CN107085904B - Forest fire danger class determination method and system based on single classification SVM - Google Patents
Forest fire danger class determination method and system based on single classification SVM Download PDFInfo
- Publication number
- CN107085904B CN107085904B CN201710209365.9A CN201710209365A CN107085904B CN 107085904 B CN107085904 B CN 107085904B CN 201710209365 A CN201710209365 A CN 201710209365A CN 107085904 B CN107085904 B CN 107085904B
- Authority
- CN
- China
- Prior art keywords
- forest fire
- risk
- sample
- fire
- probability
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B17/00—Fire alarms; Alarms responsive to explosion
- G08B17/005—Fire alarms; Alarms responsive to explosion for forest fires, e.g. detecting fires spread over a large or outdoors area
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B31/00—Predictive alarm systems characterised by extrapolation or other computation using updated historic data
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Emergency Management (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biodiversity & Conservation Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of forest fire danger class determination method and system based on single classification SVM, comprising the following steps: using day as sample unit, chooses the sample that fire occurs as modeling sample according to fire data；Obtain the corresponding meteorological factor of the modeling sample；Based on the corresponding meteorological factor of the modeling sample, single classification SVM model is constructed；Risk of forest fire occurrence Probability Model is constructed, i.e., is mapped to the intermediate sample exported value interval of the distance of the hypersphere centre of sphere into model of single classification SVM model [0,1] using activation primitive, mapping result is risk of forest fire probability of happening；The risk of forest fire probability of happening of sample to be tested is calculated, and forest fire danger class is determined according to the risk of forest fire probability of happening of the sample to be tested.Of the invention is effectively overcome based on singly the forest fire danger class determination method of classification SVM and system due to there is class imbalance in forest fire sample set, and the accuracy of the judgement of risk of forest fire is improved.
Description
Technical field
The present invention relates to a kind of grade determination method and systems, more particularly to a kind of Forest Fire based on single classification SVM
Dangerous grade determination method and system.
Background technique
Forest fire is a kind of sudden strong, destructive natural calamity big, disposition relief is more difficult, high destructive
Determine the importance of forest fire prevention and control.Therefore, how scientific and effective to forest fire progress early-warning and predicting, to greatest extent
Reduction forest fire generation and by the loss of forest fire bring be always forest department, China and scientific research department very
The problem of concern.
Risk of forest fire is a kind of measurement of a possibility that forest fire occurs and sprawling easy degree.The hair of forest fire
Life, development are closely related with meteorological condition.Therefore carry out forest fire danger forecasting work and be unable to do without element of making weather observations in real time, it is can
The comprehensive function of combustion things and the contextual integration factor.
Forest fire danger class with it is a variety of factor-related.In the prior art, to the prediction of fire size class mainly according to meteorology
The factors such as the factor and combustible situation carry out.Wherein, meteorological factor includes air themperature, relative humidity, illumination, precipitation, wind
Speed etc.,.
In forest fire field, traditional prediction of forest fire disaster method is usually the meteorological data chosen in a period of time,
Using day as sample unit, fire is occurred to each sample mark ' being ' or 'No'；Then all samples are sent into two classifiers
It is trained, obtains a fire prediction model.But due to the particularity of forest fire, the positive sample number that fire occurs is tight
Again less than the negative sample number that fire does not occur, leading to the sample set, there are serious class imbalances.The machine learning of standard
For classification method when handling unbalanced data classification problem, the comprehensive tendency multisample class of classification judgement leads to few sample class classification
Precision is very low.And in forest fire protection, the positive sample as few sample is only the sample of real concern.Therefore, existing forest
Fire prediction technique is unable to reach the requirement to prediction of forest fire disaster accuracy.
Summary of the invention
In view of the foregoing deficiencies of prior art, the purpose of the present invention is to provide a kind of based on the gloomy of single classification SVM
Woods fire size class determination method and system are utilized using several meteorological factors as the impact factor for judging forest fire danger class
The sample that fire does not occur carries out the study of risk of forest fire occurrence Probability Model based on single classification SVM, and then determines risk of forest fire
Grade effectively overcomes due to there is class imbalance in forest fire sample set, improves sentencing for risk of forest fire
Fixed accuracy.
In order to achieve the above objects and other related objects, the present invention provides a kind of risk of forest fire etc. based on single classification SVM
Grade determination method, comprising the following steps: using day as sample unit, the sample that fire occurs is chosen according to fire data as modeling
Sample；Obtain the corresponding meteorological factor of the modeling sample；Based on the corresponding meteorological factor of the modeling sample, single classification is constructed
SVM model；Risk of forest fire occurrence Probability Model is constructed, i.e., is exported single the intermediate of classification SVM model using activation primitive
Sample into model the value interval of the distance of the hypersphere centre of sphere map to [0,1], mapping result is risk of forest fire
Probability；The risk of forest fire probability of happening of sample to be tested is calculated, and is determined according to the risk of forest fire probability of happening of the sample to be tested
Forest fire danger class.
In one embodiment of the invention, the meteorological factor includes intra day ward, daily maximum temperature, daily minimal tcmperature, day
Average relative humidity, per day wind speed, precipitation yesterday, first three everyday precipitation aggregate value, first three days relative humidity average value,
First three days temperature aggregate value, before today when continuous number of days of the precipitation less than or equal to 5 millimeters, 20 before intra day ward be less than etc.
Intra day ward is less than or equal to 0.5 millimeter of continuous number of days before when 3 millimeters of continuous numbers of days, 20.
In one embodiment of the invention, when constructing single classification SVM model, Waikato intellectual analysis environmental level is used
Single classification SVM model, using default parameters.
In one embodiment of the invention, the activation primitive is used
In one embodiment of the invention, when determining forest fire danger class according to the risk of forest fire probability of happening of sample to be tested,
Follow following principle:
When risk of forest fire probability of happening is at section [0,0.2], judgement forest fire danger class is fire level-one；
When risk of forest fire probability of happening section (0.2,0.4] when, determine forest fire danger class be fire second level；
When risk of forest fire probability of happening section (0.4,0.6] when, determine forest fire danger class be fire three-level；
When risk of forest fire probability of happening section (0.6,0.8] when, determine forest fire danger class be fire level Four；
When risk of forest fire probability of happening section (0.8,1] when, determine forest fire danger class be fire Pyatyi.
Meanwhile the present invention also provides a kind of forest fire danger class decision-making systems based on single classification SVM, including choose mould
Block obtains module, single classification SVM model construction module, risk of forest fire occurrence Probability Model building module and determination module；
The selection module is used to choose the sample that fire occurs according to fire data using day as sample unit as modeling
Sample；
The acquisition module is for obtaining the corresponding meteorological factor of modeling sample；
Single classification SVM model construction module is used to be based on the corresponding meteorological factor of modeling sample, constructs single classification SVM
Model；
The risk of forest fire occurrence Probability Model building module is for constructing risk of forest fire occurrence Probability Model, i.e., using sharp
Function living maps to the intermediate sample exported value interval of the distance of the hypersphere centre of sphere into model of single classification SVM model
[0,1], mapping result are risk of forest fire probability of happening；
The determination module is used to calculate the risk of forest fire probability of happening of sample to be tested, and according to the Forest Fire of sample to be tested
Dangerous probability of happening determines forest fire danger class.
In one embodiment of the invention, the meteorological factor includes intra day ward, daily maximum temperature, daily minimal tcmperature, day
Average relative humidity, per day wind speed, precipitation yesterday, first three everyday precipitation aggregate value, first three days relative humidity average value,
First three days temperature aggregate value, before today when continuous number of days of the precipitation less than or equal to 5 millimeters, 20 before intra day ward be less than etc.
Intra day ward is less than or equal to 0.5 millimeter of continuous number of days before when 3 millimeters of continuous numbers of days, 20.
In one embodiment of the invention, when single classification SVM model construction module constructs single classification SVM model,
Using single classification SVM model of Waikato intellectual analysis environmental level, using default parameters.
In one embodiment of the invention, the activation primitive is used
In one embodiment of the invention, the determination module determines forest according to the risk of forest fire probability of happening of sample to be tested
When fire size class, it then follows following principle:
When risk of forest fire probability of happening is at section [0,0.2], judgement forest fire danger class is fire level-one；
When risk of forest fire probability of happening section (0.2,0.4] when, determine forest fire danger class be fire second level；
When risk of forest fire probability of happening section (0.4,0.6] when, determine forest fire danger class be fire three-level；
When risk of forest fire probability of happening section (0.6,0.8] when, determine forest fire danger class be fire level Four；
When risk of forest fire probability of happening section (0.8,1] when, determine forest fire danger class be fire Pyatyi.
As described above, the forest fire danger class determination method and system of the invention based on single classification SVM, has with following
Beneficial effect:
(1) using several meteorological factors as the impact factor for judging forest fire danger class, the sample that fire does not occur is utilized
The study of risk of forest fire occurrence Probability Model is carried out based on single classification SVM, and then determines forest fire danger class；
(2) it effectively overcomes due to there is class imbalance in forest fire sample set；
(3) judgement that forest fire danger class is carried out from the angle of fire probability, improves the judgement of risk of forest fire
Accuracy, with more science.
Detailed description of the invention
Fig. 1 is shown as the flow chart of the forest fire danger class determination method of the invention based on single classification SVM；
Fig. 2 is shown as the structural schematic diagram of the forest fire danger class decision-making system of the invention based on single classification SVM.
Component label instructions
1 chooses module
2 obtain module
3 single classification SVM model construction modules
4 risk of forest fire occurrence Probability Models construct module
5 determination modules
Specific embodiment
Illustrate embodiments of the present invention below by way of specific specific example, those skilled in the art can be by this specification
Other advantages and efficacy of the present invention can be easily understood for disclosed content.The present invention can also pass through in addition different specific realities
The mode of applying is embodied or practiced, the various details in this specification can also based on different viewpoints and application, without departing from
Various modifications or alterations are carried out under spirit of the invention.It should be noted that in the absence of conflict, following embodiment and implementation
Feature in example can be combined with each other.
It should be noted that illustrating the basic structure that only the invention is illustrated in a schematic way provided in following embodiment
Think, only shown in schema then with related component in the present invention rather than component count, shape and size when according to actual implementation
Draw, when actual implementation kenel, quantity and the ratio of each component can arbitrarily change for one kind, and its assembly layout kenel
It is likely more complexity.
In machine learning field, support vector machines (Support Vector Machine, SVM) is have supervision
Model is practised, commonly used to carry out pattern-recognition, classification and regression analysis.The main thought of SVM may be summarized to be following:
1) linear can a point situation analyzed；It the case where for linearly inseparable, is calculated by using Nonlinear Mapping
The sample of low-dimensional input space linearly inseparable is converted high-dimensional feature space by method makes its linear separability, so that higher-dimension is special
Sign space carries out linear analysis using nonlinear characteristic of the linear algorithm to sample and is possibly realized；
2) the construction optimum segmentation hyperplane in feature space is based on structural risk minimization theory, so that learner
Global optimization is obtained, and certain upper bound is met with some probability in the expected risk of entire sample space.
Single classification SVM (One-class Support Vector Machine, OCSVM) belongs to discrimination model, is output
It as a result is the hardness judgement of "Yes" or "No".In the present invention, it in order to determine forest fire danger class, realizes and forest fire is occurred
A possibility that and sprawling easy degree carry out the measurement of quantification, using activation primitive by the intermediate output data of OCSVM algorithm
Forest fire probability of happening is converted to, and forest fire danger class is determined according to forest fire probability of happening, by way of quantization
Determine forest fire danger class.
The habit of fire danger prediction is applied in order to adapt to fire prevention direction department and the masses, the fire degree of forest is artificially divided
It is specific as shown in table 1 for five fire size classes.
Table 1, fire size class
Referring to Fig.1, the forest fire danger class determination method of the invention based on single classification SVM the following steps are included:
Step S1, using day as sample unit, the sample that fire occurs is chosen as modeling sample according to fire data.
Specifically, sample was labeled according to fire data, the sample mark of fire will occur for a sample with one day
Note is positive sample, and the sample that fire does not occur is labeled as negative sample.Positive sample is only chosen to sentence as subsequent forest fire danger class
Fixed modeling sample.
Therefore, step S1 the following steps are included:
11) using day as sample unit, the sample that fire occurs is labeled as positive sample, the sample that fire does not occur is marked
For negative sample.
12) positive sample is chosen as modeling sample.
Step S2, the corresponding meteorological factor of modeling sample is obtained.
Specifically, according to meteorological data, meteorological factor corresponding to each sample is obtained.In the present invention, meteorological factor
Including intra day ward, daily maximum temperature, daily minimal tcmperature, per day relative humidity, per day wind speed, precipitation yesterday, first three
Precipitation is less than or equal to before precipitation aggregate value, first three days relative humidity average value, first three days temperature aggregate value, today everyday
When 5 millimeters of continuous number of days, 20 before intra day ward less than or equal to 3 millimeters continuous number of days, 20 when before intra day ward be less than etc.
In 0.5 millimeter of continuous number of days.This impact factor of 12 meteorological factors as forest fire danger class.
Step S3, it is based on the corresponding meteorological factor of modeling sample, constructs single classification SVM model.
Specifically, using platform Waikato Environment for Knowledge Analysis (Weka, Waikato
Intellectual analysis environment) the single classification SVM model of algorithm packet weka.classifiers.functions.LibSVM building.Weka
Be it is a it is free, non-commercialization, based on the machine learning and data mining software increased income under JAVA environment.In Weka
In, parameter SVMType is set as one-class SVM, i.e. SVMType type is selected as single classification SVM, and other parameters use silent
Recognize setting.
Therefore, in step s3, the corresponding single classification SVM algorithm of meteorological factor input of modeling sample is trained, i.e.,
Single classification SVM model can be obtained.
Step S4, construct risk of forest fire occurrence Probability Model, i.e., it is using activation primitive that the centre of single classification SVM model is defeated
The value interval of the distance of the hypersphere centre of sphere into model of sample out maps to [0,1], and mapping result is risk of forest fire hair
Raw probability.
Single classification SVM model output is the rigid court verdict of "Yes" or "No", can not determine forest fire danger class.It is single
Classification SVM algorithm is only trained the sample that fire occurs, and one hypersphere small as far as possible of study is wrapped up all samples
This.When new forecast sample is fallen in outside hypersphere, then determine that fire occurs；When new forecast sample is fallen in inside hypersphere
When, then determine that fire does not occur.Therefore in step S4, the intermediate output result-sample for choosing single classification SVM model surpasses into model
The distance of the spherical surface centre of sphere is as output, and to sample, the distance of the hypersphere centre of sphere into model maps via activation primitive,
With by the value interval of sample distance of the hypersphere centre of sphere into model by [0, ∞) map to [0,1].What is obtained after mapping reflects
Penetrating result is risk of forest fire probability of happening.
Specifically, activation primitive is utilizedTo the sample exported among single classification SVM model to mould
The distance d of the hypersphere centre of sphere is mapped in type, so that the value interval of sample distance d of the hypersphere centre of sphere into model is extremely
[0,1].Mapping result f (d) is risk of forest fire probability of happening.
Step S5, the risk of forest fire probability of happening of sample to be tested is calculated, and is occurred generally according to the risk of forest fire of sample to be tested
Rate determines forest fire danger class.
Specifically, the meteorological factor of sample to be tested is inputted into risk of forest fire occurrence Probability Model, obtains risk of forest fire
Probability, and forest fire danger class is determined according to the risk of forest fire probability of happening.Forest Fire is determined according to risk of forest fire probability of happening
When dangerous grade, it then follows following principle:
When risk of forest fire probability of happening is at section [0,0.2], judgement forest fire danger class is fire level-one；
When risk of forest fire probability of happening section (0.2,0.4] when, determine forest fire danger class be fire second level；
When risk of forest fire probability of happening section (0.4,0.6] when, determine forest fire danger class be fire three-level；
When risk of forest fire probability of happening section (0.6,0.8] when, determine forest fire danger class be fire level Four；
When risk of forest fire probability of happening section (0.8,1] when, determine forest fire danger class be fire Pyatyi.
Referring to Fig. 2, the forest fire danger class decision-making system of the invention based on single classification SVM includes choosing module 1, obtaining
Module 2, single classification SVM model construction module 3, risk of forest fire occurrence Probability Model building module 4 and determination module 5.
It chooses module 1 and is used to choose the sample that fire occurs according to fire data using day as sample unit as modeling sample
This.
Specifically, sample was labeled according to fire data, the sample mark of fire will occur for a sample with one day
Note is positive sample, and the sample that fire does not occur is labeled as negative sample.Positive sample is only chosen to sentence as subsequent forest fire danger class
Fixed modeling sample.
Therefore, it chooses module 1 and executes following operation:
11) using day as sample unit, the sample that fire occurs is labeled as positive sample, the sample that fire does not occur is marked
For negative sample.
12) positive sample is chosen as modeling sample.
It obtains module 2 and is connected with module 1 is chosen, for obtaining the corresponding meteorological factor of modeling sample.
Specifically, according to meteorological data, meteorological factor corresponding to each sample is obtained.In the present invention, meteorological factor
Including intra day ward, daily maximum temperature, daily minimal tcmperature, per day relative humidity, per day wind speed, precipitation yesterday, first three
Precipitation is less than or equal to before precipitation aggregate value, first three days relative humidity average value, first three days temperature aggregate value, today everyday
When 5 millimeters of continuous number of days, 20 before intra day ward less than or equal to 3 millimeters continuous number of days, 20 when before intra day ward be less than etc.
In 0.5 millimeter of continuous number of days.This impact factor of 12 meteorological factors as forest fire danger class.
Single classification SVM model construction module 3 is connected with module 2 is obtained, for being based on the corresponding meteorological factor of modeling sample,
The single classification SVM model of building.
Specifically, using the algorithm of platform Waikato Environment for Knowledge Analysis (Weka)
Wrap the single classification SVM model of weka.classifiers.functions.LibSVM building.Weka is a free, non-commercial
Change, based on the machine learning and data mining software increased income under JAVA environment.In Weka, parameter SVMType is set as
One-class SVM, i.e. SVMType type are selected as single classification SVM, and other parameters use default setting.
Therefore, when the single classification SVM model of building, by the corresponding single classification SVM algorithm of meteorological factor input of modeling sample into
Single classification SVM model can be obtained in row training.
Risk of forest fire occurrence Probability Model constructs module 4 and is connected with single classification SVM model construction module 3, gloomy for constructing
Forest fires danger occurrence Probability Model utilizes the intermediate sample that exports into model hypersphere of the activation primitive by single classification SVM model
The value interval of the distance of the face centre of sphere maps to [0,1], and mapping result is risk of forest fire probability of happening.
Single classification SVM model output is the rigid court verdict of "Yes" or "No", can not determine forest fire danger class.It is single
Classification SVM algorithm is only trained the sample that fire occurs, and one hypersphere small as far as possible of study is wrapped up all samples
This.When new forecast sample is fallen in outside hypersphere, then determine that fire occurs；When new forecast sample is fallen in inside hypersphere
When, then determine that fire does not occur.Therefore in risk of forest fire occurrence Probability Model building module, choose in single classification SVM model
Between export result-sample hypersphere centre of sphere into model distance as output, and via activation primitive to sample into model
The distance of the hypersphere centre of sphere is mapped, by the value interval of sample distance of the hypersphere centre of sphere into model by [0, ∞) reflect
It is incident upon [0,1].The mapping result obtained after mapping is risk of forest fire probability of happening.
Specifically, activation primitive is utilizedTo the sample exported among single classification SVM model to mould
The distance d of the hypersphere centre of sphere is mapped in type, so that the value interval of sample distance d of the hypersphere centre of sphere into model is extremely
[0,1].Mapping result f (d) is risk of forest fire probability of happening.
Determination module 5 is connected with risk of forest fire occurrence Probability Model building module 4, for calculating the Forest Fire of sample to be tested
Dangerous probability of happening, and forest fire danger class is determined according to the risk of forest fire probability of happening of sample to be tested.
Specifically, the meteorological factor of sample to be tested is inputted into risk of forest fire occurrence Probability Model, obtains risk of forest fire
Probability, and forest fire danger class is determined according to the risk of forest fire probability of happening.Forest Fire is determined according to risk of forest fire probability of happening
When dangerous grade, it then follows following principle:
When risk of forest fire probability of happening is at section [0,0.2], judgement forest fire danger class is fire level-one；
When risk of forest fire probability of happening section (0.2,0.4] when, determine forest fire danger class be fire second level；
When risk of forest fire probability of happening section (0.4,0.6] when, determine forest fire danger class be fire three-level；
When risk of forest fire probability of happening section (0.6,0.8] when, determine forest fire danger class be fire level Four；
When risk of forest fire probability of happening section (0.8,1] when, determine forest fire danger class be fire Pyatyi.
In conclusion the forest fire danger class determination method and system of the invention based on single classification SVM is with several gas
As the factor is as the impact factor for judging forest fire danger class, using the sample that fire does not occur be based on single classification SVM carry out it is gloomy
The study of forest fires danger occurrence Probability Model, and then determine forest fire danger class；It effectively overcomes due to forest fire sample set
In and there are problems that class imbalance；The judgement that forest fire danger class is carried out from the angle of fire probability, improves forest
The accuracy of the judgement of fire, with more science.So the present invention effectively overcomes various shortcoming in the prior art and has
High industrial utilization value.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
The personage for knowing this technology all without departing from the spirit and scope of the present invention, carries out modifications and changes to above-described embodiment.Cause
This, institute is complete without departing from the spirit and technical ideas disclosed in the present invention by those of ordinary skill in the art such as
At all equivalent modifications or change, should be covered by the claims of the present invention.
Claims (10)
1. a kind of forest fire danger class determination method based on single classification SVM, it is characterised in that: the following steps are included:
Using day as sample unit, the sample that fire occurs is chosen as modeling sample according to fire data；
Obtain the corresponding meteorological factor of the modeling sample；
Based on the corresponding meteorological factor of the modeling sample, single classification SVM model is constructed；
Construct risk of forest fire occurrence Probability Model, i.e., the sample that will be exported among single classification SVM model using activation primitive
This value interval of the distance of the hypersphere centre of sphere into model maps to [0,1], and mapping result is risk of forest fire probability of happening；
The risk of forest fire probability of happening of sample to be tested is calculated, and gloomy according to the judgement of the risk of forest fire probability of happening of the sample to be tested
Woods fire size class.
2. the forest fire danger class determination method according to claim 1 based on single classification SVM, it is characterised in that: described
Meteorological factor includes intra day ward, daily maximum temperature, daily minimal tcmperature, per day relative humidity, per day wind speed, precipitation yesterday
Amount, precipitation before first three precipitation aggregate value, first three days relative humidity average value, first three days temperature aggregate value, today everyday
Before when continuous number of days less than or equal to 5 millimeters, 20 when continuous number of days of the intra day ward less than or equal to 3 millimeters, 20 before daily precipitation
Amount is less than or equal to 0.5 millimeter of continuous number of days.
3. the forest fire danger class determination method according to claim 1 based on single classification SVM, it is characterised in that: building
When single classification SVM model, the algorithm packet weka.classifiers.funct of Waikato intellectual analysis environmental level is used
The single classification SVM model of ions.LibSVM building, parameter SVMType are set as one-class SVM, and other parameters use default
Setting.
4. the forest fire danger class determination method according to claim 1 based on single classification SVM, it is characterised in that: described
Activation primitive uses
5. the forest fire danger class determination method according to claim 1 based on single classification SVM, it is characterised in that: according to
When the risk of forest fire probability of happening of sample to be tested determines forest fire danger class, it then follows following principle:
When risk of forest fire probability of happening is at section [0,0.2], judgement forest fire danger class is fire level-one；
When risk of forest fire probability of happening section (0.2,0.4] when, determine forest fire danger class be fire second level；
When risk of forest fire probability of happening section (0.4,0.6] when, determine forest fire danger class be fire three-level；
When risk of forest fire probability of happening section (0.6,0.8] when, determine forest fire danger class be fire level Four；
When risk of forest fire probability of happening section (0.8,1] when, determine forest fire danger class be fire Pyatyi.
6. a kind of forest fire danger class decision-making system based on single classification SVM, it is characterised in that: including choosing module, obtaining mould
Block, single classification SVM model construction module, risk of forest fire occurrence Probability Model building module and determination module；
The selection module is used to choose the sample that fire occurs according to fire data using day as sample unit as modeling sample
This；
The acquisition module is for obtaining the corresponding meteorological factor of modeling sample；
Single classification SVM model construction module is used to be based on the corresponding meteorological factor of modeling sample, constructs single classification SVM mould
Type；
The risk of forest fire occurrence Probability Model building module utilizes activation letter for constructing risk of forest fire occurrence Probability Model
Number the intermediate sample exported value interval of the distance of the hypersphere centre of sphere into model of single classification SVM model is mapped to [0,
1], mapping result is risk of forest fire probability of happening；
The determination module is used to calculate the risk of forest fire probability of happening of sample to be tested, and is sent out according to the risk of forest fire of sample to be tested
Raw probability determines forest fire danger class.
7. the forest fire danger class decision-making system according to claim 6 based on single classification SVM, it is characterised in that: described
Meteorological factor includes intra day ward, daily maximum temperature, daily minimal tcmperature, per day relative humidity, per day wind speed, precipitation yesterday
Amount, precipitation before first three precipitation aggregate value, first three days relative humidity average value, first three days temperature aggregate value, today everyday
Before when continuous number of days less than or equal to 5 millimeters, 20 when continuous number of days of the intra day ward less than or equal to 3 millimeters, 20 before daily precipitation
Amount is less than or equal to 0.5 millimeter of continuous number of days.
8. the forest fire danger class decision-making system according to claim 6 based on single classification SVM, it is characterised in that: described
When single classification SVM model construction module constructs single classification SVM model, the algorithm of Waikato intellectual analysis environmental level is used
The single classification SVM model of weka.classifiers.functions.LibSVM building is wrapped, parameter SVMType is set as one-
Class SVM, other parameters use default setting.
9. the forest fire danger class decision-making system according to claim 6 based on single classification SVM, it is characterised in that: described
Activation primitive uses
10. the forest fire danger class decision-making system according to claim 6 based on single classification SVM, it is characterised in that: described
When determination module determines forest fire danger class according to the risk of forest fire probability of happening of sample to be tested, it then follows following principle:
When risk of forest fire probability of happening is at section [0,0.2], judgement forest fire danger class is fire level-one；
When risk of forest fire probability of happening section (0.2,0.4] when, determine forest fire danger class be fire second level；
When risk of forest fire probability of happening section (0.4,0.6] when, determine forest fire danger class be fire three-level；
When risk of forest fire probability of happening section (0.6,0.8] when, determine forest fire danger class be fire level Four；
When risk of forest fire probability of happening section (0.8,1] when, determine forest fire danger class be fire Pyatyi.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710209365.9A CN107085904B (en) | 2017-03-31 | 2017-03-31 | Forest fire danger class determination method and system based on single classification SVM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710209365.9A CN107085904B (en) | 2017-03-31 | 2017-03-31 | Forest fire danger class determination method and system based on single classification SVM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107085904A CN107085904A (en) | 2017-08-22 |
| CN107085904B true CN107085904B (en) | 2019-07-30 |
Family
ID=59615290
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710209365.9A Active CN107085904B (en) | 2017-03-31 | 2017-03-31 | Forest fire danger class determination method and system based on single classification SVM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107085904B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109492928B (en) * | 2018-11-23 | 2021-08-06 | 北京林业大学 | Method and device for constructing forest fire risk prediction model |
| CN109685266A (en) * | 2018-12-21 | 2019-04-26 | 长安大学 | A kind of lithium battery bin fire prediction method and system based on SVM |
| CN110046738A (en) * | 2019-01-28 | 2019-07-23 | 南京林业大学 | A kind of forest fire prediction technique based on artificial intelligence perceptron model |
| CN110163251A (en) * | 2019-04-15 | 2019-08-23 | 深圳市中电数通智慧安全科技股份有限公司 | A kind of Optimum Identification Method of fire hazard rating, device and terminal device |
| CN112132321A (en) * | 2020-08-25 | 2020-12-25 | 航天信德智图（北京）科技有限公司 | Method for predicting and analyzing forest fire based on machine learning |
| CN112990562A (en) * | 2021-03-05 | 2021-06-18 | 中国铁塔股份有限公司黑龙江省分公司 | Forest fire prevention grade prediction algorithm based on community characteristics |
| CN113298311A (en) * | 2021-06-08 | 2021-08-24 | 平安国际智慧城市科技股份有限公司 | Forest fire danger level prediction method, device, equipment and readable storage medium |
| CN117152893B (en) * | 2023-10-31 | 2023-12-29 | 广州市林业和园林科学研究院 | Forest disaster prevention method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011032117A1 (en) * | 2009-09-13 | 2011-03-17 | Delacom Detection Systems, Llc | Method and system for wildfire detection using a visible range camera |
| CN102842199A (en) * | 2012-08-30 | 2012-12-26 | 广州中国科学院工业技术研究院 | Fire identification method and system |
| CN102867386A (en) * | 2012-09-10 | 2013-01-09 | 南京恩博科技有限公司 | Intelligent video analysis-based forest smoke and fire detection method and special system thereof |
| CN103971114A (en) * | 2014-04-23 | 2014-08-06 | 天津航天中为数据***科技有限公司 | Forest fire detection method based on aerial remote sensing |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8913783B2 (en) * | 2009-10-29 | 2014-12-16 | Sri International | 3-D model based method for detecting and classifying vehicles in aerial imagery |
-
2017
- 2017-03-31 CN CN201710209365.9A patent/CN107085904B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011032117A1 (en) * | 2009-09-13 | 2011-03-17 | Delacom Detection Systems, Llc | Method and system for wildfire detection using a visible range camera |
| CN102842199A (en) * | 2012-08-30 | 2012-12-26 | 广州中国科学院工业技术研究院 | Fire identification method and system |
| CN102867386A (en) * | 2012-09-10 | 2013-01-09 | 南京恩博科技有限公司 | Intelligent video analysis-based forest smoke and fire detection method and special system thereof |
| CN103971114A (en) * | 2014-04-23 | 2014-08-06 | 天津航天中为数据***科技有限公司 | Forest fire detection method based on aerial remote sensing |
Non-Patent Citations (3)
| Title |
|---|
| Artificial Intelligence for Forest Fire Prediction;George E. Sakr et.al;《2010 IEEE/ASME International Conference on Advanced Intelligent Mechatronics》;20100731;第1311-1316页 |
| SVM方法在森林火险预测中的应用;黄玉霞等;《林业科学》;20071031;第43卷(第10期);第77-82页 |
| 基于支持向量机方法的森林火险预测研究;许志卿等;《中国农学通报》;20120531;第28卷(第13期);第126-131页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107085904A (en) | 2017-08-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107085904B (en) | Forest fire danger class determination method and system based on single classification SVM | |
| KR102159692B1 (en) | solar photovoltatic power generation forecasting apparatus and method based on big data analysis | |
| TWI662422B (en) | Air quality prediction method based on machine learning model | |
| CN109508360B (en) | Geographical multivariate stream data space-time autocorrelation analysis method based on cellular automaton | |
| CN114626512B (en) | High-temperature disaster forecasting method based on directed graph neural network | |
| CN107133960A (en) | Image crack dividing method based on depth convolutional neural networks | |
| CN102323972B (en) | Method for managing phased array radar resource | |
| CN102956023A (en) | Bayes classification-based method for fusing traditional meteorological data with perception data | |
| CN106650977A (en) | Short-term power prediction method used for newly-built wind farm | |
| CN109973331A (en) | A kind of fan blade of wind generating set fault diagnosis algorithm based on bp neural network | |
| CN109902340A (en) | It is a kind of meter and complicated meteorology coupled characteristic multi-source-lotus combine scene generating method | |
| CN109615147A (en) | A kind of following 72 hours air pollution forecasting method for early warning | |
| Fu et al. | An innovative decision making method for air quality monitoring based on big data-assisted artificial intelligence technique | |
| CN113987912A (en) | Pollutant on-line monitoring system based on geographic information | |
| CN106779222A (en) | Airport ground stand-by period Forecasting Methodology and device | |
| Ao et al. | Hybrid model of air quality prediction using k-means clustering and deep neural network | |
| CN106570592A (en) | Artificial-neural-network-based intelligent numerical value forecasting correction system | |
| Wenxiang et al. | Optimization of deep learning model for coastal chlorophyll a dynamic forecast | |
| CN113552855B (en) | Industrial equipment dynamic threshold setting method and device, electronic equipment and storage medium | |
| CN106295175B (en) | Station meteorological data missing value fill method based on svd algorithm | |
| CN112101608A (en) | Offshore wind farm site selection method and device | |
| Zhu et al. | Optimized air quality prediction model based on neural network | |
| CN115481788A (en) | Load prediction method and system for phase change energy storage system | |
| CN114552570A (en) | Offshore wind power prediction management system | |
| CN113627831A (en) | Method and device for determining house checking sequence, terminal equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |