Disaster trend prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of disaster situation prediction, in particular to a disaster situation trend prediction method and device, electronic equipment and a storage medium. Performing waveform feature extraction on data in each monitoring data set according to a plurality of standard factor waveform sets, and constructing a plurality of feature vectors obtained by extraction into a feature matrix; pooling the plurality of feature matrixes to obtain a plurality of pooling matrixes; and finally, inputting the plurality of pooling matrixes into a disaster trend prediction model to obtain a plurality of disaster predictions. According to the embodiment of the invention, the waveform feature matrix is extracted from the monitoring data by referring to the standard waveform, the waveform feature matrix is pooled and then sent to the prediction model, and the evolution condition of the disaster situation of the monitoring point in a future period of time is given. Therefore, the prediction result is reliable and accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of disaster prediction, and in particular to a disaster trend prediction method, device, electronic equipment and storage medium. Background Art

[0002] Meteorological disasters refer to the direct or indirect damage caused by the atmosphere to human life safety, production construction, national defense construction, etc. Natural disasters can cause losses ranging from several million to several billion yuan, and can also cause countless casualties in the disaster area.

[0003] Take frost as an example. It is a common agricultural meteorological disaster. It refers to the sudden drop in air temperature and the sudden drop in surface temperature to below 0℃, which damages crops or even kills them. Frost usually occurs in autumn, winter and spring. It is different from "frost". "Frost" is white ice crystals that condense directly on objects when the water vapor in the air near the ground reaches saturation and the ground temperature is below 0℃. Frost does not necessarily occur when frost occurs, and frost does not necessarily occur when frost occurs.

[0004] At present, the trend forecast of meteorological disasters is mainly based on risk estimation. For example, risk assessment is conducted on high-risk areas of meteorological disasters through weather phenomena, and early warning instructions are issued when necessary. This method is not accurate enough in terms of the occurrence area and the evolution of the disaster, and has little guiding significance.

[0005] Based on this, it is necessary to develop and design a disaster trend prediction method. Summary of the invention

[0006] The embodiments of the present invention provide a disaster trend prediction method, device, electronic device and storage medium, which are used to solve the problem that the disaster trend prediction results in the prior art are not accurate enough.

[0007] In a first aspect, an embodiment of the present invention provides a disaster trend prediction method, comprising:

[0008] Acquire multiple monitoring data sets, where each monitoring data set corresponds to a monitoring point, and the monitoring data set represents the volatility of factors affecting the change of disaster situation over time;

[0009] Extracting waveform features of the data in each monitoring data set according to a plurality of standard factor waveform sets, and constructing a plurality of feature vectors obtained by extraction into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the fluctuation characteristics of the factor;

[0010] Pooling multiple feature matrices respectively to obtain multiple pooling matrices, where each pooling matrix corresponds to a monitoring point;

[0011] The multiple pooling matrices are input into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point.

[0012] In a possible implementation, the waveform features of the data in each monitoring data set are extracted according to the multiple standard factor waveform sets, and the multiple feature vectors obtained by extraction are constructed into a feature matrix, including:

[0013] traversally extracting a standard factor waveform set from the plurality of standard factor waveform sets as a reference waveform set;

[0014] Extracting the monitoring data queue of the factors corresponding to the reference waveform set from the monitoring data set as the data queue to be processed;

[0015] Cutting the queue of data to be processed into a plurality of first data segments, wherein the length of the first data segment is the same as the length of the waveform segment in the reference waveform set;

[0016] Comparing the plurality of first data segments for similarity using waveform segments in the reference waveform set to obtain a plurality of similarity value queues, wherein each similarity value queue corresponds to a waveform segment in the reference waveform set;

[0017] Calculating the mean of each similarity value queue in the multiple similarity value queues to obtain multiple queue means;

[0018] Adding a vector constructed according to the plurality of queue means to a feature matrix;

[0019] If the traversal of the plurality of standard factor waveform sets is not completed, the process jumps to the step of traversally extracting a standard factor waveform set from the plurality of standard factor waveform sets as a reference waveform set.

[0020] In a possible implementation, the using waveform segments from the reference waveform set to perform similarity comparison on the plurality of first data segments to obtain a plurality of similarity value queues includes:

[0021] Extracting first data segments in sequence from the plurality of first data segments as first data segments to be processed;

[0022] A plurality of similarity values ​​are determined according to a first formula, the first data segment to be processed, and waveform segments in the reference waveform set, wherein the first formula is:

[0023]

[0024] Where SM iis the similarity value between the i-th waveform segment in the reference waveform set and the first data segment to be processed, ws(sn) is the sn-th data of the first data segment to be processed, sws i (sn) is the sn-th data of the i-th waveform segment in the reference waveform set, and snmax is the total number of data in the waveform segments of the reference waveform set;

[0025] adding the plurality of similarity values ​​to similarity value queues corresponding to the waveform segments in the reference waveform set respectively;

[0026] If the traversal of the multiple first data segments is not completed, jump to the step of sequentially extracting first data segments from the multiple first data segments as first data segments to be processed.

[0027] In a possible implementation, the standard factor waveform set is obtained by clustering and standardizing multiple historical factor monitoring data segments, including:

[0028] Obtain multiple historical factor monitoring data segments and clustering parameters;

[0029] Clustering the plurality of historical factor monitoring data segments according to the clustering parameters to obtain a plurality of clusters;

[0030] constructing the class centers of the multiple classes into multiple class vectors;

[0031] Taking the unit vectors of the plurality of vector-like elements as a plurality of first vectors;

[0032] The correlation of each of the plurality of first vectors is calculated according to a second formula to obtain a plurality of correlation values, wherein the second formula is:

[0033]

[0034] In the formula, Orth i is the association value of the i-th first vector, cnmax is the total number of first vectors, A i is the i-th first vector, A cn is the cn-th vector except the ith first vector among the plurality of first vectors;

[0035] If there is a correlation value higher than the threshold value among the multiple correlation values, the clustering parameter is adjusted with the goal of adjusting the number of clusters, and the process jumps to the step of clustering the multiple historical factor monitoring data segments according to the clustering parameter to obtain multiple clusters;

[0036] Otherwise, constructing a standard waveform segment according to multiple elements of each first vector, thereby obtaining multiple standard waveform segments;

[0037] The plurality of standard waveform segments are constructed as a standard factor waveform set.

[0038] In a possible implementation, the multiple feature matrices are pooled respectively to obtain multiple pooling matrices, including:

[0039] For each of the plurality of feature matrices, the following steps are performed respectively:

[0040] Sequentially take out data blocks from the feature matrix as the first data block;

[0041] Take out a plurality of first arrays from a first dimension of the first data block, wherein the first dimension is the dimension where the feature vector is constructed as a feature matrix;

[0042] Calculate the mean of each first array respectively, and construct the obtained multiple means into a second vector;

[0043] Adding the second vector to the pooling matrix according to the position of the first data block in the feature matrix;

[0044] If the traversal of the characteristic matrix is ​​not completed, the process jumps to the step of sequentially taking out data blocks from the characteristic matrix as the first data block.

[0045] In a possible implementation, the disaster trend prediction model is constructed based on multiple sample sets, including:

[0046] Acquire multiple coefficient arrays, a first model, and multiple sample sets, wherein each sample set includes a sample pooling matrix and a sample disaster data queue corresponding to the sample pooling matrix, the first model has multiple undetermined coefficients, independent variable inputs, and multiple dependent variable outputs, each coefficient array has the same number of data as the multiple coefficients, and the data of the coefficient array is randomly generated;

[0047] Substituting the multiple coefficient arrays into the first model respectively to obtain multiple second models;

[0048] traversally extracting a sample set from the multiple sample sets as a test sample set;

[0049] Inputting the sample pooling matrix of the test sample set as independent variables into the multiple second models respectively to obtain multiple disaster indication queues, wherein each disaster indication queue corresponds to a second model, and the disaster indication queue is constructed according to the multiple dependent variable outputs of the second model;

[0050] Determine a plurality of indication deviations according to the plurality of disaster indication queues and the sample disaster data queues of the test sample set, wherein each indication deviation corresponds to a disaster indication queue;

[0051] adding the plurality of indicated deviations to a plurality of deviation queues respectively, wherein each deviation queue corresponds to a coefficient array;

[0052] If there is no queue in the multiple deviation queues that satisfies the condition of being continuously below the deviation threshold, then for each coefficient array, adjustment is performed according to the state optimal coefficient array and the historical optimal coefficient array, wherein the state optimal coefficient array is the coefficient array corresponding to the minimum value of the multiple indicated deviations, the historical optimal coefficient array is the coefficient array corresponding to the minimum value in the target deviation queue, and the target deviation queue is the deviation queue corresponding to the coefficient array;

[0053] Otherwise, the corresponding coefficient array of the queue that satisfies the condition of being continuously lower than the deviation threshold is used as the target array;

[0054] The second model corresponding to the target array is used as a disaster trend prediction model.

[0055] In a possible implementation, the first model is:

[0056]

[0057] Where FDD(fm) is the fmth dependent variable output of the first model, α fm,idx is the first coefficient of the fm·idxth order, X is the combined independent variable, idxMax is the maximum index number, fm is the fmth bias coefficient, γ m,n is the m·nth second coefficient, PM(m,n) is the element in the mth row and nth column of the input matrix, nMax is the total number of rows of the input independent variables, and nMax is the total number of columns of the input independent variables;

[0058] For each coefficient array, adjustment is performed according to the state optimal coefficient array and the historical optimal coefficient array, including:

[0059] For each coefficient array, adjustment is performed according to the third formula, the state optimal coefficient array, and the historical optimal coefficient array, wherein the third formula is:

[0060]

[0061] In the formula, w (t+1) (fm) is the fmth coefficient of the coefficient array after the t+1th adjustment, α (t) (fm) is the fmth coefficient of the coefficient array after the tth adjustment, D optn and D opth are the first distance and the second distance respectively, k1 and k2 are the first coefficient and the second coefficient respectively, α optn (fm) and α optn(fm) are the fmth coefficient of the state optimal coefficient array and the fmth coefficient of the historical optimal coefficient array respectively;

[0062] The determining of a plurality of indication deviations according to the plurality of disaster indication queues and the sample disaster data queue of the test sample set comprises:

[0063] According to the fourth formula, the multiple disaster indication queues and the sample disaster data queues of the test sample set, multiple indication deviations are determined, wherein the fourth formula is:

[0064]

[0065] Where Bias is the indication bias, FDD(fm) is the fmth data in the disaster indication queue, RD(fm) is the fmth data in the sample disaster data queue of the test sample set, and fmMax is the total number of data in the disaster indication queue.

[0066] In a second aspect, an embodiment of the present invention provides a disaster trend prediction device, which is used to implement the disaster trend prediction method described in the first aspect or any possible implementation of the first aspect, wherein the disaster trend prediction device includes:

[0067] A monitoring data acquisition module is used to acquire multiple monitoring data sets, wherein each monitoring data set corresponds to a monitoring point, and the monitoring data set represents the volatility of factors affecting the change of disaster situation over time;

[0068] A feature extraction module is used to extract waveform features of the data in each monitoring data set according to multiple standard factor waveform sets, and construct multiple feature vectors obtained by extraction into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the fluctuation characteristics of the factor;

[0069] A pooling module is used to pool multiple feature matrices respectively to obtain multiple pooling matrices, where each pooling matrix corresponds to a monitoring point;

[0070] as well as,

[0071] The disaster prediction module is used to input the multiple pooling matrices into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point.

[0072] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the steps of the method described in the first aspect or any possible implementation method of the first aspect.

[0073] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect or any possible implementation of the first aspect are implemented.

[0074] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0075] The embodiment of the present invention discloses a method for disaster trend prediction, which first obtains multiple monitoring data sets, wherein each monitoring data set corresponds to a monitoring point, and the monitoring data set characterizes the volatility of factors affecting disaster changes over time; then, waveform features are extracted from the data in each monitoring data set according to multiple standard factor waveform sets, and the multiple feature vectors obtained by extraction are constructed into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the volatility characteristics of the factor; then, the multiple feature matrices are pooled respectively to obtain multiple pooled matrices, wherein each pooled matrix corresponds to a monitoring point; finally, the multiple pooled matrices are input into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point. The implementation mode of the present invention extracts the monitoring data into a waveform feature matrix by referring to the standard waveform, and sends it to the prediction model after pooling, so as to give the evolution of the disaster situation at the monitoring point in the future period of time. Since the implementation mode of the present invention performs prediction by extracting features from multi-dimensional data for the points, the prediction result is more reliable and accurate. Moreover, the implementation mode of the present invention does not adopt a general artificial intelligence algorithm to improve the prediction, and there is no problem of requiring massive samples and more human resources to invest in improving the model. It has good universality and is easy to promote. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0077] Figure 1 is a flow chart of a disaster trend prediction method provided by an embodiment of the present invention;

[0078] Figure 2 It is a principle diagram of the characteristic matrix construction process provided by an embodiment of the present invention;

[0079] Figure 3It is a schematic diagram of the process of constructing a standard factor waveform set provided by an embodiment of the present invention;

[0080] Figure 4 It is a schematic diagram of the disaster trend prediction model construction process provided by an embodiment of the present invention;

[0081] Figure 5 It is a functional block diagram of a disaster trend prediction device provided by an embodiment of the present invention;

[0082] Figure 6 It is a functional block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0083] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, and methods are omitted so as not to obscure the description of the present invention with unnecessary details.

[0084] In order to make the purpose, technical solutions and advantages of the present invention more clear, a specific implementation method will be described below in conjunction with the accompanying drawings.

[0085] The following is a detailed description of an embodiment of the present invention. This example is implemented based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0086] Figure 1 A flowchart of a disaster trend prediction method provided in an embodiment of the present invention.

[0087] like Figure 1 As shown, it shows a flowchart of the disaster situation trend prediction method provided by an embodiment of the present invention, which is described in detail as follows:

[0088] In step 101, a plurality of monitoring data sets are obtained, wherein each monitoring data set corresponds to a monitoring point, and the monitoring data set represents the volatility of factors affecting the change of disaster situation over time.

[0089] In step 102, waveform features are extracted from the data in each monitoring data set according to a plurality of standard factor waveform sets, and a plurality of feature vectors obtained by extraction are constructed into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the fluctuation characteristics of the factor.

[0090] In some implementations, step 102 includes:

[0091] traversally extracting a standard factor waveform set from the plurality of standard factor waveform sets as a reference waveform set;

[0092] Extracting the monitoring data queue of the factors corresponding to the reference waveform set from the monitoring data set as the data queue to be processed;

[0093] Cutting the queue of data to be processed into a plurality of first data segments, wherein the length of the first data segment is the same as the length of the waveform segment in the reference waveform set;

[0094] Comparing the plurality of first data segments for similarity using waveform segments in the reference waveform set to obtain a plurality of similarity value queues, wherein each similarity value queue corresponds to a waveform segment in the reference waveform set;

[0095] Calculating the mean of each similarity value queue in the multiple similarity value queues to obtain multiple queue means;

[0096] Adding a vector constructed according to the plurality of queue means to a feature matrix;

[0097] If the traversal of the plurality of standard factor waveform sets is not completed, the process jumps to the step of traversally extracting a standard factor waveform set from the plurality of standard factor waveform sets as a reference waveform set.

[0098] In some implementations, the using waveform segments from the reference waveform set to perform similarity comparison on the plurality of first data segments to obtain a plurality of similarity value queues includes:

[0099] Extracting first data segments in sequence from the plurality of first data segments as first data segments to be processed;

[0100] A plurality of similarity values ​​are determined according to a first formula, the first data segment to be processed, and waveform segments in the reference waveform set, wherein the first formula is:

[0101]

[0102] Where SM i is the similarity value between the i-th waveform segment in the reference waveform set and the first data segment to be processed, ws(sn) is the sn-th data of the first data segment to be processed, sws i (sn) is the sn-th data of the i-th waveform segment in the reference waveform set, and snmax is the total number of data in the waveform segments of the reference waveform set;

[0103] adding the plurality of similarity values ​​to similarity value queues corresponding to the waveform segments in the reference waveform set respectively;

[0104] If the traversal of the multiple first data segments is not completed, jump to the step of sequentially extracting first data segments from the multiple first data segments as first data segments to be processed.

[0105] In some implementations, the standard factor waveform set is obtained by clustering and standardizing a plurality of historical factor monitoring data segments, including:

[0106] Obtain multiple historical factor monitoring data segments and clustering parameters;

[0107] Clustering the plurality of historical factor monitoring data segments according to the clustering parameters to obtain a plurality of clusters;

[0108] constructing the class centers of the multiple classes into multiple class vectors;

[0109] Taking the unit vectors of the plurality of vector-like elements as a plurality of first vectors;

[0110] The correlation of each of the plurality of first vectors is calculated according to a second formula to obtain a plurality of correlation values, wherein the second formula is:

[0111]

[0112] In the formula, Orth i is the association value of the i-th first vector, cnmax is the total number of first vectors, A i is the i-th first vector, A cn is the cn-th vector except the ith first vector among the plurality of first vectors;

[0113] If there is a correlation value higher than the threshold value among the multiple correlation values, the clustering parameter is adjusted with the goal of adjusting the number of clusters, and the process jumps to the step of clustering the multiple historical factor monitoring data segments according to the clustering parameter to obtain multiple clusters;

[0114] Otherwise, constructing a standard waveform segment according to multiple elements of each first vector, thereby obtaining multiple standard waveform segments;

[0115] The plurality of standard waveform segments are constructed as a standard factor waveform set.

[0116] For example, as mentioned above, meteorological disasters are generally highly regionalized. The current mainstream way to predict the trend of meteorological disasters is to evaluate high-risk areas under severe meteorological conditions. When the evaluation conclusion shows that the risk is high, an early warning is issued for the high-risk areas. However, this method has certain uncertainty and blindness, and the area and time period where the disaster occurs are not accurately estimated.

[0117] One improved method is to use artificial intelligence algorithms, for example, deep learning algorithms to predict meteorological disasters and have achieved certain results. However, deep learning algorithms are usually a more universal algorithm, which requires the preparation of a large number of samples, and requires a lot of manpower and material resources for training and algorithm optimization. It is not suitable for regional and small sample predictions.

[0118] The implementation mode of the present invention intends to monitor data through multi-dimensional factors, extract features related to disaster changes, and use a pre-constructed feature-disaster prediction model based on these features to predict disaster trends at multiple locations. The complexity of the model and the complexity of the model construction process are smaller than those of the deep learning method, and it is highly targeted and easy to promote.

[0119] To achieve the above purpose, the embodiment of the present invention first monitors multiple monitoring points and constructs a feature matrix based on the data obtained from the monitoring points. Specifically, a monitoring point needs to monitor data related to the disaster and organize it into a monitoring data queue. Multiple queues are constructed as a monitoring data set of the monitoring point. For example, for frost prediction, its factors include sunshine intensity, temperature, humidity, surface temperature, etc. Taking temperature as an example, temperature data is continuously collected, and these temperature data are sorted according to time nodes to obtain a temperature data queue. Similarly, a sunshine intensity data queue, a humidity data queue, and a surface temperature data queue are obtained. These data queues of the monitoring point are constructed as a monitoring data set.

[0120] We can see that the prediction based on these monitoring data sets is relatively complex due to the large amount of data. The embodiments of the present invention are intended to extract the main features related to the disaster situation from these data queues, and more importantly, reduce the dimension of the data, thereby reducing the complexity of the model and the difficulty of application.

[0121] Based on the above starting point, the embodiment of the present invention constructs a standard factor waveform set for each factor, uses the waveform set to analyze the similarity value between the data queue and the waveform in the waveform set, regenerates the feature vector based on the similarity value, and finally constructs a matrix.

[0122] like Figure 2 As shown, the figure shows a schematic diagram of the characteristic matrix construction process provided by an embodiment of the present invention. The monitoring data queue is first divided into multiple monitoring data segments 201 (first data segments). Each monitoring data segment 201 is calculated with a waveform data segment in a standard factor waveform set 202 (reference waveform segments in the waveform set) for similarity calculation. In an application scenario, the first formula is used:

[0123]

[0124] Where SM iis the similarity value between the i-th waveform segment in the reference waveform set and the first data segment to be processed, ws(sn) is the sn-th data of the first data segment to be processed, sws i (sn) is the sn-th data of the i-th waveform segment in the reference waveform set, and snmax is the total number of data in the waveform segments of the reference waveform set.

[0125] In this way, multiple similarity values ​​are obtained, and these multiple similarity values ​​are added to the similarity value queue 203 corresponding to the waveform data segment. Repeat the above operation until all the multiple monitoring data segments 201 are processed, and then multiple complete similarity value queues 203 are obtained. These multiple similarity value queues 203 take the mean of the queue respectively, and obtain multiple means to construct a mean vector 204. After the mean vector 204 is added to the feature matrix 205, the next monitoring data queue is extracted to repeat the above steps (for example, the sunshine intensity data queue, the temperature data queue, the humidity data queue and the surface temperature data queue are extracted in turn), and the complete feature matrix 205 is obtained.

[0126] It can be seen that after the above operations, the monitoring data is standardized by calculating the similarity value with the standard waveform, and because the features are extracted through the mean operation, the amount of data is greatly compressed.

[0127] In terms of constructing a standard waveform, a variety of methods can be implemented. In one application scenario, the standard waveform is obtained by clustering, standardization, and modifying the clustering method through correlation analysis.

[0128] Specifically, if Figure 3 As shown, multiple historical factor monitoring data segments (for example, humidity data queues constructed by historical humidity monitoring data) are clustered into multiple classes 301 (the clustering method can be a common clustering method such as DBSCAN and kmeans), the class center of each class is taken as a vector, and the vector is normalized to obtain a standardized vector 302, and each standardized vector 302 uses the second formula to calculate the association value:

[0129]

[0130] In the formula, Orth i is the association value of the i-th first vector, cnmax is the total number of first vectors, A i is the i-th first vector, A cn is the cn-th vector among the multiple first vectors except the ith first vector.

[0131] If the correlation values ​​of all the standardized vectors 302 are less than the threshold, the standardized vectors 302 are used as standard waveform data segments and constructed into a standard factor waveform set. Otherwise, the clustering parameters are modified and clustering is performed again. Generally speaking, the clustering parameters are adjusted toward a smaller number of clusters.

[0132] In step 103, multiple feature matrices are pooled respectively to obtain multiple pooling matrices, where each pooling matrix corresponds to a monitoring point.

[0133] In some implementations, step 103 includes:

[0134] For each of the plurality of feature matrices, the following steps are performed respectively:

[0135] Sequentially take out data blocks from the feature matrix as the first data block;

[0136] Take out a plurality of first arrays from a first dimension of the first data block, wherein the first dimension is the dimension where the feature vector is constructed as a feature matrix;

[0137] Calculate the mean of each first array respectively, and construct the obtained multiple means into a second vector;

[0138] Adding the second vector to the pooling matrix according to the position of the first data block in the feature matrix;

[0139] If the traversal of the characteristic matrix is ​​not completed, the process jumps to the step of sequentially taking out data blocks from the characteristic matrix as the first data block.

[0140] Exemplarily, in terms of pooling, the embodiment of the present invention pools each feature matrix according to the dimension where the feature vector is located, such as Figure 2 As shown, the feature vector (mean vector 204) shown in the figure is a column vector. During pooling, a data block is taken out from the feature matrix 205, the mean of each data block is calculated by column, and the mean is placed in the corresponding column to obtain a second vector. Then, the second vector is placed in the pooling matrix according to the position of the taken data block in the feature matrix 205. The above steps are repeated to obtain the pooling matrix.

[0141] The role of the pooling matrix is ​​to reduce the dimension of the data and eliminate small-value data in the feature matrix, making the matrix more representative.

[0142] In step 104, the multiple pooling matrices are input into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point.

[0143] In some implementations, the disaster trend prediction model is constructed based on multiple sample sets, including:

[0144] Acquire multiple coefficient arrays, a first model, and multiple sample sets, wherein each sample set includes a sample pooling matrix and a sample disaster data queue corresponding to the sample pooling matrix, the first model has multiple undetermined coefficients, independent variable inputs, and multiple dependent variable outputs, each coefficient array has the same number of data as the multiple coefficients, and the data of the coefficient array is randomly generated;

[0145] Substituting the multiple coefficient arrays into the first model respectively to obtain multiple second models;

[0146] traversally extracting a sample set from the multiple sample sets as a test sample set;

[0147] Inputting the sample pooling matrix of the test sample set as independent variables into the multiple second models respectively to obtain multiple disaster indication queues, wherein each disaster indication queue corresponds to a second model, and the disaster indication queue is constructed according to the multiple dependent variable outputs of the second model;

[0148] Determine a plurality of indication deviations according to the plurality of disaster indication queues and the sample disaster data queues of the test sample set, wherein each indication deviation corresponds to a disaster indication queue;

[0149] adding the plurality of indicated deviations to a plurality of deviation queues respectively, wherein each deviation queue corresponds to a coefficient array;

[0150] If there is no queue in the multiple deviation queues that satisfies the condition of being continuously below the deviation threshold, then for each coefficient array, adjustment is performed according to the state optimal coefficient array and the historical optimal coefficient array, wherein the state optimal coefficient array is the coefficient array corresponding to the minimum value of the multiple indicated deviations, the historical optimal coefficient array is the coefficient array corresponding to the minimum value in the target deviation queue, and the target deviation queue is the deviation queue corresponding to the coefficient array;

[0151] Otherwise, the corresponding coefficient array of the queue that satisfies the condition of being continuously lower than the deviation threshold is used as the target array;

[0152] The second model corresponding to the target array is used as a disaster trend prediction model.

[0153] In some embodiments, the first model is:

[0154]

[0155] Where FDD(fm) is the fmth dependent variable output of the first model, α fm,idxis the first coefficient of the fm·idxth order, X is the combined independent variable, idxMax is the maximum exponential order, β fm is the fmth bias coefficient, γ m,n is the m·nth second coefficient, PM(m,n) is the element in the mth row and nth column of the input matrix, mMax is the total number of rows of the input independent variables, and nMax is the total number of columns of the input independent variables;

[0156] For each coefficient array, adjustment is performed according to the state optimal coefficient array and the historical optimal coefficient array, including:

[0157] For each coefficient array, adjustment is performed according to the third formula, the state optimal coefficient array, and the historical optimal coefficient array, wherein the third formula is:

[0158]

[0159] In the formula, w (t+1) (fm) is the fmth coefficient of the coefficient array after the t+1th adjustment, α (t) (fm) is the fmth coefficient of the coefficient array after the tth adjustment, D optn and D opth are the first distance and the second distance respectively, k1 and k2 are the first coefficient and the second coefficient respectively, α optn (fm) and α optn (fm) are the fmth coefficient of the state optimal coefficient array and the fmth coefficient of the historical optimal coefficient array respectively;

[0160] The determining of a plurality of indication deviations according to the plurality of disaster indication queues and the sample disaster data queue of the test sample set comprises:

[0161] According to the fourth formula, the multiple disaster indication queues and the sample disaster data queues of the test sample set, multiple indication deviations are determined, wherein the fourth formula is:

[0162]

[0163] Where Bias is the indication bias, FDD(fm) is the fmth data in the disaster indication queue, RD(fm) is the fmth data in the sample disaster data queue of the test sample set, and fmMax is the total number of data in the disaster indication queue.

[0164] Exemplarily, the pooling matrices obtained in the above steps are respectively input into the model to obtain the prediction results of multiple monitoring points. Here, the construction process of the model has to be explained.

[0165] The model gives the trend of disaster changes at the point in the future period of time based on the pooling matrix (for example, the frost level of frost at different time points in the future). Therefore, when the model is constructed, the applied data is also a (sample) pooling matrix. In the embodiment of the present invention, the sample pooling matrix and the sample disaster data queue corresponding to the sample pooling matrix (disaster data later than the sample pooling matrix time period) are constructed as a sample set.

[0166] The model of the embodiment of the present invention is:

[0167]

[0168] Where FDD(fm) is the fmth dependent variable output of the first model, α fm,idx is the first coefficient of the fm·idxth order, X is the combined independent variable, idxMax is the maximum exponential order, β fm is the fmth bias coefficient, γ m,n is the m·nth second coefficient, PM(m,n) is the element in the mth row and nth column of the input matrix, mMax is the total number of rows of the input independent variables, and nMax is the total number of columns of the input independent variables.

[0169] From the above model, it can be seen that the model has many coefficients whose values ​​need to be determined. Figure 4 As shown, the embodiment of the present invention obtains and initializes multiple arrays 402, and these multiple arrays 402 are respectively substituted into the above-mentioned model 401 to obtain multiple intermediate models 403 (second models).

[0170] The sample pooling matrix in the sample set 404 is distributedly input into the multiple intermediate models 403 to obtain multiple disaster indication queues 402 (the disaster indication queues and the intermediate models are in a one-to-one relationship). We can know that the deviations between the multiple disaster indication queues 402 and the sample disaster data queues in the sample set 404 are different. The embodiment of the present invention uses the following formula to calculate the deviation:

[0171]

[0172] Where Bias is the indication bias, FDD(fm) is the fmth data in the disaster indication queue, RD(fm) is the fmth data in the sample disaster data queue of the test sample set, and fmMax is the total number of data in the disaster indication queue.

[0173] That is to say, each intermediate model 403 corresponds to a deviation, and these deviations are added to the deviation queue of the intermediate model 403 (the deviation queue is not shown). If the last few digits in these deviation queues are all lower than the deviation threshold (for example, the last five digits of the deviation queue are all lower than the threshold 0.03, 3%), then these arrays 402 need to be modified. After the modification is completed, a new sample set 404 is taken out and the above steps are repeated. Otherwise, the intermediate model 403 corresponding to the deviation queue whose last few digits are all lower than the deviation threshold is selected as the model with successful iteration, which is the disaster trend prediction model.

[0174] In terms of modifying the array, the embodiment of the present invention refers to the current optimal intermediate model 403 and the historical optimal intermediate model (the historical optimal intermediate model of the modified array) to make modifications, and applies the following formula:

[0175]

[0176] In the formula, w (t+1) (fm) is the fmth coefficient of the coefficient array after the t+1th adjustment, α (t) (fm) is the fmth coefficient of the coefficient array after the tth adjustment, D optn and D opth are the first distance and the second distance respectively, k1 and k2 are the first coefficient and the second coefficient respectively, α optn (fm) and α optn (fm) are the fmth coefficient of the state optimal coefficient array and the fmth coefficient of the historical optimal coefficient array respectively.

[0177] After the above steps, the model is successfully constructed, and the pooling matrix is ​​substituted into the model to obtain the disaster trend at future time nodes. After the pooling matrices of multiple monitoring points have obtained trends, when they are expressed in a graphical way, the evolution of the disaster can be intuitively observed.

[0178] The disaster trend prediction method of the present invention is implemented as follows: first, multiple monitoring data sets are obtained, wherein each monitoring data set corresponds to a monitoring point, and the monitoring data set characterizes the volatility of factors affecting disaster changes over time; then, waveform features are extracted from the data in each monitoring data set according to multiple standard factor waveform sets, and the multiple feature vectors obtained by extraction are constructed into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the volatility characteristics of the factor; then, the multiple feature matrices are pooled respectively to obtain multiple pooled matrices, wherein each pooled matrix corresponds to a monitoring point; finally, the multiple pooled matrices are input into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point. The implementation mode of the present invention extracts the monitoring data into a waveform feature matrix by referring to the standard waveform, and sends it to the prediction model after pooling, so as to give the evolution of the disaster situation at the monitoring point in the future period of time. Since the implementation mode of the present invention performs prediction by extracting features from multi-dimensional data for the points, the prediction result is more reliable and accurate. Moreover, the implementation mode of the present invention does not adopt a general artificial intelligence algorithm to improve the prediction, and there is no problem of requiring massive samples and more human resources to invest in improving the model. It has good universality and is easy to promote.

[0179] It should be understood that the size of the serial numbers of the steps in the above implementation does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the implementation method of the present invention.

[0180] The following is an implementation of the device of the present invention. For details not described in detail, reference may be made to the corresponding method implementation described above.

[0181] Figure 5 is a functional block diagram of a disaster trend prediction device provided by an embodiment of the present invention, referring to Figure 5 The disaster trend prediction device includes: a monitoring data acquisition module 501, a feature extraction module 502, a pooling module 503 and a disaster prediction module 504, wherein:

[0182] The monitoring data acquisition module 501 is used to acquire multiple monitoring data sets, wherein each monitoring data set corresponds to a monitoring point, and the monitoring data set represents the volatility of factors affecting the change of disaster situation over time;

[0183] A feature extraction module 502 is used to extract waveform features from the data in each monitoring data set according to a plurality of standard factor waveform sets, and construct a plurality of feature vectors obtained by extraction into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the fluctuation characteristics of the factor;

[0184] A pooling module 503, used to pool the multiple feature matrices respectively to obtain multiple pooling matrices, wherein each pooling matrix corresponds to a monitoring point;

[0185] The disaster prediction module 504 is used to input the multiple pooling matrices into the disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point.

[0186] Figure 6 is a functional block diagram of an electronic device provided by an embodiment of the present invention. Figure 6 As shown, the electronic device 6 of this embodiment includes: a processor 600 and a memory 601, wherein the memory 601 stores a computer program 602 that can be run on the processor 600. When the processor 600 executes the computer program 602, the steps in the above-mentioned disaster trend prediction method and the embodiment are implemented, for example Figure 1 Steps 101 to 104 are shown.

[0187] Exemplarily, the computer program 602 may be divided into one or more modules / units, and the one or more modules / units are stored in the memory 601 and executed by the processor 600 to implement the present invention.

[0188] The electronic device 6 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 6 may include, but is not limited to, a processor 600 and a memory 601. Those skilled in the art will appreciate that Figure 6 It is only an example of the electronic device 6 and does not constitute a limitation of the electronic device 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 6 may also include input and output devices, network access devices, buses, etc.

[0189] The processor 600 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0190] The memory 601 may be an internal storage unit of the electronic device 6, such as a hard disk or memory of the electronic device 6. The memory 601 may also be an external storage device of the electronic device 6, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 6. Further, the memory 601 may also include both an internal storage unit and an external storage device of the electronic device 6. The memory 601 is used to store the computer program 602 and other programs and data required by the electronic device 6. The memory 601 may also be used to temporarily store data that has been output or is to be output.

[0191] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the implementation method can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method implementation method, which will not be repeated here.

[0192] In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0193] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0194] In the embodiments provided by the present invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0195] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0196] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0197] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned implementation method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various methods and device implementation methods. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.

[0198] The above-described embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A disaster trend prediction method, characterized in that: include: Acquire multiple monitoring data sets, where each monitoring data set corresponds to a monitoring point, and the monitoring data set represents the volatility of factors affecting the change of disaster situation over time; Extracting waveform features from the data in each monitoring data set according to a plurality of standard factor waveform sets, and constructing a plurality of feature vectors obtained by extraction into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the fluctuation characteristics of the factor; Pooling multiple feature matrices respectively to obtain multiple pooling matrices, where each pooling matrix corresponds to a monitoring point; The multiple pooling matrices are input into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point.

2. The disaster trend prediction method according to claim 1, characterized in that: The waveform features of the data in each monitoring data set are extracted according to the multiple standard factor waveform sets, and the multiple feature vectors obtained by extraction are constructed into a feature matrix, including: traversally extracting a standard factor waveform set from the plurality of standard factor waveform sets as a reference waveform set; Extracting the monitoring data queue of the factors corresponding to the reference waveform set from the monitoring data set as the data queue to be processed; Cutting the queue of data to be processed into a plurality of first data segments, wherein the length of the first data segment is the same as the length of the waveform segment in the reference waveform set; Performing similarity comparison on the plurality of first data segments using waveform segments in the reference waveform set to obtain a plurality of similarity value queues, wherein each similarity value queue corresponds to a waveform segment in the reference waveform set; Calculating the mean of each similarity value queue in the multiple similarity value queues to obtain multiple queue means; Adding a vector constructed according to the plurality of queue means to a feature matrix; If the traversal of the plurality of standard factor waveform sets is not completed, the process jumps to the step of traversally extracting a standard factor waveform set from the plurality of standard factor waveform sets as a reference waveform set.

3. The disaster trend prediction method according to claim 1, characterized in that: The method of comparing the plurality of first data segments by similarity using waveform segments in the reference waveform set to obtain a plurality of similarity value queues includes: Extracting first data segments in sequence from the plurality of first data segments as first data segments to be processed; A plurality of similarity values ​​are determined according to a first formula, the first data segment to be processed, and waveform segments in the reference waveform set, wherein the first formula is: Where SM i is the similarity value between the i-th waveform segment in the reference waveform set and the first data segment to be processed, ws(sn) is the sn-th data of the first data segment to be processed, sws i (sn) is the sn-th data of the i-th waveform segment in the reference waveform set, and snmax is the total number of data in the waveform segments of the reference waveform set; adding the plurality of similarity values ​​to similarity value queues corresponding to the waveform segments in the reference waveform set respectively; If the traversal of the multiple first data segments is not completed, jump to the step of sequentially extracting first data segments from the multiple first data segments as first data segments to be processed.

4. The disaster trend prediction method according to claim 1, characterized in that: The standard factor waveform set is obtained by clustering and standardizing multiple historical factor monitoring data segments, including: Obtain multiple historical factor monitoring data segments and clustering parameters; Clustering the plurality of historical factor monitoring data segments according to the clustering parameters to obtain a plurality of clusters; constructing the class centers of the multiple classes into multiple class vectors; Taking the unit vectors of the plurality of vector-like elements as a plurality of first vectors; The correlation of each of the plurality of first vectors is calculated according to a second formula to obtain a plurality of correlation values, wherein the second formula is: In the formula, Orth i is the association value of the i-th first vector, cnmax is the total number of first vectors, A i is the i-th first vector, A cn is the cn-th vector except the ith first vector among the plurality of first vectors; If there is a correlation value higher than the threshold value among the multiple correlation values, the clustering parameter is adjusted with the goal of adjusting the number of clusters, and the process jumps to the step of clustering the multiple historical factor monitoring data segments according to the clustering parameter to obtain multiple clusters; Otherwise, constructing a standard waveform segment according to multiple elements of each first vector, thereby obtaining multiple standard waveform segments; The plurality of standard waveform segments are constructed as a standard factor waveform set.

5. The disaster trend prediction method according to claim 1, characterized in that: The multiple feature matrices are pooled respectively to obtain multiple pooling matrices, including: For each of the plurality of feature matrices, the following steps are performed respectively: Sequentially take out data blocks from the feature matrix as the first data block; Take out a plurality of first arrays from a first dimension of the first data block, wherein the first dimension is the dimension where the feature vector is constructed as a feature matrix; Calculate the mean of each first array respectively, and construct the obtained multiple means into a second vector; Adding the second vector to the pooling matrix according to the position of the first data block in the feature matrix; If the traversal of the characteristic matrix is ​​not completed, the process jumps to the step of sequentially taking out data blocks from the characteristic matrix as the first data block.

6. The disaster trend prediction method according to any one of claims 1 to 5, characterized in that: The disaster trend prediction model is constructed based on multiple sample sets, including: Acquire multiple coefficient arrays, a first model, and multiple sample sets, wherein each sample set includes a sample pooling matrix and a sample disaster data queue corresponding to the sample pooling matrix, the first model has multiple undetermined coefficients, independent variable inputs, and multiple dependent variable outputs, each coefficient array has the same number of data as the multiple coefficients, and the data of the coefficient array is randomly generated; Substituting the multiple coefficient arrays into the first model respectively to obtain multiple second models; traversally extracting a sample set from the multiple sample sets as a test sample set; Inputting the sample pooling matrix of the test sample set as independent variables into the multiple second models respectively to obtain multiple disaster indication queues, wherein each disaster indication queue corresponds to a second model, and the disaster indication queue is constructed according to the multiple dependent variable outputs of the second model; Determine a plurality of indication deviations according to the plurality of disaster indication queues and the sample disaster data queues of the test sample set, wherein each indication deviation corresponds to a disaster indication queue; adding the plurality of indicated deviations to a plurality of deviation queues respectively, wherein each deviation queue corresponds to a coefficient array; If there is no queue in the multiple deviation queues that satisfies the condition of being continuously below the deviation threshold, then for each coefficient array, adjustment is performed according to the state optimal coefficient array and the historical optimal coefficient array, wherein the state optimal coefficient array is the coefficient array corresponding to the minimum value of the multiple indicated deviations, the historical optimal coefficient array is the coefficient array corresponding to the minimum value in the target deviation queue, and the target deviation queue is the deviation queue corresponding to the coefficient array; Otherwise, the corresponding coefficient array of the queue that satisfies the condition of being continuously lower than the deviation threshold is used as the target array; The second model corresponding to the target array is used as a disaster trend prediction model.

7. The disaster trend prediction method according to claim 6, characterized in that: The first model is: Where FDD(fm) is the fmth dependent variable output of the first model, α fm,idx is the first coefficient of the fm·idxth order, X is the combined independent variable, idxMax is the maximum exponential order, β fm is the fmth bias coefficient, γ m,n is the m·nth second coefficient, PM(m,n) is the element in the mth row and nth column of the input matrix, mMax is the total number of rows of the input independent variables, and nMax is the total number of columns of the input independent variables; For each coefficient array, adjustment is performed according to the state optimal coefficient array and the historical optimal coefficient array, including: For each coefficient array, adjustment is performed according to the third formula, the state optimal coefficient array, and the historical optimal coefficient array, wherein the third formula is: In the formula, w (t+1) (fm) is the fmth coefficient of the coefficient array after the t+1th adjustment, α (t) (fm) is the fmth coefficient of the coefficient array after the tth adjustment, D optn and D opth are the first distance and the second distance respectively, k1 and k2 are the first coefficient and the second coefficient respectively, α optn (fm) and α optn (fm) are the fmth coefficient of the state optimal coefficient array and the fmth coefficient of the historical optimal coefficient array respectively; The determining of a plurality of indication deviations according to the plurality of disaster indication queues and the sample disaster data queue of the test sample set comprises: According to the fourth formula, the multiple disaster indication queues and the sample disaster data queues of the test sample set, multiple indication deviations are determined, wherein the fourth formula is: Where Bias is the indication bias, FDD(fm) is the fmth data in the disaster indication queue, RD(fm) is the fmth data in the sample disaster data queue of the test sample set, and fmMax is the total number of data in the disaster indication queue.

8. A disaster trend prediction device, characterized in that: Used to implement the disaster trend prediction method according to any one of claims 1 to 7, the disaster trend prediction device comprises: A monitoring data acquisition module is used to acquire multiple monitoring data sets, wherein each monitoring data set corresponds to a monitoring point, and the monitoring data set represents the volatility of factors affecting the change of disaster situation over time; A feature extraction module is used to extract waveform features of the data in each monitoring data set according to multiple standard factor waveform sets, and construct multiple feature vectors obtained by extraction into a feature matrix, wherein each standard factor waveform set corresponds to a factor, and the standard factor waveform set characterizes the fluctuation characteristics of the factor; A pooling module is used to pool multiple feature matrices respectively to obtain multiple pooling matrices, where each pooling matrix corresponds to a monitoring point; as well as, The disaster prediction module is used to input the multiple pooling matrices into a disaster trend prediction model to obtain multiple disaster predictions, wherein the disaster trend prediction model has multiple outputs, and each disaster prediction corresponds to a monitoring point.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method as claimed in any one of claims 1 to 7 are implemented.

Images