CN108510191B - Mangrove ecological health evaluation method based on stacking noise reduction automatic coding algorithm - Google Patents
Mangrove ecological health evaluation method based on stacking noise reduction automatic coding algorithm Download PDFInfo
- Publication number
- CN108510191B CN108510191B CN201810281244.XA CN201810281244A CN108510191B CN 108510191 B CN108510191 B CN 108510191B CN 201810281244 A CN201810281244 A CN 201810281244A CN 108510191 B CN108510191 B CN 108510191B
- Authority
- CN
- China
- Prior art keywords
- layer
- mangrove forest
- data set
- information
- forest ecological
- 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
- 240000002044 Rhizophora apiculata Species 0.000 title claims abstract description 99
- 230000036541 health Effects 0.000 title claims abstract description 30
- 230000009467 reduction Effects 0.000 title claims abstract description 20
- 238000011156 evaluation Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims abstract description 15
- 230000007704 transition Effects 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 14
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 12
- 238000005516 engineering process Methods 0.000 claims description 12
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 9
- 229910052760 oxygen Inorganic materials 0.000 claims description 9
- 239000001301 oxygen Substances 0.000 claims description 9
- 241000607479 Yersinia pestis Species 0.000 claims description 8
- 150000001875 compounds Chemical class 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 239000002689 soil Substances 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 241000251468 Actinopterygii Species 0.000 claims description 4
- 241000254173 Coleoptera Species 0.000 claims description 4
- 241000195493 Cryptophyta Species 0.000 claims description 4
- 241000238557 Decapoda Species 0.000 claims description 4
- 241000255777 Lepidoptera Species 0.000 claims description 4
- 241001465754 Metazoa Species 0.000 claims description 4
- 244000005700 microbiome Species 0.000 claims description 4
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 claims description 3
- 101150099612 Esrrb gene Proteins 0.000 claims description 3
- 229910002651 NO3 Inorganic materials 0.000 claims description 3
- NHNBFGGVMKEFGY-UHFFFAOYSA-N Nitrate Chemical compound [O-][N+]([O-])=O NHNBFGGVMKEFGY-UHFFFAOYSA-N 0.000 claims description 3
- IOVCWXUNBOPUCH-UHFFFAOYSA-M Nitrite anion Chemical compound [O-]N=O IOVCWXUNBOPUCH-UHFFFAOYSA-M 0.000 claims description 3
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 3
- XKMRRTOUMJRJIA-UHFFFAOYSA-N ammonia nh3 Chemical compound N.N XKMRRTOUMJRJIA-UHFFFAOYSA-N 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 229910052799 carbon Inorganic materials 0.000 claims description 3
- 229930002875 chlorophyll Natural products 0.000 claims description 3
- 235000019804 chlorophyll Nutrition 0.000 claims description 3
- ATNHDLDRLWWWCB-AENOIHSZSA-M chlorophyll a Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 ATNHDLDRLWWWCB-AENOIHSZSA-M 0.000 claims description 3
- 239000003208 petroleum Substances 0.000 claims description 3
- 229910052698 phosphorus Inorganic materials 0.000 claims description 3
- 239000011574 phosphorus Substances 0.000 claims description 3
- 239000013049 sediment Substances 0.000 claims description 3
- 150000003568 thioethers Chemical class 0.000 claims description 3
- 230000008021 deposition Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 2
- 230000004083 survival effect Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 235000020068 maotai Nutrition 0.000 description 1
- 239000013535 sea water Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 229910021654 trace metal Inorganic materials 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A20/00—Water conservation; Efficient water supply; Efficient water use
- Y02A20/152—Water filtration
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a mangrove forest ecological health evaluation method based on a stacking noise reduction automatic coding algorithm, which mainly comprises the following steps: 1) and determining the mangrove forest ecological protection area. 2) And establishing a mangrove forest ecological index data set. 3) Data set X and data set Y are preprocessed. Taking the normalized data set X as the data set R1.4) The data in data set X is trained using a stacked noise reduction auto-encoding algorithm. 5) And establishing a mangrove forest ecological index prediction model. 6) And predicting the future ecological health condition of the mangrove forest in the mangrove forest ecological protection area by utilizing the mangrove forest ecological index prediction model, the linear training set S and the mangrove forest ecological health evaluation grade judgment table. The method utilizes the predicted important ecological indexes and combines the mangrove forest ecological health evaluation grade judgment table to finally realize accurate prediction of the mangrove forest ecological health condition.
Description
Technical Field
The invention relates to a data mining technology and a deep learning method, in particular to a mangrove forest ecological health evaluation method based on a stacking noise reduction automatic coding algorithm.
Background
The mangrove forest is a woody plant community which grows in the intertidal zone of the tropical and subtropical coast and is periodically immersed by seawater, is a complex ecosystem with land and sea characteristics and an important ecological key zone of the coast, and plays an irreplaceable role in improving gulfs, preventing waves and protecting dikes, purifying pollution, protecting wetland diversity and the like. The coast of Guangxi is an important mangrove forest distribution area in mainland of China, the land line length is 1490km, the area of the mangrove forest is the second nationwide, and the district with the largest mangrove forest distribution area per unit land line length is the province. The Guangxi is provided with 3 mangrove natural protection areas (the Shankou national natural protection area of Guangxi Hepu, the Beilun river national natural protection area of Guangxi City harbor and the Maotai province natural protection area of Guangxi Qinzhou). According to the current environmental status report of the ecological protection area of the river mouth in northern deluxe in the past eight years, mangrove forest ecological indexes including water environment, deposition environment, plankton, phytoplankton, intertidal zone organisms, mangrove forest communities and birds can be known. Wherein the indexes of plankton, phytoplankton, intertidal zone organisms, mangrove community and birds can most directly reflect the ecological health of mangrove. The analysis shows that the indexes are closely related, for example, the water environment has great influence on the survival of zooplankton, phytoplankton and intertidal zone organisms, and the deposition environment has great influence on the survival of zooplankton and phytoplankton. However, the sensing and image technology can only collect part of mangrove forest ecological indexes, such as water environment, deposition environment pests, pests and the like. Therefore, the sensing and image technology is difficult to accurately evaluate and predict the ecological health condition of the mangrove forest.
Disclosure of Invention
The present invention is directed to solving the problems of the prior art.
The technical scheme adopted for achieving the purpose of the invention is that the mangrove forest ecological health evaluation method based on the stacking noise reduction automatic coding algorithm mainly comprises the following steps:
1) and determining the mangrove forest ecological protection area.
2) And establishing a mangrove forest ecological index data set.
And taking mangrove forest ecological index data information I acquired by the mangrove forest ecological protection area in eta years by using a sensing and image technology as a data set X. The mangrove forest ecological index data information I mainly comprises water quality information, sediment pH, soil pH and soil granularity indexes.
The water quality information mainly comprises water temperature, salinity, pH, chlorophyll, ammonia nitrogen, nitrate, nitrite, inorganic phosphorus, petroleum and chemical oxygen demand.
The deposits comprise mainly organic carbon and sulphides.
Data set X is shown below:
in the formula (I), the compound is shown in the specification,is mangrove forest ecological index data information I.
And taking mangrove forest ecological index data information II acquired manually in the ecological protection area within eta years as a data set Y. The mangrove forest ecological index data information II mainly comprises fish information, shrimp information, microorganism information, algae information, Chemical Oxygen Demand (COD) information, Biochemical Oxygen Demand (BOD) information, mangrove forest pest information, community type, benthonic animal information, phytoplankton information, intertidal zone biological information, lepidoptera information and coleoptera information.
Data set Y is shown below:
in the formula (I), the compound is shown in the specification,is mangrove forest ecological index data information II.
3) Data set X and data set Y are preprocessed. The preprocessing mainly comprises denoising and normalization. CountingThe data set after X normalization is R1。
4) The data in data set X is trained using a stacked noise reduction auto-encoding algorithm.
The main steps for training the data in the data set X are as follows:
4.1) determining the network structure of the stacking noise reduction automatic coding algorithm. And setting the network to have M layers, wherein the number of input layers is 1, the number of transition layers is M-2, and the number of output layers is 1.
4.2) data set R1As an input layer to the network.
4.3) setting the initial weight of the transition layer. Training the transition layer by adopting a noise reduction automatic coding algorithm to obtain a training result H1。
4.4) training results H1As the input layer for the next layer. Repeating the step 3 to obtain a training result H2。
4.5) repeating the step 3 and the step 4M-2 times to finally obtain the output result of the M layer, namely the output layer. And recording the output result as a data set V, and finishing the training.
5) And establishing a mangrove forest ecological index prediction model.
The method mainly comprises the following steps of:
5.1) designing an inverse decision algorithm with the number of layers p. The reverse decision algorithm mainly comprises an input layer, a transition layer and an output layer. The number of the input layers is 1. The number of the transition layers is p-1. The number of the output layers is 1.
And taking the data set V and the data set Y as a training set D of the mangrove forest ecological index prediction model. Training set D is as follows:
D={(V1,Y1),(V2,Y2)……(Vm,Ym)}。 (3)
in the formula, ViIs row i of data set V. Y isiIs row i of data set Y.
5.2) inputting the data set V as a characteristic.
5.3) recording the weight between the ith and jth layerHeavy is Wij. Bias of j-th layer is thetaj。
Wherein, Wij∈[-1,0]。θj∈[-1,0]。
5.4) input I of the j-th layerjAs follows:
in the formula, WijIs the weight between the ith and jth layers. ThetajIs the bias of the j-th layer. O isiIs the output of the ith layer. i is the level number. j is the level number.
Output of j-th layer OjAs follows:
in the formula IjIs the input to the j-th layer. j is the level number.
Input I through the j-th layerjAnd output of j-th layer OjForward transmission to obtain the training result of the first layer as S1。
5.5) adding S1As an input layer for the next layer. Defining an error function Err1j. Error function Err1jAs follows:
in the formula, WjkIs the weight between the j-th and k-th layers. O isjIs the output of the j-th layer. j is the level number. And k is a layer sequence number.
Weight change amount Δ WijAs follows:
ΔWij=(l)Err1jOi。 (7)
in the formula, OiIs the output of the ith layer. i is the level number. Err1jIs an error function.
Offset change deltajAs follows:
Δθj=(l)Err1j。 (8)
in the formula, j is a hierarchical sequence number. (l) Is a learning coefficient. Err1jIs an error function.
Repeating the step 3 and the step 4 to obtain a training result S of the layer 22。
5.6) repeating the steps 3 to 5p-1 times to finally obtain the p layer, namely the output result Y of the output layer1。
5.7) for the output layer, an error function Err2 is definedj。
Err2j=Oj(I-Oj)(1-Oj)。 (9)
In the formula, OjIs the output of the j-th layer. j is the level number.
In the negative gradient direction, with a minimum Err2jAnd a minimum of thetajOn the basis of the values, the weight W is adjusted using the formula (10) and the formula (11)ijAnd an offset thetaj。
W′ij=Wij+ΔWij。 (10)
In the formula, WijIs the weight between the ith and jth layers. Δ WijIs the amount of weight change. W'ijTo adjust the weight between the ith layer and the jth layer.
θj=θj+Δθj。 (11)
In the formula, thetajIs the bias of the j-th layer. Delta thetajIs the amount of change in the bias. Theta'jTo adjust the bias of the j-th layer after the adjustment.
5.8) judging the weight W 'between the ith layer and the jth layer after adjustment'ijWhether or not it is below a threshold value epsilon1. Determining offset theta 'between the ith and jth layers after adjustment'jWhether or not it is below a threshold value epsilon2。
5.9) if W'ij≥ε1Then W'ijValue of as WijValue, repeating equation (10) to retrieve the weight W 'between the adjusted ith and jth layers'ijAnd step 5.8 is repeated.
If theta'j≥ε2Then theta 'will be'jValue of (2) is θ'jValue, repeat equation (11) to recover the offset θ 'between the ith and jth layers after adjustment'jAnd step 5.8 is repeated.
If W'ij<ε1And θ'j<ε2And if the result is true, finishing training to obtain the mangrove forest ecological index prediction model.
6) And predicting the future ecological health condition of the mangrove forest in the mangrove forest ecological protection area by utilizing the mangrove forest ecological index prediction model, the linear training set S and the mangrove forest ecological health evaluation grade judgment table.
The effects of the present invention are undoubted. The unsupervised deep feature learning is used for preprocessing the mangrove forest ecological big data, a deep neural network consisting of a plurality of layers of noise reduction automatic decoders is used for processing in an unsupervised mode, and the deep structure and the law in the mangrove forest ecological big data are obtained, so that the formed data representation is easier to understand by a deep learning algorithm of a prediction model, and the prediction accuracy of a reverse decision algorithm is obviously improved. Meanwhile, a reverse decision network model is established, and data information of plankton, phytoplankton, intertidal organisms and mangrove communities in the next time period is accurately predicted through water environment, deposition environment and pest data measured by current time period sensing and images. Thereby obtaining the correlation between the indexes which can be collected by the sensing and image technology and the important indexes which can not be collected. The invention utilizes the predicted important ecological indexes and combines the mangrove forest ecological health evaluation grade judgment table to finally realize the aim of predicting the mangrove forest ecological health condition.
Drawings
FIG. 1 is a general flow chart of the mangrove forest ecological health status prediction technology;
FIG. 2 is a block diagram of a reverse decision algorithm;
FIG. 3 is a comparison of predicted data and actual data;
FIG. 4 is a comparison graph of the result of ecological index prediction using the kernel technique algorithm.
Detailed Description
The present invention is further illustrated by the following examples, but it should not be construed that the scope of the above-described subject matter is limited to the following examples. Various substitutions and alterations can be made without departing from the technical idea of the invention and the scope of the invention is covered by the present invention according to the common technical knowledge and the conventional means in the field.
Example 1:
referring to fig. 1 to 4, a mangrove forest ecological health evaluation method based on a stacking noise reduction automatic coding algorithm mainly comprises the following steps:
1) and determining the mangrove forest ecological protection area. In this example, the ecological protection area of the northern river estuary is selected.
2) And establishing a mangrove forest ecological index data set.
And taking mangrove forest ecological index data information I acquired by the mangrove forest ecological protection area in the last 7 years by using a sensing and image technology as a data set X. The mangrove forest ecological index data information I mainly comprises water quality information, sediment pH, soil pH and soil granularity indexes.
The water quality information mainly comprises water temperature, salinity, pH, chlorophyll, ammonia nitrogen, nitrate, nitrite, inorganic phosphorus, petroleum and chemical oxygen demand.
The indexes which can most directly reflect the ecological health condition of the mangrove such as fish, shrimp, microorganism and algae, COD, BOD and other trace metal atoms, mangrove insect pests, community types, benthonic animals, phytoplankton and intertidal zone organisms, lepidoptera and coleoptera information are difficult to acquire through sensing and image technology, and must be manually acquired for analysis, thus being time-consuming and labor-consuming. Therefore, it is necessary to establish a model to obtain the correlation between the indexes which can be acquired by sensing and image technology and the important indexes which cannot be acquired.
The deposits comprise mainly organic carbon and sulphides.
In the formula (I), the compound is shown in the specification,is mangrove forest ecological index data information I. Data with the same attribute is specified as a column in the training set, and different samples are specified as a row in the training set.
And taking mangrove forest ecological index data information II manually collected in the ecological protection area in the last 7 years as a data set Y. The mangrove forest ecological index data information II mainly comprises fish information, shrimp information, microorganism information, algae information, Chemical Oxygen Demand (COD) information, Biochemical Oxygen Demand (BOD) information, mangrove forest pest information, community type, benthonic animal information, phytoplankton information, intertidal zone biological information, lepidoptera information and coleoptera information.
In the formula (I), the compound is shown in the specification,is mangrove forest ecological index data information II. Data with the same attribute is specified as a column in the training set, and different samples are specified as a row in the training set. Y isi∈RL。
3) Data set T and data set Y are preprocessed. The preprocessing mainly comprises denoising and normalization. The unit of the column information related to the data having a request for the unit is unified, and m represents the dimension after the unit is unified. Taking the normalized data set X as the data set R1。
Standardize min-max:one column of the data set X, max and min being the maximum and minimum values in the column, XiFor the normalized data column information, each column in the data set X is normalized respectively, so that the data of each column is mapped to [0,1]Within, the normalization is finished to obtain a data set R1。
4) The data in data set X is trained using a stacked noise reduction auto-encoding algorithm.
The main steps for training the data in the data set X are as follows:
4.1) determining the network structure of the stacking noise reduction automatic coding algorithm. And setting the network to have M layers, wherein the number of input layers is 1, the number of transition layers is M-2, and the number of output layers is 1.
4.2) data set R1As an input layer to the network.
4.3) setting the initial weight H of the transition layer. Training the transition layer by adopting a noise reduction automatic coding algorithm to obtain a training result H1。
H1=RW。 (4)
Output set H of transition layer1Reconstructing an and R by decoding1The same size specification signal F. Decoding the expression:
f and R1The reconstruction error between is as follows:
4.4) training results H1As the input layer for the next layer. Repeating the step 3 to obtain a training result H2。
4.5) repeating the step 3 and the step 4M-2 times to finally obtain the output result of the M layer, namely the output layer. And recording the output result as a data set V, and finishing the training.
5) And establishing a mangrove forest ecological index prediction model.
The method mainly comprises the following steps of:
5.1) designing an inverse decision algorithm with the number of layers p. The total number of layers of the reverse decision algorithm is the number of transition layers plus the number of output layers. The main idea of the reverse decision algorithm is to divide the learning process into two stages:
the first stage is the forward propagation process of the information flow. When input information propagates and is processed in the path of "input layer → transition layer → output layer", the actual output value of each layer is calculated.
The second stage is the back propagation process of the error. When the desired output value is not obtained at the output layer, the difference (i.e., error) between the actual output and the desired output propagates in the path "output layer → hidden layer → input layer". Specifically, the error is distributed to each layer, so as to obtain error signals of each layer, and the error signals are used as the basis for correcting each connection weight. The repeated application of these two processes ultimately minimizes the error.
The reverse decision algorithm mainly comprises an input layer, a transition layer and an output layer. The number of the input layers is 1. The number of the transition layers is p-1. The number of the output layers is 1.
And taking a linear training set S and a data set Y as a training set D of the mangrove forest ecological index prediction model. Training set D is as follows:
D={(V1,Y1),(V2,Y2)……(Vm,Ym)}。 (7)
in the formula, ViIs the ith row of the linear training set V. Y isiIs row i of data set Y.
5.2) inputting the data set V as a characteristic.
5.3) note the weight between the ith and jth layers as Wij. Bias of j-th layer is thetaj。
Wherein, Wij∈[-1,0]。θj∈[-1,0]。
5.4) input I of the j-th layerjAs follows:
in the formula, WijIs the weight between the ith and jth layers. ThetajIs the bias of the j-th layer. O isiIs the output of the ith layer. i is the level number. j is the level number.
Output of j-th layer OjAs follows:
in the formula IjIs the input to the j-th layer. j is the level number.
Input I through the j-th layerjAnd output of j-th layer OjForward transmission to obtain the training result of the first layer as S1。
5.5) adding S1As an input layer for the next layer. Defining an error function Err1j. Error function Err1jAs follows:
in the formula, WjkIs the weight between the j-th and k-th layers. O isjIs the output of the j-th layer. j is the level number. And k is a layer sequence number.
Weight change amount Δ WijAs follows:
ΔWij=(l)ErrjOi。 (11)
in the formula, OiIs the output of the ith layer. i is the level number.
Offset change deltajAs follows:
Δθj=(l)Errj。 (12)
in the formula, j is a hierarchical sequence number. (l) Is a learning coefficient. Err1jIs an error function.
Repeating the steps 3 and 4 to obtain the layer 2Training result S of2。
5.6) repeating the steps 3 to 5p-1 times to finally obtain the p layer, namely the output result Y of the output layer1。
5.7) for the output layer, an error function Err2 is definedj。
Err2j=Oj(I-Oj)(1-Oj)。 (13)
In the formula, OjIs the output of the j-th layer. j is the level number.
In the negative gradient direction, with a minimum Err2jAnd minimum ErriOn the basis of the values, the weight W is adjusted using the formula (14) and the formula (15)ijAnd an offset thetaj. Passing minimum Err2jCalculating to obtain a weight WijAnd then updating the weight.
W′ij=Wij+ΔWij。 (14)
In the formula, WijIs the weight between the ith and jth layers. Δ WijIs the amount of weight change. W'ijTo adjust the weight between the ith layer and the jth layer.
θ′j=θj+Δθj。 (15)
In the formula, thetajIs the bias of the j-th layer. Delta thetajIs the amount of change in the bias. Theta'jTo adjust the bias of the j-th layer after the adjustment.
5.8) judging the weight W 'between the ith layer and the jth layer after adjustment'ijWhether or not it is below a threshold value epsilon1. Determining offset theta 'between the ith and jth layers after adjustment'jWhether or not it is below a threshold value epsilon2。
5.9) if W'ij≥ε1Then W'ijValue of as WijValue, repeat equation (14) to recover the adjusted weight W 'between the ith and jth layers'ijAnd step 5.8 is repeated.
If theta'j≥ε2Then theta 'will be'jValue of (2) is θ'jValue, repeating equation (15) to recover the offset θ 'between the ith and jth layers after adjustment'jAnd repeating step 5.8。
If W'ij<ε1And θ'j<ε2And if the result is true, finishing training to obtain the mangrove forest ecological index prediction model.
Part of mangrove forest ecological index data information acquired by northern river estuary ecological protection area sensing and image technology in July 2016 is used as a test set T, the output of the model is compared with relevant parameters acquired manually in August, and the accuracy of prediction by adopting a reverse decision algorithm is higher as can be seen from the prediction result, so that the model is verified. As shown in fig. 3 and 4. Meanwhile, fig. 4 also illustrates that the prediction accuracy of the reverse decision algorithm is significantly improved by utilizing the kernel technique algorithm to perform linearization processing.
6) And predicting the future ecological health condition of the mangrove forest in the mangrove forest ecological protection area by utilizing the mangrove forest ecological index prediction model, the linear training set S and the mangrove forest ecological health evaluation grade judgment table. The mangrove forest ecological health evaluation grade judgment table grades the ecological health of the mangrove forest, for example, the health is 1 grade, the sub-health is 2 grade, the unhealthy is 3 grade, and the critical is 4 grade.
According to the invention, an accurate mangrove forest ecological index prediction model is constructed through a reverse decision algorithm, a kernel skill algorithm is added in front of the prediction model to process mangrove forest ecological big data, so that the finally constructed data representation is easier to understand by a deep learning algorithm of the ecological prediction model, and the prediction accuracy of the reverse decision prediction model is obviously improved.
Meanwhile, the mangrove forest ecological index prediction model can accurately predict the data information of plankton, phytoplankton, intertidal zone organisms and mangrove forest communities in the next time period through the water environment, deposition environment and pest data measured by sensing and image in the current time period, thereby obtaining the correlation between the indexes which can be acquired by sensing and image technology and the important indexes which can not be acquired. And finally, the predicted important ecological indexes are utilized, and the target of predicting the ecological health condition of the mangrove forest is realized by combining the mangrove forest ecological health evaluation grade judgment table.
Claims (4)
1. A mangrove forest ecological health evaluation method based on a stacking noise reduction automatic coding algorithm is characterized by mainly comprising the following steps:
1) determining an ecological protection area of the mangrove forest;
2) establishing a mangrove forest ecological index data set;
taking mangrove forest ecological index data information I acquired by the mangrove forest ecological protection area in eta years by using a sensing and image technology as a data set X;
mangrove forest ecological index data information II acquired manually in the ecological protection area within eta years is used as a data set Y;
3) preprocessing a data set X and a data set Y; the preprocessing mainly comprises denoising and normalization; recording the data set after X normalization as R1;
4) Training data in the data set X by using a stacking noise reduction automatic coding algorithm to obtain a data set V;
5) establishing a mangrove forest ecological index prediction model;
the method mainly comprises the following steps of:
5.1) designing a reverse decision algorithm with the number of layers p; the reverse decision algorithm mainly comprises an input layer, a transition layer and an output layer; the number of the input layers is 1; the number of the transition layers is p-1; the number of the output layers is 1;
taking a data set V and a data set Y as a training set D of the mangrove forest ecological index prediction model; training set D is as follows:
D={(V1,Y1),(V2,Y2)……(Vm,Ym)}; (1)
in the formula, ViRow i of the data set V; y isiRow i of data set Y;
5.2) inputting the data set V as a characteristic;
5.3) note the weight between the ith and jth layers as Wij(ii) a Bias of j-th layer is thetaj;
Wherein, Wij∈[-1,0];θj∈[-1,0];
5.4) input I of the j-th layerjAs follows:
in the formula, WijIs the weight between the ith layer and the jth layer; thetajIs the bias of the j-th layer; o isiIs the output of the ith layer; i is a level sequence number; j is a hierarchical sequence number;
output of j-th layer OjAs follows:
in the formula IjIs the input of the j-th layer; j is a hierarchical sequence number;
input I through the j-th layerjAnd output of j-th layer OjForward transmission to obtain the training result of the first layer as S1;
5.5) adding S1An input layer as a next layer; defining an error function Err1j(ii) a Error function Err1jAs follows:
in the formula, WjkIs the weight between the j-th layer and the k-th layer; o isjIs the output of the j-th layer; j is a hierarchical sequence number; k is a hierarchical sequence number;
weight change amount Δ WijAs follows:
ΔWij=(l)Err1jOi; (5)
in the formula, OiIs the output of the ith layer; i is a level sequence number; (l) Is a learning coefficient; err1jIs an error function;
offset change deltajAs follows:
Δθj=(l)Err1j; (6)
in the formula, j is a level serial number; (l) Is a learning coefficient; err1jIs an error function;
repeating the step 5.3) and the step 5.4) to obtain a training result S of the layer 22;
5.6) repeating the step 5.3) to the step 5.5) for p-1 times to finally obtain the output result Y of the p-th layer, namely the output layer1;
5.7) for the output layer, an error function Err2 is definedj;
Err2j=Oj(I-Oj)(1-Oj); (7)
In the formula, OjIs the output of the j-th layer; j is a hierarchical sequence number;
in the negative gradient direction, with a minimum Err2jAnd a minimum of thetajOn the basis of the values, the weight W is adjusted using the formula (8) and the formula (9)ijAnd an offset thetaj;
W′ij=Wij+ΔWij; (8)
In the formula, WijIs the weight between the ith layer and the jth layer; Δ WijIs the weight change amount; w'ijThe weight between the ith layer and the jth layer after adjustment;
θ′j=θj+Δθj; (9)
in the formula, thetajIs the bias of the j-th layer; delta thetajIs a bias change amount; theta'jIs the bias of the adjusted j-th layer;
5.8) judging the weight W 'between the ith layer and the jth layer after adjustment'ijWhether or not it is below a threshold value epsilon1(ii) a Determining offset theta 'between the ith and jth layers after adjustment'jWhether or not it is below a threshold value epsilon2;
5.9) if W'ij≥ε1Then W'ijValue of as WijValue, repeating the formula (8), and recovering the weight W 'between the ith layer and the jth layer after adjustment'ijAnd repeating step 5.8);
if theta'j≥ε2Then theta 'will be'jValue of as θjValue, repetitionEquation (9), retrieve offset θ 'between the i-th and j-th layers after adjustment'jAnd repeating step 5.8);
if W'ij<ε1And θ'j<ε2If the mangrove forest ecological index is established, training is finished to obtain a mangrove forest ecological index prediction model;
6) and predicting the future ecological health condition of the mangrove forest in the mangrove forest ecological protection area by utilizing the mangrove forest ecological index prediction model, the data set V and a mangrove forest ecological health evaluation grade judgment table.
2. The mangrove forest ecological health assessment method based on the stacking noise reduction automatic coding algorithm according to claim 1, characterized in that the data set X is as follows:
in the formula (I), the compound is shown in the specification,obtaining mangrove forest ecological index data information I;
data set X is shown below:
3. The mangrove forest ecological health assessment method based on the stacking noise reduction automatic coding algorithm according to claim 1, characterized in that: the mangrove forest ecological index data information I mainly comprises water quality information, sediment pH, soil pH and soil granularity indexes;
the water quality information mainly comprises water temperature, salinity, pH, chlorophyll, ammonia nitrogen, nitrate, nitrite, inorganic phosphorus, petroleum and chemical oxygen demand;
the deposit comprises primarily organic carbon and sulfides;
the mangrove forest ecological index data information II mainly comprises fish information, shrimp information, microorganism information, algae information, Chemical Oxygen Demand (COD) information, Biochemical Oxygen Demand (BOD) information, mangrove forest pest information, community type, benthonic animal information, phytoplankton information, intertidal zone biological information, lepidoptera information and coleoptera information.
4. The mangrove forest ecological health assessment method based on the stacking noise reduction automatic coding algorithm according to claim 1, characterized in that the main steps of training the data in the data set X are as follows:
1) determining a network structure of a stacking denoising automatic coding algorithm; setting the network to have M layers, wherein the number of input layers is 1, the number of transition layers is M-2, and the number of output layers is 1;
2) data set R1As an input layer of the network;
3) setting an initial weight of the transition layer; training the transition layer by adopting a noise reduction automatic coding algorithm to obtain a training result H1;
4) Will train the result H1An input layer as a next layer; repeating the step 3) to obtain a training result H2;
5) Repeating the step 3) and the step 4) for M-2 times to finally obtain an output result of the Mth layer, namely the output layer; and recording the output result as a data set V, and finishing the training.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810281244.XA CN108510191B (en) | 2018-04-02 | 2018-04-02 | Mangrove ecological health evaluation method based on stacking noise reduction automatic coding algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810281244.XA CN108510191B (en) | 2018-04-02 | 2018-04-02 | Mangrove ecological health evaluation method based on stacking noise reduction automatic coding algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108510191A CN108510191A (en) | 2018-09-07 |
CN108510191B true CN108510191B (en) | 2021-10-08 |
Family
ID=63379644
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810281244.XA Active CN108510191B (en) | 2018-04-02 | 2018-04-02 | Mangrove ecological health evaluation method based on stacking noise reduction automatic coding algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108510191B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110222832A (en) * | 2019-06-19 | 2019-09-10 | 中国水产科学研究院东海水产研究所 | Entrance of Changjiang River salt marshes macrobenthos habitat simulation prediction technique |
CN111413622B (en) * | 2020-04-03 | 2022-04-15 | 重庆大学 | Lithium battery life prediction method based on stacking noise reduction automatic coding machine |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
CN106777879A (en) * | 2016-11-16 | 2017-05-31 | 北京大学深圳研究生院 | Mangrove water environment health Evaluation method based on Biological indicators |
CN107101828A (en) * | 2017-05-27 | 2017-08-29 | 辽宁大学 | The bearing residual life Forecasting Methodology that a kind of encoder health degree is evaluated |
CN107730006A (en) * | 2017-09-13 | 2018-02-23 | 重庆电子工程职业学院 | A kind of nearly zero energy consumption controller of building based on regenerative resource big data deep learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10255628B2 (en) * | 2015-11-06 | 2019-04-09 | Adobe Inc. | Item recommendations via deep collaborative filtering |
-
2018
- 2018-04-02 CN CN201810281244.XA patent/CN108510191B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
CN106777879A (en) * | 2016-11-16 | 2017-05-31 | 北京大学深圳研究生院 | Mangrove water environment health Evaluation method based on Biological indicators |
CN107101828A (en) * | 2017-05-27 | 2017-08-29 | 辽宁大学 | The bearing residual life Forecasting Methodology that a kind of encoder health degree is evaluated |
CN107730006A (en) * | 2017-09-13 | 2018-02-23 | 重庆电子工程职业学院 | A kind of nearly zero energy consumption controller of building based on regenerative resource big data deep learning |
Non-Patent Citations (3)
Title |
---|
Deep belief network-based drug identication;Huihua Yang etal;《Journal of Innovative Optical Health Sciences》;20160628;第1-10页 * |
基于SDA与SVR混合模型的迁移学习预测算法;任俊 等;《计算机科学》;20180131;第280-284、312页 * |
基于时空优化深度神经网络的AQI等级预测;董婷 等;《计算机工程与应用》;20171231;第17-23、41页 * |
Also Published As
Publication number | Publication date |
---|---|
CN108510191A (en) | 2018-09-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jennings et al. | Trawling disturbance can modify benthic production processes | |
Gjelland et al. | Environmental influence on transmitter detection probability in biotelemetry: developing a general model of acoustic transmission | |
Brown et al. | Habitat suitability index models for eight fish and invertebrate species in Casco and Sheepscot Bays, Maine | |
Austin et al. | Intraspecific variation in movement patterns: modeling individual behaviour in a large marine predator | |
Heddam | Simultaneous modelling and forecasting of hourly dissolved oxygen concentration (DO) using radial basis function neural network (RBFNN) based approach: a case study from the Klamath River, Oregon, USA | |
Vaughan et al. | Combining surveys of river habitats and river birds to appraise riverine hydromorphology | |
Mishra et al. | Coastal water quality monitoring and modelling off Chennai city | |
CN108510191B (en) | Mangrove ecological health evaluation method based on stacking noise reduction automatic coding algorithm | |
Marchant | The use of taxonomic distinctness to assess environmental disturbance of insect communities from running water | |
Callaway | Historical data reveal 30-year persistence of benthic fauna associations in heavily modified waterbody | |
CN108520348B (en) | Ecological index prediction method based on mangrove forest ecological big data | |
CN114242156A (en) | Real-time prediction method and system for relative abundance of pathogenic vibrios on marine micro-plastic | |
Kennish | Estuarine research, monitoring, and resource protection | |
White et al. | More than a whistle: Automated detection of marine sound sources with a convolutional neural network | |
Fietz et al. | Sedimentology and neoichnology of a muddy translating point bar in the fluvio-tidal transition, Serpentine River, BC, Canada | |
Kodama et al. | Long‐term changes in the assemblage of demersal fishes and invertebrates in relation to environmental variations in Tokyo Bay, Japan | |
Knecht et al. | The impact of zooplankton calcifiers on the marine carbon cycle | |
Boswell et al. | Evaluation of target strength–fish length equation choices for estimating estuarine fish biomass | |
CN106845870B (en) | Fuzzy-level power plant warm-water drainage fishery damage assessment method based on grey correlation | |
Courtney et al. | Improving mortality rate estimates for management of the Queensland saucer scallop fishery | |
Guerra-Castro et al. | Spatial variability and human disturbance of sponge assemblages associated with mangrove roots in the southern Caribbean | |
RS et al. | Influence of salinity on the meiofaunal distribution in a hypersaline lake along the southeast coast of India | |
CN113221651B (en) | Submarine sediment classification method using acoustic propagation data and unsupervised machine learning | |
Tsikopoulou et al. | Can intensive fish farming for 20 years induce changes in benthic ecosystems on a scale of waterbody? An assessment from Cephalonia bay (Ionian Sea) | |
Stephenson et al. | Developing updated predictive models for benthic taxa and communities across Chatham Rise and Campbell Plateau using photographic survey data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |