CN117408167A - Debris flow disaster vulnerability prediction method based on deep neural network - Google Patents

Debris flow disaster vulnerability prediction method based on deep neural network Download PDF

Info

Publication number
CN117408167A
CN117408167A CN202311722664.4A CN202311722664A CN117408167A CN 117408167 A CN117408167 A CN 117408167A CN 202311722664 A CN202311722664 A CN 202311722664A CN 117408167 A CN117408167 A CN 117408167A
Authority
CN
China
Prior art keywords
neural network
debris flow
data
representing
training set
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311722664.4A
Other languages
Chinese (zh)
Inventor
赵建壮
杨磊
赵超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sichuan Energy Geological Survey And Research Institute
Original Assignee
Sichuan Energy Geological Survey And Research Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sichuan Energy Geological Survey And Research Institute filed Critical Sichuan Energy Geological Survey And Research Institute
Priority to CN202311722664.4A priority Critical patent/CN117408167A/en
Publication of CN117408167A publication Critical patent/CN117408167A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
    • 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
    • G06N3/048—Activation functions
    • 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
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Medical Informatics (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a debris flow disaster vulnerability prediction method based on a deep neural network, which comprises the following steps: acquiring mud-rock flow cataloging data of a preset area, and acquiring a training set and a testing set according to the mud-rock flow cataloging data; based on the training set, integrating a convolutional neural network and a cyclic neural network which are built in advance by a Stacking integration method, and building a deep neural network prediction model; the method comprises the steps of inputting a test set into a deep neural network prediction model to obtain a predicted result of the debris flow disaster vulnerability in a preset area, combining a convolutional neural network and a cyclic neural network through a Stacking integration method to obtain a more accurate deep neural network prediction model, wherein the deep neural network prediction model has better generalization capability and higher prediction precision, and the Stacking integration method can obtain a more reliable predicted result of the debris flow disaster vulnerability when modeling is performed by using extremely limited samples, and shows better performance in time consumption and accuracy.

Description

Debris flow disaster vulnerability prediction method based on deep neural network
Technical Field
The invention relates to the field of artificial intelligence testing, in particular to a debris flow disaster vulnerability prediction method based on a deep neural network.
Background
The mud-rock flow disaster seriously threatens the sustainable development of national economy and society, so that the mud-rock flow disaster prediction has important significance in mud-rock flow disaster prevention and reduction practice, and can be directly used for national economy construction, but the mud-rock flow disaster is numerous in influencing factors and complex in relation, and along with the development of economy, the economic loss caused by mud-rock flow increases along with the trend.
The traditional debris flow disaster vulnerability prediction method has the problems that the variables are difficult to accurately count data, so that the prediction is inaccurate, the existing debris flow vulnerability prediction method through displacement-time sequence modeling has randomness, the meanings of all components are required to be manually determined for splitting prediction, and the workload is increased.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a debris flow disaster vulnerability prediction method based on a deep neural network, which comprises the following steps:
acquiring mud-rock flow cataloging data of a preset area, and acquiring a training set and a testing set according to the mud-rock flow cataloging data;
based on the training set, integrating a convolutional neural network and a cyclic neural network which are built in advance by a Stacking integration method, and building a deep neural network prediction model;
and inputting the test set into the deep neural network prediction model to obtain a predicted result of the susceptibility of the debris flow disaster in the preset area.
Preferably, the acquiring the mud-rock flow catalogue data of the preset area, and acquiring the training set and the testing set according to the mud-rock flow catalogue data, includes:
acquiring mud-rock flow catalogue data of a preset area, and determining mud-rock flow evaluation factors through colinear screening according to the mud-rock flow catalogue data;
resampling the debris flow evaluation factor layer to the same spatial resolution to generate a same-frequency layer;
classifying the same-frequency layers according to the mud-rock flow historical experience value to obtain layer subcategories;
assigning values to the layer sub-categories, and determining the specific gravity of the debris flow evaluation factors;
constructing a debris flow susceptibility database according to the specific gravity of the debris flow evaluation factors, the debris flow evaluation factors and the debris flow catalogue data;
based on the debris flow vulnerability database, sample data of the debris flow and the non-debris flow are obtained, and the sample data are divided to obtain a training set and a testing set.
Preferably, the integrating the convolutional neural network and the cyclic neural network which are built in advance by a Stacking integration method based on the training set, and building a deep neural network prediction model comprises the following steps:
dividing the training set into sets with the same preset number and size through a base classifier based on the training set;
wherein the set comprises: a first initial test set, a second initial training set, …, and an nth initial training set;
respectively inputting the set into a convolutional neural network and a cyclic neural network which are constructed in advance, and training a primary classifier by using the second initial training set, the … and the nth initial training set;
testing the first initial test set through the primary classifier to obtain a convolution first test result in a convolution neural network and a circulation first test result in a circulation neural network;
setting the first convolution test result as a first subset in a training set of a meta-learner in the convolution neural network, respectively taking the second initial training set, the … initial training set and the nth initial training set as initial testing sets, and carrying out primary classifier training for n-1 times by using all sets outside the current initial testing set to obtain a second convolution test result, … and an nth convolution test result;
setting the convolved second test results, … and convolved nth test results to a second subset, … and nth subset of the training set of the meta learner;
determining a first target training set according to the first subset, the second subset, … and the nth subset, and selecting first prediction feature data according to the first target training set;
setting the first test result of the circulation as a first subset in a training set of a meta learner in the circulation neural network, respectively taking the second initial training set, the … initial training set and the nth initial training set as initial testing sets, and carrying out primary classifier training for n-1 times by using all sets except the current initial testing set to obtain a second test result of the circulation, … and a nth test result of the circulation;
setting the looped second test result, … and looped nth test result to a second subset, … and nth subset of the training set of meta learner;
determining a second target training set according to the first subset, the second subset, … and the nth subset, and selecting second prediction feature data according to the second target training set;
integrating the first prediction feature data and the second prediction feature data by a Stacking integration method to obtain debris flow prediction data;
and constructing a deep neural network prediction model according to the debris flow prediction data.
Preferably, the training steps of the pre-constructed convolutional neural network are as follows:
s1: inputting pre-acquired debris flow training sample data into an input layer of a convolutional neural network, and carrying out multidimensional arrangement on the debris flow training sample data through the input layer to obtain a preset dimension array;
s2: inputting the preset dimension array into a convolution kernel of a convolution layer through a convolution window, and performing inner product operation according to the sequence from left to right and from top to bottom to obtain a convolution characteristic diagram;
s3: carrying out pooling operation on the convolution feature map through a pooling layer to obtain pooled data;
s4: according to the pooled data, extracting features through a full-connection layer to obtain high-dimensional features, and mapping the high-dimensional features to a low-dimensional feature space to obtain preset dimensional feature data;
s5: and when the loss function in the convolutional neural network is smaller than or equal to a preset convergence value, outputting the preset dimension characteristic data through an output layer to obtain debris flow output data, otherwise, returning to the step S2 to continue training.
Preferably, the inner product operation has the following formula:
，
in the method, in the process of the invention,an output value representing a j-th convolution kernel;representing a nonlinear activation function;representing the weight of the mud-rock flow training sample data;representing input data corresponding to the convolution window;representing the bias of pooled data corresponding to the jth convolution kernel;representing the total dimension;representing the number of dimensions;representing the number of convolution kernels;representing the total number of convolution kernels.
Preferably, the training steps of the pre-constructed recurrent neural network are as follows:
r1: sequencing pre-acquired mud-rock flow training sample data according to a time sequence to acquire sample sequence data;
r2: carrying out importance assignment on the debris flow evaluation factors in the sample sequence data by a factor selection method to obtain input sequence data;
r3: inputting the input sequence data into a cyclic neural network, and activating the input sequence data through a cyclic layer of the cyclic neural network to obtain output data;
r4: and when the loss function in the cyclic neural network is smaller than or equal to a preset convergence value, outputting the output data through an output layer, otherwise, returning to the step R2 to continue training.
Preferably, the cyclic neural network calculation model is as follows:
；
wherein,；
in the method, in the process of the invention,time stepCorresponding output data;representing an activation function;a second weight matrix representing input sequence data;representing input sequence data;a bias representing input sequence data;a first weight matrix representing input sequence data;representing a time stepThe corresponding input sequence data;representing debris flow evaluation factors;representing an importance parameter;representing a sample sequence data bias vector;representing a time step.
Preferably, the inputting the test set into the deep neural network prediction model to obtain a predicted result of the susceptibility to the debris flow disaster in the preset area includes:
transmitting the state parameters of the test set and the deep neural network prediction model to a memory module for calculation to obtain a calculation result of a one-dimensional column vector;
based on the calculation result of the one-dimensional column vector, reserving output information to be reserved in the test set, and outputting state parameters of a memory module at the current moment and state parameters of an hidden layer in the deep neural network prediction model at the current moment;
and calculating the calculation result of the one-dimensional column vector through an activation formula of a linear function to obtain a predicted result of the susceptibility of the debris flow disaster in the preset area.
Preferably, the activation formula of the linear function is as follows:
；
in the method, in the process of the invention,representing a predicted result of the susceptibility of the debris flow disaster in a preset area;representing an activation function;a weight value representing data in the test set;representing the calculation result of the one-dimensional column vector;representing the prediction constants.
Preferably, after the test set is input into the deep neural network prediction model to obtain a predicted result of the susceptibility to the debris flow disaster in the preset area, the method further includes:
according to the predicted result of the debris flow disaster vulnerability in the preset area, performing a simulation experiment to obtain an experimental result;
and carrying out the accuracy evaluation of the debris flow disaster susceptibility prediction according to the experimental result.
Compared with the closest prior art, the invention has the following beneficial effects:
a debris flow disaster vulnerability prediction method based on a deep neural network comprises the following steps: acquiring mud-rock flow cataloging data of a preset area, and acquiring a training set and a testing set according to the mud-rock flow cataloging data; based on the training set, integrating a convolutional neural network and a cyclic neural network which are built in advance by a Stacking integration method, and building a deep neural network prediction model; the method comprises the steps of inputting a test set into a deep neural network prediction model to obtain a predicted result of the debris flow disaster vulnerability in a preset area, combining a convolutional neural network and a cyclic neural network through a Stacking integration method to obtain a more accurate deep neural network prediction model, wherein the deep neural network prediction model has better generalization capability and higher prediction precision, and the Stacking integration method can obtain a more reliable predicted result of the debris flow disaster vulnerability when modeling is performed by using extremely limited samples, and shows better performance in time consumption and accuracy.
Drawings
FIG. 1 is a schematic flow chart of a debris flow disaster vulnerability prediction method based on a deep neural network;
fig. 2 is a training flow chart of a convolutional neural network in the debris flow disaster vulnerability prediction method based on the deep neural network provided by the invention;
fig. 3 is a training flow chart of a cyclic neural network in the debris flow disaster vulnerability prediction method based on the deep neural network.
Detailed Description
The following describes the embodiments of the present invention in further detail with reference to the drawings.
Example 1
The flow chart of the debris flow disaster vulnerability prediction method based on the deep neural network provided by the invention is shown in fig. 1, and comprises the following steps:
step 1: acquiring mud-rock flow cataloging data of a preset area, and acquiring a training set and a testing set according to the mud-rock flow cataloging data;
step 2: based on the training set, integrating a convolutional neural network and a cyclic neural network which are built in advance by a Stacking integration method, and building a deep neural network prediction model;
step 3: and inputting the test set into the deep neural network prediction model to obtain a predicted result of the susceptibility of the debris flow disaster in the preset area.
Specifically, the step 1 includes:
acquiring mud-rock flow catalogue data of a preset area, and determining mud-rock flow evaluation factors through colinear screening according to the mud-rock flow catalogue data;
resampling the debris flow evaluation factor layer to the same spatial resolution to generate a same-frequency layer;
classifying the same-frequency layers according to the mud-rock flow historical experience value to obtain layer subcategories;
assigning values to the layer sub-categories, and determining the specific gravity of the debris flow evaluation factors;
constructing a debris flow susceptibility database according to the specific gravity of the debris flow evaluation factors, the debris flow evaluation factors and the debris flow catalogue data;
based on the debris flow vulnerability database, sample data of the debris flow and the non-debris flow are obtained, and the sample data are divided to obtain a training set and a testing set.
The selection of the debris flow evaluation factors is important for debris flow susceptibility evaluation;
according to the current field investigation and the existing related research, and comprehensively considering the factors such as topography, basic geology, hydrologic conditions and the like of a research area, finally selecting 20 debris flow influencing factors through debris flow catalogue data of a preset area;
wherein the 20 debris flow influencing factors at least comprise one or more of the following: digital elevation model data, slope direction, grade, drainage basin area, drainage basin grade, curvature, slope morphology factor, geographic location index, terrain roughness index, terrain surface curvature, terrain surface texture, terrain humidity index factor, lithology factor, fault distance factor, rainfall factor, magnitude factor, water system distance factor, land utilization factor, vegetation normalization index, and normalized water index;
to ensure spatial consistency of the factors, the factor layers are resampled to the same spatial resolution;
on the basis, the continuous layers are reclassified according to historical documents and expert experience, and then all the layer subcategories are assigned, so that the accuracy of data can be ensured, and the time for data analysis can be saved.
Specifically, the step 2 includes:
dividing the training set into sets with the same preset number and size through a base classifier based on the training set;
wherein the set comprises: a first initial test set, a second initial training set, …, and an nth initial training set;
respectively inputting the set into a convolutional neural network and a cyclic neural network which are constructed in advance, and training a primary classifier by using the second initial training set, the … and the nth initial training set;
testing the first initial test set through the primary classifier to obtain a convolution first test result in a convolution neural network and a circulation first test result in a circulation neural network;
setting the first convolution test result as a first subset in a training set of a meta-learner in the convolution neural network, respectively taking the second initial training set, the … initial training set and the nth initial training set as initial testing sets, and carrying out primary classifier training for n-1 times by using all sets outside the current initial testing set to obtain a second convolution test result, … and an nth convolution test result;
setting the convolved second test results, … and convolved nth test results to a second subset, … and nth subset of the training set of the meta learner;
determining a first target training set according to the first subset, the second subset, … and the nth subset, and selecting first prediction feature data according to the first target training set;
setting the first test result of the circulation as a first subset in a training set of a meta learner in the circulation neural network, respectively taking the second initial training set, the … initial training set and the nth initial training set as initial testing sets, and carrying out primary classifier training for n-1 times by using all sets except the current initial testing set to obtain a second test result of the circulation, … and a nth test result of the circulation;
setting the looped second test result, … and looped nth test result to a second subset, … and nth subset of the training set of meta learner;
determining a second target training set according to the first subset, the second subset, … and the nth subset, and selecting second prediction feature data according to the second target training set;
integrating the first prediction feature data and the second prediction feature data by a Stacking integration method to obtain debris flow prediction data;
and constructing a deep neural network prediction model according to the debris flow prediction data.
As shown in fig. 2, the training steps of the convolutional neural network constructed in advance are as follows:
s1: inputting pre-acquired debris flow training sample data into an input layer of a convolutional neural network, and carrying out multidimensional arrangement on the debris flow training sample data through the input layer to obtain a preset dimension array;
s2: inputting the preset dimension array into a convolution kernel of a convolution layer through a convolution window, and performing inner product operation according to the sequence from left to right and from top to bottom to obtain a convolution characteristic diagram;
s3: carrying out pooling operation on the convolution feature map through a pooling layer to obtain pooled data;
s4: according to the pooled data, extracting features through a full-connection layer to obtain high-dimensional features, and mapping the high-dimensional features to a low-dimensional feature space to obtain preset dimensional feature data;
s5: and when the loss function in the convolutional neural network is smaller than or equal to a preset convergence value, outputting the preset dimension characteristic data through an output layer to obtain debris flow output data, otherwise, returning to the step S2 to continue training.
The inner product operation has the following calculation formula:
；
in the method, in the process of the invention,an output value representing a j-th convolution kernel;representing a nonlinear activation function;representing the weight of the mud-rock flow training sample data;representing input data corresponding to the convolution window;representing the bias of pooled data corresponding to the jth convolution kernel;representing the total dimension;representing the number of dimensions;representing the number of convolution kernels;representing the total number of convolution kernels.
The Convolutional Neural Network (CNN) is provided with an input layer and an output layer which are shared with a common neural network, and the Convolutional Neural Network (CNN) structurally comprises a convolutional layer, a pooling layer and a full-connection layer;
the pooling operation in the pooling layer not only can improve the calculation efficiency by reducing the dimension of the convolution feature map output by the convolution layer, but also can keep the extracted features in translation invariance;
the most common pooling operation is maximum pooling, and the calculation formula is as follows:
；
in the method, in the process of the invention,representing pooled data;indicating that the sub-pooling operation is in placeUpper corresponding toA convolution feature map of the dimension array;representing the pooled positions.
Wherein the fully connected layer is also one of Convolutional Neural Network (CNN) hidden layers, and the neural units of the fully connected layer are connected with the neurons of the upper layer;
the full connection layer can be regarded as a special classifier, and the purpose of the full connection layer is to map high-dimensional features extracted by convolution and pooling operation into a low-dimensional feature space;
the steps optimize the calculation steps and ensure the accuracy of output data.
As shown in fig. 3, the training steps of the pre-constructed recurrent neural network are as follows:
r1: sequencing pre-acquired mud-rock flow training sample data according to a time sequence to acquire sample sequence data;
r2: carrying out importance assignment on the debris flow evaluation factors in the sample sequence data by a factor selection method to obtain input sequence data;
r3: inputting the input sequence data into a cyclic neural network, and activating the input sequence data through a cyclic layer of the cyclic neural network to obtain output data;
r4: and when the loss function in the cyclic neural network is smaller than or equal to a preset convergence value, outputting the output data through an output layer, otherwise, returning to the step R2 to continue training.
The cyclic neural network calculation model is as follows:
；
wherein,；
in the method, in the process of the invention,time stepCorresponding output data;representing an activation function;a second weight matrix representing input sequence data;representing input sequence data;a bias representing input sequence data;a first weight matrix representing input sequence data;representing a time stepThe corresponding input sequence data;representing debris flow evaluation factors;representing an importance parameter;representing sample orderColumn data bias vector;representing a time step.
Wherein Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) models are iteratively trained based on a gradient descent algorithm until the loss function converges, and as the number of trains increases, the loss value decreases until a low level is reached, indicating that the training process is satisfactory, at which point the training may be ended.
Specifically, the step 3 includes:
transmitting the state parameters of the test set and the deep neural network prediction model to a memory module for calculation to obtain a calculation result of a one-dimensional column vector;
based on the calculation result of the one-dimensional column vector, reserving output information to be reserved in the test set, and outputting state parameters of a memory module at the current moment and state parameters of an hidden layer in the deep neural network prediction model at the current moment;
and calculating the calculation result of the one-dimensional column vector through an activation formula of a linear function to obtain a predicted result of the susceptibility of the debris flow disaster in the preset area.
The activation formula of the linear function is as follows:
；
in the method, in the process of the invention,representing a predicted result of the susceptibility of the debris flow disaster in a preset area;representing an activation function;a weight value representing data in the test set;representing the calculation result of the one-dimensional column vector;representing the prediction constants.
Specifically, after the step 3, the method further includes:
according to the predicted result of the debris flow disaster vulnerability in the preset area, performing a simulation experiment to obtain an experimental result;
and carrying out the accuracy evaluation of the debris flow disaster susceptibility prediction according to the experimental result.
Wherein, 10 pieces of simulation experiment data are screened out from sample data of the debris flow and the non-debris flow obtained from the debris flow susceptibility database;
inputting the 10 simulation experiment data into a deep neural network prediction model for simulation experiment to obtain an experiment result;
the obtained predicted result and experimental result of the susceptibility to the debris flow disaster in the preset area are subjected to difference making to obtain an error value=0.0059；
Wherein,；
in the method, in the process of the invention,representing an error value;representing a predicted result of the susceptibility of the debris flow disaster in a preset area;representing a simulation result obtained by a simulation experiment;
wherein,=0.5779；=0.5720；=0.5881，the method comprises the steps of representing initial data of the susceptibility of debris flow disasters in a preset area;
because ofAnd (2) andthe accuracy of predicting the susceptibility of the debris flow disaster can be evaluated to be excellent;
therefore, the method can be used for integrating the convolutional neural network and the cyclic neural network which are built in advance by a Stacking integration method, and the built deep neural network prediction model is more accurate in predicting the debris flow easy disasters;
the performance of the deep neural network prediction model in the aspects of time consumption and accuracy in operation is improved, the workload in actual operation is effectively reduced, and the reliability of a prediction result is improved.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that the foregoing embodiments are merely for illustrating the technical solution of the present invention and not for limiting the scope of protection thereof, and although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that various changes, modifications or equivalents may be made to the specific embodiments of the application after reading the present invention, and these changes, modifications or equivalents are within the scope of protection of the claims appended hereto.

Claims (10)

1. The debris flow disaster vulnerability prediction method based on the deep neural network is characterized by comprising the following steps of:
acquiring mud-rock flow cataloging data of a preset area, and acquiring a training set and a testing set according to the mud-rock flow cataloging data;
based on the training set, integrating a convolutional neural network and a cyclic neural network which are built in advance by a Stacking integration method, and building a deep neural network prediction model;
and inputting the test set into the deep neural network prediction model to obtain a predicted result of the susceptibility of the debris flow disaster in the preset area.
2. The method for predicting the susceptibility to the debris flow disaster based on the deep neural network according to claim 1, wherein the step of obtaining the debris flow catalogue data of the preset area, and obtaining the training set and the testing set according to the debris flow catalogue data, comprises the steps of:
acquiring mud-rock flow catalogue data of a preset area, and determining mud-rock flow evaluation factors through colinear screening according to the mud-rock flow catalogue data;
resampling the debris flow evaluation factor layer to the same spatial resolution to generate a same-frequency layer;
classifying the same-frequency layers according to the mud-rock flow historical experience value to obtain layer subcategories;
assigning values to the layer sub-categories, and determining the specific gravity of the debris flow evaluation factors;
constructing a debris flow susceptibility database according to the specific gravity of the debris flow evaluation factors, the debris flow evaluation factors and the debris flow catalogue data;
based on the debris flow vulnerability database, sample data of the debris flow and the non-debris flow are obtained, and the sample data are divided to obtain a training set and a testing set.
3. The debris flow disaster vulnerability prediction method based on the deep neural network according to claim 1, wherein the integrating the convolutional neural network and the cyclic neural network which are constructed in advance by a Stacking integration method based on the training set, and constructing the deep neural network prediction model comprises:
dividing the training set into sets with the same preset number and size through a base classifier based on the training set;
wherein the set comprises: a first initial test set, a second initial training set, …, and an nth initial training set;
respectively inputting the set into a convolutional neural network and a cyclic neural network which are constructed in advance, and training a primary classifier by using the second initial training set, the … and the nth initial training set;
testing the first initial test set through the primary classifier to obtain a convolution first test result in a convolution neural network and a circulation first test result in a circulation neural network;
setting the first convolution test result as a first subset in a training set of a meta-learner in the convolution neural network, respectively taking the second initial training set, the … initial training set and the nth initial training set as initial testing sets, and carrying out primary classifier training for n-1 times by using all sets outside the current initial testing set to obtain a second convolution test result, … and an nth convolution test result;
setting the convolved second test results, … and convolved nth test results to a second subset, … and nth subset of the training set of the meta learner;
determining a first target training set according to the first subset, the second subset, … and the nth subset, and selecting first prediction feature data according to the first target training set;
setting the first test result of the circulation as a first subset in a training set of a meta learner in the circulation neural network, respectively taking the second initial training set, the … initial training set and the nth initial training set as initial testing sets, and carrying out primary classifier training for n-1 times by using all sets except the current initial testing set to obtain a second test result of the circulation, … and a nth test result of the circulation;
setting the looped second test result, … and looped nth test result to a second subset, … and nth subset of the training set of meta learner;
determining a second target training set according to the first subset, the second subset, … and the nth subset, and selecting second prediction feature data according to the second target training set;
integrating the first prediction feature data and the second prediction feature data by a Stacking integration method to obtain debris flow prediction data;
and constructing a deep neural network prediction model according to the debris flow prediction data.
4. The debris flow disaster vulnerability prediction method based on deep neural network as claimed in claim 1, wherein the training step of the pre-constructed convolutional neural network is as follows:
s1: inputting pre-acquired debris flow training sample data into an input layer of a convolutional neural network, and carrying out multidimensional arrangement on the debris flow training sample data through the input layer to obtain a preset dimension array;
s2: inputting the preset dimension array into a convolution kernel of a convolution layer through a convolution window, and performing inner product operation according to the sequence from left to right and from top to bottom to obtain a convolution characteristic diagram;
s3: carrying out pooling operation on the convolution feature map through a pooling layer to obtain pooled data;
s4: according to the pooled data, extracting features through a full-connection layer to obtain high-dimensional features, and mapping the high-dimensional features to a low-dimensional feature space to obtain preset dimensional feature data;
s5: and when the loss function in the convolutional neural network is smaller than or equal to a preset convergence value, outputting the preset dimension characteristic data through an output layer to obtain debris flow output data, otherwise, returning to the step S2 to continue training.
5. The method for predicting the susceptibility to debris flow disasters based on a deep neural network according to claim 4, wherein the inner product calculation formula is as follows:
；
in the method, in the process of the invention,an output value representing a j-th convolution kernel; />Representing a nonlinear activation function; />Representing the weight of the mud-rock flow training sample data; />Representing input data corresponding to the convolution window; />Representing the bias of pooled data corresponding to the jth convolution kernel; />Representing the total dimension; />Representing the number of dimensions; />Representing the number of convolution kernels; />Representing the total number of convolution kernels.
6. The debris flow disaster vulnerability prediction method based on deep neural network as claimed in claim 1, wherein the training step of the pre-constructed recurrent neural network is as follows:
r1: sequencing pre-acquired mud-rock flow training sample data according to a time sequence to acquire sample sequence data;
r2: carrying out importance assignment on the debris flow evaluation factors in the sample sequence data by a factor selection method to obtain input sequence data;
r3: inputting the input sequence data into a cyclic neural network, and activating the input sequence data through a cyclic layer of the cyclic neural network to obtain output data;
r4: and when the loss function in the cyclic neural network is smaller than or equal to a preset convergence value, outputting the output data through an output layer, otherwise, returning to the step R2 to continue training.
7. The debris flow disaster vulnerability prediction method based on the deep neural network according to claim 6, wherein the cyclic neural network calculation model is as follows:
；
wherein,；
in the method, in the process of the invention,time step->Corresponding output data; />Representing an activation function; />A second weight matrix representing input sequence data; />Representing input sequence data; />A bias representing input sequence data; />First representing input sequence dataA weight matrix; />Representing a time step +.>The corresponding input sequence data; />Representing debris flow evaluation factors; />Representing an importance parameter; />Representing a sample sequence data bias vector; />Representing a time step.
8. The method for predicting the susceptibility to debris flow disasters based on the deep neural network according to claim 1, wherein the step of inputting the test set into the deep neural network prediction model to obtain the predicted result of the susceptibility to debris flow disasters in the preset area comprises the steps of:
transmitting the state parameters of the test set and the deep neural network prediction model to a memory module for calculation to obtain a calculation result of a one-dimensional column vector;
based on the calculation result of the one-dimensional column vector, reserving output information to be reserved in the test set, and outputting state parameters of a memory module at the current moment and state parameters of an hidden layer in the deep neural network prediction model at the current moment;
and calculating the calculation result of the one-dimensional column vector through an activation formula of a linear function to obtain a predicted result of the susceptibility of the debris flow disaster in the preset area.
9. The debris flow disaster vulnerability prediction method based on the deep neural network according to claim 8, wherein an activation formula of the linear function is as follows:
；
in the method, in the process of the invention,representing a predicted result of the susceptibility of the debris flow disaster in a preset area; />Representing an activation function; />A weight value representing data in the test set; />Representing the calculation result of the one-dimensional column vector; />Representing the prediction constants.
10. The method for predicting the susceptibility to debris flow disasters based on a deep neural network according to claim 1, wherein after inputting the test set into the deep neural network prediction model to obtain the predicted result of the susceptibility to debris flow disasters in a preset area, the method further comprises:
according to the predicted result of the debris flow disaster vulnerability in the preset area, performing a simulation experiment to obtain an experimental result;
and carrying out the accuracy evaluation of the debris flow disaster susceptibility prediction according to the experimental result.
CN202311722664.4A 2023-12-15 2023-12-15 Debris flow disaster vulnerability prediction method based on deep neural network Pending CN117408167A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311722664.4A CN117408167A (en) 2023-12-15 2023-12-15 Debris flow disaster vulnerability prediction method based on deep neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311722664.4A CN117408167A (en) 2023-12-15 2023-12-15 Debris flow disaster vulnerability prediction method based on deep neural network

Publications (1)

Publication Number Publication Date
CN117408167A true CN117408167A (en) 2024-01-16

Family

ID=89492984

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311722664.4A Pending CN117408167A (en) 2023-12-15 2023-12-15 Debris flow disaster vulnerability prediction method based on deep neural network

Country Status (1)

Country Link
CN (1) CN117408167A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117782226A (en) * 2024-02-23 2024-03-29 四川省能源地质调查研究所 Mine safety early warning system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110348624A (en) * 2019-07-04 2019-10-18 内蒙古工业大学 A kind of classification of sandstorm intensity prediction technique based on Stacking Integrated Strategy
CN111639783A (en) * 2020-04-17 2020-09-08 中国电力科学研究院有限公司 Line loss prediction method and system based on LSTM neural network
CN111968019A (en) * 2020-07-31 2020-11-20 中国地质大学（武汉） Landslide hazard susceptibility prediction method based on CNN and RNN integration
WO2023020896A1 (en) * 2021-08-19 2023-02-23 Iceye Oy Image difference identification

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110348624A (en) * 2019-07-04 2019-10-18 内蒙古工业大学 A kind of classification of sandstorm intensity prediction technique based on Stacking Integrated Strategy
CN111639783A (en) * 2020-04-17 2020-09-08 中国电力科学研究院有限公司 Line loss prediction method and system based on LSTM neural network
CN111968019A (en) * 2020-07-31 2020-11-20 中国地质大学（武汉） Landslide hazard susceptibility prediction method based on CNN and RNN integration
WO2023020896A1 (en) * 2021-08-19 2023-02-23 Iceye Oy Image difference identification

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
WENJUAN LI等: "Stacking ensemble of deep learning methods for landslide susceptibility mapping in the Three Gorges Reservoir area, China", STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT, pages 1 - 22 *
王毅等: "融合深度神经网络的三峡库区滑坡灾害易发性预测", 资源环境与工程, vol. 35, no. 5, pages 652 - 660 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117782226A (en) * 2024-02-23 2024-03-29 四川省能源地质调查研究所 Mine safety early warning system
CN117782226B (en) * 2024-02-23 2024-05-14 四川省能源地质调查研究所 Mine safety early warning system

Similar Documents

Publication Publication Date Title
CN112382352B (en) Method for quickly evaluating structural characteristics of metal organic framework material based on machine learning
CN109800863B (en) Logging phase identification method based on fuzzy theory and neural network
CN112819207B (en) Geological disaster space prediction method, system and storage medium based on similarity measurement
CN110442666B (en) Mineral resource prediction method and system based on neural network model
CN112001110A (en) Structural damage identification monitoring method based on vibration signal space real-time recursive graph convolutional neural network
CN117408167A (en) Debris flow disaster vulnerability prediction method based on deep neural network
CN116108758A (en) Landslide susceptibility evaluation method
CN108764527B (en) Screening method for soil organic carbon library time-space dynamic prediction optimal environment variables
CN110533239A (en) A kind of smart city air quality high-precision measuring method
CN111428419A (en) Suspended sediment concentration prediction method and device, computer equipment and storage medium
Zhuang et al. Long-lead prediction of extreme precipitation cluster via a spatiotemporal convolutional neural network
CN116187835A (en) Data-driven-based method and system for estimating theoretical line loss interval of transformer area
Yuan An improved K-means clustering algorithm for global earthquake catalogs and earthquake magnitude prediction
CN112329804A (en) Naive Bayes lithofacies classification integrated learning method and device based on feature randomness
CN117078114A (en) Water quality evaluation method and system for water-bearing lakes under influence of diversion engineering
CN115660221B (en) Oil and gas reservoir economic recoverable reserve assessment method and system based on hybrid neural network
CN116720079A (en) Wind driven generator fault mode identification method and system based on multi-feature fusion
CN114819344A (en) Global space-time meteorological agricultural disaster prediction method based on key influence factors
CN113205274A (en) Quantitative ranking method for construction quality
CN113627522A (en) Image classification method, device and equipment based on relational network and storage medium
CN111210870A (en) Protein classification method
Wang et al. A performance evaluation framework for association mining in spatial data
CN112884286B (en) Comprehensive evaluation method for ground test scheme of aviation equipment
CN116050120B (en) Landslide hidden danger activity remote sensing evaluation modeling method, system and storage medium
CN113673015B (en) Intelligent system construction and parameter identification method for beam column end plate connection node optimization design

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