out；Bayesian network structure is obtained according to K2 structure learning algorithm；
According to the bayesian network structure, corresponding prior information is inputted, with Bayes's Parameter Learning Algorithm, study obtains each
The conditional probability distribution of network node；Gained conditional probability table is assigned to determining Bayesian Network Topology Structures, establishes ship
The Bayesian network of collision accident；
According to the method for joint tree reasoning, Bayesian Network Inference is constructed according to the Bayesian network of the Collision Accidents of Ships
Model obtains Collision Accidents of Ships consequence prediction model；
Using marine incident grade as query interface, damage sequence is extrapolated according to the Collision Accidents of Ships consequence prediction model
Probability distribution.
2. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, the accident factor data, comprising:
Accidents Disasters consequence factor: incident classification；
Environmental factor: time of casualty, wind-force, visibility, water surface situation, temperature and/or water temperature；
Ship factor: Description of Ship, captain, the beam, the age of a ship, gross tonnage, loading, crewman's number and/or all properties of ship.
3. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, the method also includes:
Study precision, the prediction accuracy level of Bayesian network model from Bayesian network, to the ship collision thing of building
Therefore consequence prediction model carries out validation verification.
4. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, further includes data cleansing step after the acquisition accident factor data, specific as follows:
Collision Accidents of Ships essential information data are arranged, for incorrect recording data present in the accident analysis report, root
The whether accurate confidence level of a certain data is judged according to evidence theory, and then the data of mistake are cleaned；
Ship basic static information is proofreaded and supplemented using ship static data.
5. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, during the acquisition bayesian network structure, further includes:
Using the method based on scoring search in the study of the bayesian network structure.
6. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, the Bayesian network node, further includes following treatment process:
Using the method for cluster, the stronger feature of correlation in the Bayesian network node is polymerized to cluster one by one, each cluster
Interior elemental characteristic is all similar characteristic, and cluster is considered as abstract node；
Each cluster that feature clustering is obtained, complete class in multiple features to an abstract characteristics codomain mapping.It will not cluster
Feature merge with abstract characteristics, new data set is generated, as the Bayesian network node after clustering processing.
7. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, further includes:
Before carrying out the search study of bayesian network structure according to K2 searching algorithm, the sequence of network node is first given；
It is ranked up using the BD score function based on Bayesian statistics.
8. a kind of Collision Accidents of Ships grade prediction technique based on Bayesian network according to claim 1, feature
It is, further includes:
Based between the correlation analysis variable, experience is modeled in conjunction with expert to determine the sequence of Bayesian network node.
9. a kind of Collision Accidents of Ships grade forecast system based on Bayesian network, which is characterized in that inputted including data single
Member, endpoint processing unit, topological determination unit, probability distribution determination unit, model construction unit and predicting unit, wherein
The data input cell therefrom obtains accident factor data for obtaining maritime accidents forecast survey report；
The endpoint processing unit, for carrying out the accident factor data at preliminary discretization according to Joint Distribution feature
Each influence factor and damage sequence element are carried out sensitivity analysis, merge some discrete set of each influence factor, obtain by reason
To discretization set as a result, as Bayesian network node；
The topology determination unit, for modeling the empirically determined shellfish in conjunction with expert based between the correlation analysis variable
The sequence of this network node of leaf；Machine language is write, bayesian network structure learning is carried out；It is obtained according to K2 structure learning algorithm
Bayesian network structure；
The probability distribution determination unit, for corresponding prior information being inputted, with pattra leaves according to the bayesian network structure
This Parameter Learning Algorithm, study obtain the conditional probability distribution of each network node；Gained conditional probability table is assigned to determining shellfish
This network topology structure of leaf, establishes the Bayesian network of Collision Accidents of Ships；
The model construction unit, for the method according to joint tree reasoning, according to the Bayesian network of the Collision Accidents of Ships
Network constructs Bayesian Network Inference model, obtains Collision Accidents of Ships consequence prediction model；
The predicting unit, for predicting mould according to the Collision Accidents of Ships consequence using marine incident grade as query interface
Type extrapolates the probability distribution of damage sequence.
10. a kind of Collision Accidents of Ships grade forecast system based on Bayesian network according to claim 9, feature
It is, further includes:
Data cleansing unit gathers the stronger feature of correlation in the Bayesian network node for the method using cluster
At cluster one by one, the elemental characteristic in each cluster is similar characteristic, and cluster is considered as abstract node；
Each cluster that feature clustering is obtained, complete class in multiple features to an abstract characteristics codomain mapping.It will not cluster
Feature merge with abstract characteristics, new data set is generated, as the Bayesian network node after clustering processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910021019.7A CN109657880A (en) | 2019-01-09 | 2019-01-09 | A kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910021019.7A CN109657880A (en) | 2019-01-09 | 2019-01-09 | A kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109657880A true CN109657880A (en) | 2019-04-19 |
Family
ID=66119432
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910021019.7A Pending CN109657880A (en) | 2019-01-09 | 2019-01-09 | A kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109657880A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503303A (en) * | 2019-07-22 | 2019-11-26 | 中国人民解放军国防科技大学 | cloud-Bayesian network-based intelligent macro site selection method for offshore wind farm |
| CN110598969A (en) * | 2019-06-25 | 2019-12-20 | 大连海事大学 | Offshore channel emergency risk early warning method |
| CN110930024A (en) * | 2019-11-20 | 2020-03-27 | 大连海事大学 | Intelligent deep sea emergency situation analysis system and method based on Bayesian network |
| CN111275329A (en) * | 2020-01-20 | 2020-06-12 | 交通运输部水运科学研究所 | Water transport engineering construction safety accident cause analysis method |
| CN111553562A (en) * | 2020-04-07 | 2020-08-18 | 哈尔滨工程大学 | Ship collision risk degree estimation method based on improved BP neural network |
| CN112036653A (en) * | 2020-09-07 | 2020-12-04 | 江苏金鸽网络科技有限公司 | Fire risk early warning method and system based on Bayesian network |
| CN112330161A (en) * | 2020-11-09 | 2021-02-05 | 中国民航大学 | Logistics unmanned aerial vehicle failure risk assessment method based on Bayesian network |
| CN112465304A (en) * | 2020-11-07 | 2021-03-09 | 西南交通大学 | Railway turnout area train derailment accident assessment method based on Bayesian network |
| CN112949999A (en) * | 2021-02-04 | 2021-06-11 | 浙江工业大学 | High-speed traffic accident risk early warning method based on Bayesian deep learning |
| CN113657599A (en) * | 2021-08-20 | 2021-11-16 | 北京航空航天大学 | Accident cause and effect reasoning method and device, electronic equipment and readable storage medium |
| CN114662575A (en) * | 2022-03-09 | 2022-06-24 | 集美大学 | Wind power water area ship navigation risk estimation method and system and storage medium |
| CN114662979A (en) * | 2022-04-12 | 2022-06-24 | 西南交通大学 | BN-ISM model-based railway traffic accident early warning method and system |
| CN115329857A (en) * | 2022-08-04 | 2022-11-11 | 武汉理工大学 | Inland river navigation water area grading method and device, electronic equipment and storage medium |
| CN115659263A (en) * | 2022-10-14 | 2023-01-31 | 长江三峡通航管理局 | Ship control behavior risk assessment system and assessment method based on big data |
-
2019
- 2019-01-09 CN CN201910021019.7A patent/CN109657880A/en active Pending
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598969A (en) * | 2019-06-25 | 2019-12-20 | 大连海事大学 | Offshore channel emergency risk early warning method |
| CN110598969B (en) * | 2019-06-25 | 2023-03-31 | 大连海事大学 | Offshore channel emergency risk early warning method |
| CN110503303A (en) * | 2019-07-22 | 2019-11-26 | 中国人民解放军国防科技大学 | cloud-Bayesian network-based intelligent macro site selection method for offshore wind farm |
| CN110930024A (en) * | 2019-11-20 | 2020-03-27 | 大连海事大学 | Intelligent deep sea emergency situation analysis system and method based on Bayesian network |
| CN110930024B (en) * | 2019-11-20 | 2024-04-26 | 大连海事大学 | Intelligent analysis system and method for deep sea emergency situation based on Bayesian network |
| CN111275329A (en) * | 2020-01-20 | 2020-06-12 | 交通运输部水运科学研究所 | Water transport engineering construction safety accident cause analysis method |
| CN111553562A (en) * | 2020-04-07 | 2020-08-18 | 哈尔滨工程大学 | Ship collision risk degree estimation method based on improved BP neural network |
| CN112036653A (en) * | 2020-09-07 | 2020-12-04 | 江苏金鸽网络科技有限公司 | Fire risk early warning method and system based on Bayesian network |
| CN112465304A (en) * | 2020-11-07 | 2021-03-09 | 西南交通大学 | Railway turnout area train derailment accident assessment method based on Bayesian network |
| CN112330161A (en) * | 2020-11-09 | 2021-02-05 | 中国民航大学 | Logistics unmanned aerial vehicle failure risk assessment method based on Bayesian network |
| CN112949999A (en) * | 2021-02-04 | 2021-06-11 | 浙江工业大学 | High-speed traffic accident risk early warning method based on Bayesian deep learning |
| CN113657599A (en) * | 2021-08-20 | 2021-11-16 | 北京航空航天大学 | Accident cause and effect reasoning method and device, electronic equipment and readable storage medium |
| CN113657599B (en) * | 2021-08-20 | 2024-05-28 | 北京航空航天大学 | Accident cause and effect reasoning method, device, electronic equipment and readable storage medium |
| CN114662575A (en) * | 2022-03-09 | 2022-06-24 | 集美大学 | Wind power water area ship navigation risk estimation method and system and storage medium |
| CN114662575B (en) * | 2022-03-09 | 2024-06-14 | 集美大学 | Wind power water area ship navigation risk prediction method, system and storage medium |
| CN114662979A (en) * | 2022-04-12 | 2022-06-24 | 西南交通大学 | BN-ISM model-based railway traffic accident early warning method and system |
| CN115329857B (en) * | 2022-08-04 | 2024-02-09 | 武汉理工大学 | Inland navigation water area grade division method and device, electronic equipment and storage medium |
| CN115329857A (en) * | 2022-08-04 | 2022-11-11 | 武汉理工大学 | Inland river navigation water area grading method and device, electronic equipment and storage medium |
| CN115659263A (en) * | 2022-10-14 | 2023-01-31 | 长江三峡通航管理局 | Ship control behavior risk assessment system and assessment method based on big data |
| CN115659263B (en) * | 2022-10-14 | 2023-08-08 | 长江三峡通航管理局 | Ship control behavior risk assessment system and method based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109657880A (en) | A kind of Collision Accidents of Ships grade prediction technique and system based on Bayesian network | |
| CN110633855B (en) | Bridge health state detection and management decision making system and method | |
| Yu et al. | An integrated dynamic ship risk model based on Bayesian Networks and Evidential Reasoning | |
| Zhang et al. | A systematic approach for collision risk analysis based on AIS data | |
| Zhang et al. | A novel case adaptation method based on an improved integrated genetic algorithm for power grid wind disaster emergencies | |
| Li et al. | AIS data-based decision model for navigation risk in sea areas | |
| CN110636066B (en) | Network security threat situation assessment method based on unsupervised generative reasoning | |
| CN115063020B (en) | Multi-dimensional safety scheduling device and method for cascade hydropower station based on risk monitoring fusion | |
| Kang et al. | Modeling and evaluation of the oil-spill emergency response capability based on linguistic variables | |
| Huang et al. | A review on risk assessment methods for maritime transport | |
| CN111613094A (en) | Port water area ship traffic risk early warning method | |
| Bakdi et al. | Fullest COLREGs evaluation using fuzzy logic for collaborative decision-making analysis of autonomous ships in complex situations | |
| Lau et al. | A fuzzy-based decision support model for engineering asset condition monitoring–A case study of examination of water pipelines | |
| Zhang et al. | A Bayesian network-based model for risk modeling and scenario deduction of collision accidents of inland intelligent ships | |
| Liu et al. | A novel model for identifying the vessel collision risk of anchorage | |
| Sylaios et al. | A fuzzy inference system for wind-wave modeling | |
| Hadi et al. | Achieving fuel efficiency of harbour craft vessel via combined time-series and classification machine learning model with operational data | |
| CN113626929B (en) | Multi-stage multi-topology ship traffic complexity measurement method and system | |
| Booker et al. | Probabilistic reasoning about ship images | |
| Fiskin | An advanced decision-making model for determining ship domain size with a combination of MCDM and fuzzy logic | |
| CN104778633B (en) | A kind of pumping equipment safeguard protection hierarchical intelligence determination method | |
| Sakar et al. | Dynamic analysis of pilot transfer accidents | |
| Tu et al. | Evaluation of seawater quality in hangzhou bay based on TS fuzzy neural network | |
| CN113469504A (en) | Dynamic testing method, medium and system for water traffic risk | |
| CN113902327A (en) | Evaluation method and system for corrosion health state of offshore wind plant foundation structure |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190419 |