CN116843074A - Typhoon disaster damage prediction method based on CNN-LSTM model - Google Patents
Typhoon disaster damage prediction method based on CNN-LSTM model Download PDFInfo
- Publication number
- CN116843074A CN116843074A CN202310823486.8A CN202310823486A CN116843074A CN 116843074 A CN116843074 A CN 116843074A CN 202310823486 A CN202310823486 A CN 202310823486A CN 116843074 A CN116843074 A CN 116843074A
- Authority
- CN
- China
- Prior art keywords
- data
- typhoon
- disaster
- typhoon disaster
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000012360 testing method Methods 0.000 claims abstract description 29
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 17
- 230000007787 long-term memory Effects 0.000 claims abstract description 10
- 230000002265 prevention Effects 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000006403 short-term memory Effects 0.000 claims abstract 4
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000011176 pooling Methods 0.000 claims description 12
- 230000015654 memory Effects 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 6
- 238000003912 environmental pollution Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000009331 sowing Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- 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
- G06N3/0464—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
Abstract
The invention discloses a typhoon disaster damage prediction method based on a CNN-LSTM model, which comprises the steps of firstly collecting typhoon disaster data of a region to be predicted in the past year, wherein the typhoon disaster data consists of typhoon attribute data, disaster bearing body data, disaster prevention capacity data and typhoon damage data, classifying and sorting the collected data according to time sequence to obtain a data set, dividing the data set into a training set and a test set by adopting a leaving method, inputting the training set as the typhoon disaster damage data set into a convolutional neural network for extracting features, then sending the training set to a long-term and short-term memory network for processing to obtain a prediction model, testing the obtained prediction model by adopting the test set, and repeating training and testing operations of the model until the optimal typhoon disaster damage prediction model is obtained if the accuracy of the prediction model does not meet the requirement. The method solves the problem of low prediction precision in the existing typhoon disaster loss prediction method.
Description
Technical Field
The invention belongs to the technical field of typhoon disaster damage prediction, and particularly relates to a typhoon disaster damage prediction method based on a CNN-LSTM model.
Background
Typhoon disasters are a systematic, complex and multidisciplinary crossing research problem, and are always an important branch of the research field of disasters. In recent years, the contents of typhoon disaster research at home and abroad are also more and more rich, and disaster research topics such as typhoon path simulation, typhoon disaster early warning and forecasting, typhoon disaster multi-disaster chain reaction, typhoon disaster situation space-time distribution, typhoon disaster comprehensive risk evaluation, disaster pre-disaster and post-disaster loss evaluation and the like are mainly included. The disaster loss assessment is used as the key point of the current typhoon disaster research, the research method is also continuously updated, and the technical methods such as fuzzy mathematics, input-output, neural network, deep learning and the like are most widely applied.
At present, in the aspect of disaster damage assessment, prediction is mainly performed based on an artificial intelligent neural network algorithm, direct economic loss, indirect economic loss, collapsed and damaged houses, disaster-affected crop areas and the like are taken as prediction objects, a neural network model is combined to perform prediction demonstration analysis, so that the research progress of typhoon disaster-bearing body loss is promoted to a certain extent, but input factors of the neural network model are disaster-causing factors, namely typhoon attribute factors, and social factors influencing natural vulnerability of the disaster-bearing body and related to disaster prevention and reduction are ignored. In addition, the typhoon sample number is too small, so that the prediction accuracy is low. Therefore, a method for accurately predicting typhoon disaster loss is still needed.
Disclosure of Invention
Aiming at the defects, the invention discloses a typhoon disaster damage prediction method based on a CNN-LSTM model, which solves the problem of low prediction precision in the existing typhoon disaster damage prediction method.
The invention is realized by adopting the following technical scheme:
a typhoon disaster damage prediction method based on a CNN-LSTM model comprises the following steps:
(1) Collecting typhoon disaster data of the region to be predicted in the past year, wherein the typhoon disaster data consists of typhoon attribute data, disaster bearing body data, disaster prevention capability data and typhoon loss data;
the typhoon attribute data comprise maximum rainfall, maximum wind speed and central air pressure;
the disaster-bearing body data comprise population density, crop sowing area and average GDP;
the disaster prevention capability data comprise general budget expenditure, the number of sickbeds per thousand people and the number of medical care per thousand people;
the typhoon loss data comprise direct economic loss, casualties, house collapse and damage quantity, farmland damage quantity and environmental pollution data;
(2) Classifying and sorting the data collected in the step (1) according to a time sequence to obtain a data set, and then dividing the data set into a training set and a testing set by adopting a reserving method;
(3) Inputting the training set obtained in the step (2) as a typhoon disaster damage data set to a Convolutional Neural Network (CNN) to extract characteristics, then sending the training set to a long-short-term memory network (LSTM) to process the training set to obtain a prediction model, adopting the test set obtained in the step (2) to test the obtained prediction model, if the precision of the prediction model does not meet the requirement, inputting the typhoon disaster damage data set to the convolutional neural network to extract the characteristics after readjusting model parameters, sending the training set to the long-short-term memory network to process the training set to obtain a new prediction model, and repeating the operation until the precision of the prediction model meets the requirement to obtain the optimal typhoon disaster damage prediction model;
the Convolutional Neural Network (CNN) is a type of convolutional computational feedforward neural network with depth structure (see FIG. 10) consisting of an input layer, a convolutional layer, a pooling layer, a fullThe connection layer and the output layer are formed, and the input multidimensional time sequence is set asn is the length of the input time sequence, m is the feature dimension of the input, firstly, the convolution layer calculates and extracts features, different convolution kernels represent different feature extraction modes, different feature information can be obtained when the same time sequence information is subjected to local feature extraction, and one convolution kernel is assumed>(j<n,i<m) step size is 1, and padding is 0, then the convolution result of the input sequence x by using the convolution kernel can be expressed as:
equation 1:
equation 2:w in k,d For weights in the convolution kernel, x p+k,q+d Parameters in the corresponding input matrix after p and q are slid for the convolution kernel; f' is an activation function; and then, inputting the extracted characteristic information of the convolution layer into a pooling layer, wherein the pooling layer has the significance of reserving the most important characteristic information for characteristic dimension reduction and reducing the complexity of network calculation. And finally, transmitting the output of the pooling layer as input to a full-connection layer, wherein the output result of the full-connection layer is shown as a formula 3: x is x i ＝f i (w i ×x i-1 +b i ) Wherein x is i ，x i-1 Output results of two adjacent connecting layers, f i 、w i 、b i The method comprises the steps of respectively obtaining an activation function, a weight matrix and a bias matrix of an ith full-connection layer;
the long-short-term memory network (LSTM) is an improved learning algorithm of RNN, and by adding Input, forget, output gate structures, the cyclic neural network memorizes historical effective information and selects unimportant information, thereby solving the problem of long time sequence data to a certain extentPhase dependency problem, see FIG. 11, assuming an input sequence of (x 1 ,x 2 ,x 3 ,...,x t ) When the hidden layer state is (h 1 ,h 2 ,h 3 ,...,h t ) Then time t has:
equation 4: f (f) t ＝sigmod(w f ·[h t-1 ,x t ]+b f ) Equation 5: i.e t ＝sigmod(w i ·[h t-1 ,x t ]+b i ) Equation 6:equation 7: />Equation 8: o (O) t ＝sigmod(w o ·[h t-1 ,x t ]+b o ) Equation 9: h is a t ＝O t ×tanh(C t ) In f t 、i t 、O t Output information of Forget, input, output gate structures, h t-1 For the output information of the hidden layer at the previous moment, h t C for outputting hidden layer at current moment t-1 、C t History memory information of a moment and a current moment on a long-term memory chain respectively, < >>To update the feature cells of the long-term memory chain, w f 、w i 、w c 、w o Weight matrix of different connection layers respectively, b f 、b i 、b c 、b o For the bias of different connection layers, sigmod and tanh are two activation functions.
Further, the time span of typhoon disaster data collected in the step (1) is not less than 10 years.
Further, typhoon disaster data of the region to be predicted for approximately 20 years are collected in the step (1).
Further, the environmental pollution data in the step (1) include hydrologic data and air quality data.
Further, in the step (2), after normalization processing is performed on the data in the data set, a set aside method is adopted to divide the data set into a training set and a testing set. The data in the data set is normalized, so that the adaptability of the model to training data is improved, and the model is suitable for data with different scales of hundreds of millions, tens of thousands and the like.
Further, the ratio of the training set to the test set in the step (2) is (3-4): 1. The sizes of the training set and the testing set are reasonably distributed, and if the training set is too small, the trained model can not learn the rule of the whole original data set; too large a training set tends to learn the rules in the raw data, but too small a test set can result in the performance being tested being difficult to represent the true performance of the predictive model.
Further, in step (3), the obtained prediction model is tested by using the test set obtained in step (2), the average absolute error percentage and the root mean square error are used as indexes for judging the accuracy of the prediction model, the error threshold value is set to 15%, and the correlation coefficient between the input element and the result is used as a mark for judging the complexity degree of the modeling type. Setting the error thresholds to 15% respectively is a realistic problem for small samples.
Compared with the prior art, the technical scheme has the following beneficial effects:
according to the invention, aiming at the problems that sample data encountered in typhoon disaster damage prediction are small, rapid convergence cannot be carried out in model training and prediction precision is low, from historical typhoon data samples, disaster bearing body data and disaster prevention capability data are also included in addition to typhoon attribute data, a typhoon disaster damage prediction model based on combination of a Convolutional Neural Network (CNN) and a long-short-term memory network (LSTM) is adopted, the small-sample typhoon disaster damage prediction precision is improved, rapid convergence is realized, and a relatively accurate typhoon post-disaster damage prediction model is obtained.
Drawings
Fig. 1 is a flowchart of a typhoon disaster damage prediction method based on the CNN-LSTM model described in example 1.
Fig. 2 is a block diagram of the typhoon disaster damage prediction model described in example 1.
Fig. 3 is a test result of the typhoon disaster damage prediction model obtained in example 1, in which a curve 1 represents test data and a curve 2 represents actual data.
Fig. 4 shows the test errors of the typhoon disaster damage prediction model obtained in example 1.
Fig. 5 shows the test relative error of the typhoon disaster damage prediction model obtained in example 1.
Fig. 6 is a predicted value-actual value scatter diagram of the typhoon disaster damage prediction model test obtained in example 1.
FIG. 7 is a graph showing the results of normalization of the data set described in example 1, wherein curves 1-10 represent, in order, maximum wind speed, maximum rainfall, central air pressure, population density, crop seed area, average GDP, general budget expenditure, number of beds per thousand people, number of nurses per thousand people, and direct economic loss.
Fig. 8 is a graph showing significant correlation coefficients between input data and output data described in embodiment 1.
Fig. 9 is a training schedule of the predictive model described in example 1.
Fig. 10 is a schematic structural diagram of the CNN according to the present invention.
Fig. 11 is a schematic view of the LSTM structure of the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are not intended to be limiting. The specific experimental conditions and methods not specified in the following examples are generally conventional means well known to those skilled in the art.
Example 1: as shown in fig. 1-2, a typhoon disaster damage prediction method based on a CNN-LSTM model includes the following steps:
(1) Collecting typhoon disaster data in Guangxi certain city 2001-2021, wherein the typhoon disaster data consists of typhoon attribute data, disaster bearing body data, disaster prevention capability data and typhoon loss data;
the typhoon attribute data are maximum rainfall (mm), maximum wind speed (m/s) and central air pressure (kPa);
the disaster-bearing body data are mankou density (square kilometers per man), crop sowing area (kilohectare) and average human GDP (ten thousand yuan);
the disaster prevention capability data are general budget expenditure (hundred million yuan), the number of sickbeds per thousand people (sheets/thousands people) and the number of medical care per thousand people (sheets/thousands people);
the typhoon loss data are direct economic losses (billions of yuan);
(2) Classifying and sorting the data collected in the step (1) according to a time sequence to obtain a data set, taking typhoon attribute data, disaster-bearing body data and disaster prevention capability data as input data, taking typhoon loss data as output data, carrying out normalization processing on data in the data set, and dividing the data set into a training set and a testing set by adopting a leaving method, wherein the ratio of the training set to the testing set is 4:1;
(3) Inputting the training set obtained in the step (2) as a typhoon disaster damage data set to a Convolutional Neural Network (CNN) to extract characteristics, then sending the training set to a long-short-period memory network (LSTM) to process the training set to obtain a prediction model, adopting the test set obtained in the step (2) to test the obtained prediction model, if the accuracy of the prediction model does not meet the requirement, inputting the typhoon disaster damage data set to the convolutional neural network to extract characteristics after readjusting model parameters, sending the training set to the long-short-period memory network to process the training set to obtain a new prediction model, repeating the operation until the accuracy of the prediction model meets the requirement, namely obtaining an optimal typhoon disaster damage prediction model, and adopting the test set to test the obtained prediction model, wherein average absolute error percentage and root mean square error are used as indexes for judging the accuracy of the prediction model, error threshold setting is respectively 15%, and the correlation coefficient between input elements and results is used as marks for judging the modeling type complexity;
the Convolutional Neural Network (CNN) is a type of convolutional computational feedforward neural network with a deep structure (see figure 10),it is composed of input layer, convolution layer, pooling layer, full connection layer and output layer, and has input multidimensional time sequence ofn is the length of the input time sequence, m is the feature dimension of the input, firstly, the convolution layer calculates and extracts features, different convolution kernels represent different feature extraction modes, different feature information can be obtained when the same time sequence information is subjected to local feature extraction, and one convolution kernel is assumed>(j<n,i<m) step size is 1, and padding is 0, then the convolution result of the input sequence x by using the convolution kernel can be expressed as:
equation 1:
equation 2:
w in k,d For weights in the convolution kernel, x p+k,q+d Parameters in the corresponding input matrix after p and q are slid for the convolution kernel; f' is an activation function; and then, inputting the extracted characteristic information of the convolution layer into a pooling layer, wherein the pooling layer has the significance of reserving the most important characteristic information for characteristic dimension reduction and reducing the complexity of network calculation. And finally, transmitting the output of the pooling layer as input to a full-connection layer, wherein the output result of the full-connection layer is shown as a formula 3: x is x i ＝f i (w i ×x i-1 +b i ) Wherein x is i ，x i-1 Output results of two adjacent connecting layers, f i 、w i 、b i The method comprises the steps of respectively obtaining an activation function, a weight matrix and a bias matrix of an ith full-connection layer;
the long-short-term memory network (LSTM) is an improved learning algorithm of RNN, and the memory history of the recurrent neural network is realized by adding Input, forget, output gate structuresThe long-term dependence problem of time series data is solved to a certain extent by effective information and selection of unimportant information, and the method is shown in fig. 11, assuming that the input sequence is (x 1 ,x 2 ,x 3 ,...,x t ) When the hidden layer state is (h 1 ,h 2 ,h 3 ,...,h t ) Then time t has:
equation 4: f (f) t ＝sigmod(w f ·[h t-1 ,x t ]+b f ) Equation 5: i.e t ＝sigmod(w i ·[h t-1 ,x t ]+b i ) Equation 6:equation 7: />Equation 8: o (O) t ＝sigmod(w o ·[h t-1 ,x t ]+b o ) Equation 9: h is a t ＝O t ×tanh(C t ) In f t 、i t 、O t Output information of Forget, input, output gate structures, h t-1 For the output information of the hidden layer at the previous moment, h t C for outputting hidden layer at current moment t-1 、C t History memory information of a moment and a current moment on a long-term memory chain respectively, < >>To update the feature cells of the long-term memory chain, w f 、w i 、w c 、w o Weight matrix of different connection layers respectively, b f 、b i 、b c 、b o For the bias of different connection layers, sigmod and tanh are two activation functions.
According to the typhoon disaster damage prediction model obtained by the method, model training is extremely fast in convergence, and the bottleneck of minimum damage can be achieved within 0.2 s; the model adopts extremely simple expression, avoids the model overfitting condition caused by deviation data, and can realize accurate grasp of the data characteristic rule in 30 iterations.
Furthermore, it should be understood that although the present disclosure describes embodiments, not every embodiment is provided with a separate embodiment, and that this description is provided for clarity only, and that the disclosure is not limited to the embodiments described in detail below, and that the embodiments described in the examples may be combined as appropriate to form other embodiments that will be apparent to those skilled in the art.
Claims (7)
1. A typhoon disaster damage prediction method based on a CNN-LSTM model is characterized in that: the method comprises the following steps:
(1) Collecting typhoon disaster data of the region to be predicted in the past year, wherein the typhoon disaster data consists of typhoon attribute data, disaster bearing body data, disaster prevention capability data and typhoon loss data;
the typhoon attribute data comprise maximum rainfall, maximum wind speed and central air pressure;
the disaster-bearing body data comprise population density, crop sowing area and average GDP;
the disaster prevention capability data comprise general budget expenditure, the number of sickbeds per thousand people and the number of medical care per thousand people;
the typhoon loss data comprise direct economic loss, casualties, house collapse and damage quantity, farmland damage quantity and environmental pollution data;
(2) Classifying and sorting the data collected in the step (1) according to a time sequence to obtain a data set, and then dividing the data set into a training set and a testing set by adopting a reserving method;
(3) Inputting the training set obtained in the step (2) as a typhoon disaster damage data set to a convolutional neural network to extract characteristics, then sending the training set to a long-term and short-term memory network to process to obtain a prediction model, adopting the test set obtained in the step (2) to test the obtained prediction model, if the precision of the prediction model does not meet the requirement, re-adjusting model parameters, inputting the typhoon disaster damage data set to the convolutional neural network to extract characteristics, then sending the training set to the long-term and short-term memory network to process to obtain a new prediction model, and repeating the operation until the precision of the prediction model meets the requirement to obtain the optimal typhoon disaster damage prediction model;
the convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a full-connection layer and an output layer, and the input multi-dimensional time sequence is as follows:
n is the length of the input time sequence, m is the dimension of the input feature, firstly, the convolution layer calculates and extracts the feature, different convolution kernels represent different feature extraction modes, different feature information can be obtained when the local feature extraction is carried out on the same time sequence information, and one convolution kernel is assumed to be:
(j<n,i<m) step size is 1, and padding is 0, then the convolution result of the input sequence x by using the convolution kernel can be expressed as:
equation 1:
equation 2:
w in k,d For weights in the convolution kernel, x p+k,q+d Parameters in the corresponding input matrix after p and q are slid for the convolution kernel; f' is an activation function; and then, inputting the extracted characteristic information of the convolution layer into a pooling layer, wherein the pooling layer has the significance of reserving the most important characteristic information for characteristic dimension reduction and reducing the complexity of network calculation. And finally, transmitting the output of the pooling layer as input to a full-connection layer, wherein the output result of the full-connection layer is shown as a formula 3: x is x i ＝f i (w i ×x i-1 +b i ) Wherein x is i ，x i-1 Output results of two adjacent connecting layers, f i 、w i 、b i The method comprises the steps of respectively obtaining an activation function, a weight matrix and a bias matrix of an ith full-connection layer;
the long and short term memory network assumes an input sequence of (x) 1 ,x 2 ,x 3 ,...,x t ) When the hidden layer state is (h 1 ,h 2 ,h 3 ,...,h t ) Then time t has:
equation 4: f (f) t ＝sigmod(w f ·[h t-1 ,x t ]+b f ) Equation 5: i.e t ＝sigmod(w i ·[h t-1 ,x t ]+b i ) Equation 6:equation 7: />Equation 8: o (O) t ＝sigmod(w o ·[h t-1 ,x t ]+b o ) Equation 9: h is a t ＝O t ×tanh(C t ) In f t 、i t 、O t Output information of Forget, input, output gate structures, h t-1 For the output information of the hidden layer at the previous moment, h t C for outputting hidden layer at current moment t-1 、C t History memory information of a moment and a current moment on a long-term memory chain respectively, < >>To update the feature cells of the long-term memory chain, w f 、w i 、w c 、w o Weight matrix of different connection layers respectively, b f 、b i 、b c 、b o For the bias of different connection layers, sigmod and tanh are two activation functions.
2. The typhoon disaster damage prediction method based on the CNN-LSTM model according to claim 1, wherein: the time span of typhoon disaster data collected in the step (1) is not less than 10 years.
3. The typhoon disaster damage prediction method based on the CNN-LSTM model according to claim 2, wherein: and (3) collecting typhoon disaster data of the region to be predicted for approximately 20 years in the step (1).
4. The typhoon disaster damage prediction method based on the CNN-LSTM model according to claim 1, wherein: the environmental pollution data in the step (1) comprise hydrologic data and air quality data.
5. The typhoon disaster damage prediction method based on the CNN-LSTM model according to claim 1, wherein: and (2) carrying out normalization processing on data in the data set, and then dividing the data set into a training set and a testing set by adopting a leave-out method.
6. The typhoon disaster damage prediction method based on the CNN-LSTM model according to claim 1, wherein: the ratio of the training set to the testing set in the step (2) is (3-4): 1.
7. The typhoon disaster damage prediction method based on the CNN-LSTM model according to claim 6, wherein: in step (3), the obtained prediction model is tested by adopting the test set obtained in step (2), the average absolute error percentage and the root mean square error are used as indexes for judging the accuracy of the prediction model, the error threshold value is set to be 15%, and the correlation coefficient between the input element and the result is used as a mark for judging the modeling type complexity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310823486.8A CN116843074A (en) | 2023-07-06 | 2023-07-06 | Typhoon disaster damage prediction method based on CNN-LSTM model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310823486.8A CN116843074A (en) | 2023-07-06 | 2023-07-06 | Typhoon disaster damage prediction method based on CNN-LSTM model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116843074A true CN116843074A (en) | 2023-10-03 |
Family
ID=88166759
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310823486.8A Pending CN116843074A (en) | 2023-07-06 | 2023-07-06 | Typhoon disaster damage prediction method based on CNN-LSTM model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116843074A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117592789A (en) * | 2024-01-18 | 2024-02-23 | 山东金桥保安器材有限公司 | Power grid environment fire risk assessment method and equipment based on time sequence analysis |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109902885A (en) * | 2019-04-09 | 2019-06-18 | 中国人民解放军国防科技大学 | Typhoon prediction method based on deep learning mixed CNN-LSTM model |
| CN112785066A (en) * | 2021-01-27 | 2021-05-11 | 北京师范大学 | Convolution-recurrent neural network-based global wildfire sub-season time-space prediction method |
| CN113642231A (en) * | 2021-07-09 | 2021-11-12 | 西北大学 | CNN-GRU landslide displacement prediction method based on compression excitation network and application |
| CN113723595A (en) * | 2021-09-02 | 2021-11-30 | 重庆大学 | Coal mine rock burst intelligent early warning method based on microseismic event quantitative prediction |
| CN116385239A (en) * | 2023-04-07 | 2023-07-04 | 应急管理部上海消防研究所 | Emergency management method based on dynamic perception fusion of disaster site information |
-
2023
- 2023-07-06 CN CN202310823486.8A patent/CN116843074A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109902885A (en) * | 2019-04-09 | 2019-06-18 | 中国人民解放军国防科技大学 | Typhoon prediction method based on deep learning mixed CNN-LSTM model |
| CN112785066A (en) * | 2021-01-27 | 2021-05-11 | 北京师范大学 | Convolution-recurrent neural network-based global wildfire sub-season time-space prediction method |
| CN113642231A (en) * | 2021-07-09 | 2021-11-12 | 西北大学 | CNN-GRU landslide displacement prediction method based on compression excitation network and application |
| CN113723595A (en) * | 2021-09-02 | 2021-11-30 | 重庆大学 | Coal mine rock burst intelligent early warning method based on microseismic event quantitative prediction |
| CN116385239A (en) * | 2023-04-07 | 2023-07-04 | 应急管理部上海消防研究所 | Emergency management method based on dynamic perception fusion of disaster site information |
Non-Patent Citations (2)
| Title |
|---|
| 叶小岭: "基于粒子群优化BP神经网络的台风灾损预测模型研究", 《灾害学》, vol. 28, no. 4, pages 11 - 15 * |
| 戴梦奇: "基于深度学习的城市暴雨内涝水深预测方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》, pages 25 - 43 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117592789A (en) * | 2024-01-18 | 2024-02-23 | 山东金桥保安器材有限公司 | Power grid environment fire risk assessment method and equipment based on time sequence analysis |
| CN117592789B (en) * | 2024-01-18 | 2024-04-16 | 山东金桥保安器材有限公司 | Power grid environment fire risk assessment method and equipment based on time sequence analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
| CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
| CN110348624B (en) | Sand storm grade prediction method based on Stacking integration strategy | |
| CN111612066A (en) | Remote sensing image classification method based on depth fusion convolutional neural network | |
| CN110929847A (en) | Converter transformer fault diagnosis method based on deep convolutional neural network | |
| CN112735097A (en) | Regional landslide early warning method and system | |
| CN104008164A (en) | Generalized regression neural network based short-term diarrhea multi-step prediction method | |
| CN103150611A (en) | Hierarchical prediction method of II type diabetes mellitus incidence probability | |
| CN116843074A (en) | Typhoon disaster damage prediction method based on CNN-LSTM model | |
| CN114781538A (en) | Air quality prediction method and system of GA-BP neural network coupling decision tree | |
| CN113257425B (en) | Influenza prediction system, storage medium and device for optimizing LSTM and LightGBM parameters | |
| CN104881400A (en) | Semantic dependency calculating method based on associative network | |
| CN112434887B (en) | Water supply network risk prediction method combining network kernel density estimation and SVM | |
| Anatya et al. | Fruit maturity classification using convolutional neural networks method | |
| CN112651499A (en) | Structural model pruning method based on ant colony optimization algorithm and interlayer information | |
| CN115877483A (en) | Typhoon path forecasting method based on random forest and GRU | |
| CN113657726B (en) | Personnel risk analysis method based on random forest | |
| CN105675807A (en) | Evaluation method of atrazine residue based on BP neural network | |
| CN115759095A (en) | Named entity recognition method and device for tobacco plant diseases and insect pests | |
| CN114757433A (en) | Method for quickly identifying relative risk of drinking water source antibiotic resistance | |
| An et al. | Forecasting daily PM₁₀ concentration in Seoul Jong-no District by using various statistical techniques | |
| CN114254828A (en) | Power load prediction method based on hybrid convolution feature extractor and GRU | |
| CN113962431A (en) | Bus load prediction method for two-stage feature processing | |
| Seyogaa et al. | A Comparison Between Backpropagation, Holt-Winter, and Polynomial Regression Methods in Forecasting Dog Bites Cases in Bali | |
| Sahafizadeh et al. | Prediction of air pollution of Boushehr city using data mining |
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 |