Forest fire risk prediction method

Abstract

The invention provides a forest fire risk prediction method, and relates to the field of machine learning. The forest fire risk prediction process provided by the invention comprises the steps of constructing a forest fire risk assessment data set, generating a data sequence feature vector, then sequentially passing through a data feature information interaction module and a spatial neighborhood feature interaction module, then converting a data sequence feature into a time sequence feature, and finally predicting the forest fire risk. A predicted value is obtained using a time sequence feature capture module and using a full-connection neural network. The data feature information interaction module captures relevance among different types of data, the spatial neighborhood feature interaction module aggregates feature information of adjacent regions, the time sequence feature capture module processes unified time sequence feature representation, accurate modeling of long-time sequence data is achieved, and by means of the three modules, accurate modeling of the long-time sequence data is achieved. The model can consider information of three dimensions of data, neighborhood and time, and the prediction precision of the model is further enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, and in particular relates to a forest fire risk prediction method. Background Art

[0002] Forest fire refers to a fire event that occurs in a forest or woodland area. It has a significant negative impact on the ecosystem, including widespread damage to vegetation, destruction of animal habitats, and deterioration of soil quality, which may in turn cause soil erosion and pose a threat to the life and property safety of nearby residents. Therefore, forest fire risk prediction is particularly important. Through scientific risk prediction, potential high-risk areas can be effectively identified, and appropriate prevention and response measures can be taken in a timely manner, thereby reducing the probability of fire and the losses it causes.

[0003] Machine learning has shown significant advantages in forest fire prediction due to its ability to process multiple data types simultaneously, process new data in real time, and has good scalability. Using it for forest fire prediction can automatically learn and identify potential fire risks from complex data, thereby achieving more accurate predictions. However, existing forest fire prediction methods based on machine learning simply splice and embed different types of data into a feature space, ignoring the multivariate correlations between them, thereby limiting the model's performance and generalization capabilities.

[0004] Forest fire risk assessment data usually presents highly nonlinear and complex relationships. Embedding each data sequence independently into a feature space helps machine learning models to more effectively learn and utilize the potential connections between different data, pay attention to spatial neighborhood characteristics, and then convert the data sequence information into time series information. Further utilizing the information in the time dimension can enable the model to more comprehensively and accurately warn of forest fire risks, help formulate effective fire prevention measures, and reduce losses caused by forest fires. Summary of the invention

[0005] The present invention provides a forest fire risk prediction method, which aims to learn more representative data feature representations by utilizing the potential connections between different data sources at the same location. Subsequently, a graph convolutional network is used to capture and aggregate feature information of spatial neighborhoods, so that the model can identify and learn the correlations and dependencies between adjacent areas. Finally, the data sequence features are converted into time series features, combined with the capture and processing of time series information, to dynamically track changes in fire risk and achieve accurate prediction of forest fire risk.

[0006] The technical method adopted by the present invention to achieve the above-mentioned purpose specifically comprises the following steps: S1. Collect relevant monitoring data of each forest area, including temperature, precipitation, wind speed, wind direction, relative humidity, vegetation type, accumulation of combustible materials on the ground, terrain orientation, and smoke content, clean the data, improve the data quality, and construct a forest fire risk assessment data set; S2, embed each type of data sequence generated by the forest area within a certain time interval into a feature space, and concatenate them into a feature vector of this type of data; S3. Build a data feature information interaction module, use the multi-head attention mechanism to make the feature information interact with the feature information of other different types of data, capture the correlation between different types of data, and thus generate richer and more representative feature embeddings; S4. Construct a spatial neighborhood feature interaction module to learn the correlation weights between the current forest area and its neighboring areas through graph convolutional networks (GCN), and use these weights to aggregate the feature information of the neighboring areas, so that the feature representation of the current forest area contains rich spatial neighborhood comprehensive information; S5, extracting information at each moment in the data sequence information in time sequence and splicing them to obtain time series information, wherein the information at each time step contains all data information of a specific forest area at the same time point; S6. Build a time series feature capture module to extract and process dependencies in long time series data through LSTM (Long Short-Term Memory) network to enhance the model's ability to understand and express time dynamic features; S7. Construct a fully connected neural network and calculate the final feature representation to obtain the final forest fire risk prediction value.

[0007] Preferably, in step S1, the monitoring data of each forest area includes temperature, precipitation, wind speed, wind direction, relative humidity, vegetation type, accumulation of combustible materials on the ground, terrain orientation, and smoke content. The collected data are preprocessed, including removing erroneous values ​​and outliers to ensure data accuracy, filling missing values ​​to ensure data integrity, and avoiding analysis bias caused by incomplete data.

[0008] Preferably, in step S2, the forest area monitoring data input is ,in is the number of data types, is the time series length of the forest area related monitoring data, For the The data in The monitoring value sequence of this data type obtained within a certain length of time, where for The monitoring value obtained at any time will be Obtain the corresponding eigenvector by looking up the table function: ; In the formula is a lookup table function, for The new feature vector obtained by looking up the table is After all the data is mapped through the lookup table, they are spliced ​​in chronological order. The specific process is as follows: ; In the formula For the Type of data in length time The unified feature representation of the monitoring value sequence obtained in is the data dimension after the lookup table, symbol For splicing operation.

[0009] Preferably, since different types of data have completely different physical meanings, each type of data is independently mapped into a feature vector and spliced ​​in time series. This can more effectively capture and learn the correlation between different data and avoid the information loss and correlation weakening problems that may result from mixing multiple data in one feature vector. This achieves flexible extraction of data features and retention of time dependencies, providing a solid foundation for subsequent models in learning and modeling multivariate correlations between data.

[0010] Preferably, in step S3, the number of attention heads used is The multi-head attention mechanism captures the correlation between different types of data, thereby generating richer and more representative feature embeddings. First, the input feature vector matrix Generate the query matrix through three linear transformations respectively , key matrix Sum Matrix , the specific calculation process is as follows: ; ; ; In the formula , and are query matrix, key matrix and value matrix respectively, , and are trainable weight matrices used to generate queries, keys, and values, respectively. is the dimension of the vector. Then, the dot product between the query and the key is calculated to obtain the attention score matrix. The specific calculation process is as follows: ; In the formula For the The attention score matrix of the attention heads, for function, , and For the The query, key, and value of each attention head, is the dimension of the key vector, is a scaling factor used to prevent the gradient from disappearing due to excessive dot product values. For the transpose operation, for each head After calculating the attention separately, the outputs of all heads are concatenated to obtain the output of the multi-head attention mechanism. The specific calculation process is as follows: ; In the formula is the final feature output, where For the The feature representation of class data is obtained through the multi-head attention mechanism. is the weight matrix of the output transformation after splicing, For splicing operation.

[0011] Preferably, the multi-head attention mechanism projects the current data feature representation into multiple subspaces, and calculates the attention weights with other input data in each subspace, so as to dynamically select the more relevant parts of other input data for aggregation, and finally, splice the outputs of each subspace together to capture the diverse feature representations in the data.

[0012] Preferably, in step S4, the correlation weights between the current forest area and its adjacent areas are learned through a graph convolutional network, and the feature information of the adjacent areas is aggregated using these weights. The specific calculation process is as follows: ; ; In the formula is the feature matrix obtained after the graph convolutional network, is the adjacency matrix between forest regions, where is the number of forest areas divided, is the identity matrix, is the adjacency matrix with self-loops added, for The degree matrix of The sum of each row of elements, is the trainable weight matrix.

[0013] Preferably, through the graph convolutional network, the representation of each forest area not only retains its own characteristics, but also integrates the characteristics of its neighboring areas, so that it can more effectively capture the information of spatial neighborhood characteristics. This feature integration method can improve the model's understanding and expression of the relationship between regions, thereby enhancing the ability to recognize the overall spatial pattern and enhancing the robustness and accuracy of the model.

[0014] Preferably, in step S5, first extract The feature representations of various types of data at the same time are concatenated to obtain the time series feature representation ,in For the Feature representation of the moment.

[0015] Preferably, in step S6, LSTM is used to capture long-term dependencies to obtain feature representations containing historical information. The specific calculation process is as follows: ; ; ; ; ; ; In the formula is the learnable weight parameter, is the bias parameter, for function, Timestamp The input gate, forget gate and output gate of LSTM. For timestamp The candidate memory cell at For timestamp The memory unit at For timestamp The hidden state of is element-wise multiplication.

[0016] Preferably, by splicing the feature representations of different types of data within the same time step to generate a unified time series feature representation, and combining the long short-term memory mechanism of LSTM to capture the long-term dependencies and multivariate correlations in the data, it not only effectively integrates the information of multi-source data and improves the expressive power of the model, but also enhances the prediction accuracy of complex time series data through dynamic memory management and learnable weight parameters, providing a solid foundation for more accurate forest fire risk prediction.

[0017] Preferably, in step S7, the output of the LSTM at the last moment in step S6 is taken as the final feature vector representation, and the feature vector is passed through a fully connected neural network to calculate the final forest fire risk prediction value: ; In the formula is a fully connected neural network, for function, is the output of LSTM at the last moment, is the final forest fire risk prediction value.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention first obtains the feature representation of data based on data categories, uses a multi-head attention mechanism to capture the correlation between different types of data, dynamically selects related data for aggregation, and generates richer and more representative feature embeddings. Then, the correlation weights between the forest area and its adjacent areas are learned through a graph convolutional network to aggregate the feature information of the adjacent areas, thereby introducing the information of the spatial neighborhood into the feature representation. Subsequently, the data sequence features are converted into time series features, and the unified time series feature representation is processed using LSTM to achieve accurate modeling of long time series data, thereby further enhancing the prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A step-by-step diagram of a forest fire risk prediction method.

[0020] Figure 2 This is a diagram of the data feature information interaction module.

[0021] Figure 3 This is a diagram of the interaction module of some spatial neighborhood features.

[0022] Figure 4 Convert data series information into time series information graph.

[0023] Figure 5 This is a diagram showing the effect of predicting the possibility of forest fires. DETAILED DESCRIPTION

[0024] The present invention proposes a forest fire risk prediction method. The method fully utilizes the potential correlation between different data sources by constructing multi-source data sequence features of the same location. Then, a graph convolutional network is used to capture and aggregate feature information of spatial neighborhoods, so that the model can identify and learn the correlation and dependency between adjacent areas. Finally, the data sequence features are converted into time series features. Combined with the capture and processing of time series information, the method comprehensively considers the three dimensions of data, neighborhood and time, thereby obtaining a richer and more comprehensive feature representation, and realizing accurate prediction of forest fire risks. The technical solution in the embodiment of the present invention is described in detail and completely below, specifically including the following steps: Figure 1 shown.

[0025] S1. Collect relevant monitoring data for each forest area for 3 years, including temperature, precipitation, wind speed, wind direction, relative humidity, vegetation type, accumulation of combustible materials on the ground, terrain orientation, and smoke content. Clean the data, including filling missing values, removing irrelevant data, and scaling some data to improve data quality. Construct a forest fire risk assessment dataset, and finally divide the dataset into a training set and a validation set in a ratio of 7 to 3. The first 14 days of data for each day are used as historical data for method learning.

[0026] Furthermore, the relevant monitoring data of forest areas are ,in is the number of data types, is the length of the time series. kind of data, in The sequence of monitoring values ​​obtained during the day is ,in for The monitoring value obtained at any time will be Obtain the corresponding eigenvector by looking up the table function: ; In the formula is a lookup table function, for The new feature vector obtained by looking up the table is After all the data is mapped through the lookup table, they are spliced ​​in chronological order. The specific process is as follows: ; In the formula For the The data in The unified characteristic representation of the monitoring value sequence obtained within a day, symbol For splicing operation.

[0027] S3. Build a data feature information interaction module, and use the multi-head attention mechanism to make the feature vector interact with the feature vectors of other types of data, so as to capture the correlation between different types of data and generate richer and more representative feature embeddings.

[0028] Furthermore, if Figure 2 As shown, by using The multi-head attention mechanism captures the correlation between different types of data and inputs the feature vector matrix Generate the query matrix through three linear transformations respectively , key matrix Sum Matrix , the specific calculation process is as follows: ; ; ; In the formula , and are query matrix, key matrix and value matrix respectively, , and are trainable weight matrices for generating queries, keys, and values, respectively. Subsequently, the dot product between the query and the key is calculated to obtain the attention score matrix. The specific calculation process is as follows: ; In the formula For the The attention score matrix of the attention heads, for function, , and For the The query, key, and value of each attention head, is the dimension of the key vector, where is the data dimension obtained by the lookup table function, is the number of attention heads of the multi-head attention mechanism, is a scaling factor used to prevent the gradient from disappearing due to excessive dot product values. For the transpose operation, for each head After calculating the attention separately, the outputs of all heads are concatenated to obtain the output of the multi-head attention mechanism. The specific calculation process is as follows: ; In the formula is the final feature output, where For the The feature representation of class data is obtained through the multi-head attention mechanism. is the weight matrix of the output transformation, For splicing operation.

[0029] S4. Construct a spatial neighborhood feature interaction module, learn the correlation weights between the current forest area and its adjacent areas through a graph convolutional network, and use the correlation weights to aggregate the feature information of the adjacent areas into the feature information of the current forest area, so that the feature representation of the current forest area contains rich spatial neighborhood comprehensive information.

[0030] Furthermore, if Figure 3 As shown in the figure, with forest areas as nodes and whether forest areas are adjacent as edges, an adjacency matrix between forest areas is constructed, in which adjacent forest areas are marked as 1 and non-adjacent areas are marked as 0. The correlation weights between the current forest area and its adjacent areas are learned through the graph convolutional network, and these weights are used to aggregate the feature information of adjacent areas. The specific calculation process is as follows: ; ; In the formula is the feature matrix obtained after the graph convolutional network, is the adjacency matrix between forest regions, where is the number of forest areas divided, is the identity matrix, is the adjacency matrix with self-loops added, for The degree matrix of The sum of each row of elements, is the trainable weight matrix.

[0031] S5. Extract the relevant information at each moment in the data sequence in chronological order and splice the information to construct a complete time series. In this process, the data information at each time step includes all the observation data of a specific forest area at the same time point, thereby ensuring that the time series information can reflect the dynamic changes in the area.

[0032] Furthermore, if Figure 4 As shown, first extract The feature representations of various types of data at the same time are concatenated to obtain the time series feature representation ,in For the Feature representation of the moment.

[0033] S6. Build a time series feature capture module, input the time series data into the LSTM network, and extract the dependencies in the time dimension to enhance the model’s ability to understand and express time dynamic features.

[0034] Furthermore, LSTM is used to capture long-term dependencies and obtain feature representations containing historical information. The specific calculation process is as follows: ; ; ; ; ; ; In the formula is the learnable weight parameter, is the bias parameter, for function, Timestamp The input gate, forget gate and output gate of LSTM. Timestamp The candidate memory cell at Timestamp The memory unit at For timestamp The hidden state of is element-wise multiplication.

[0035] S7. Construct a fully connected neural network and calculate the final feature representation to obtain the final forest fire risk prediction value.

[0036] Furthermore, in step S7, the output of the LSTM at the last moment in step S6 is taken as the final feature vector representation, and the feature vector is passed through a fully connected neural network to calculate the final forest fire risk prediction value: ; In the formula is a fully connected neural network, for function, is the output of LSTM at the last moment, is the final forest fire risk prediction value.

[0037] Furthermore, the method uses Python 3.9 language, uses the PyTorch framework in the CUDA 12 environment, and is trained on an NVIDIA RTX 3090 GPU. The batch size during training is 64 and the learning rate is .

[0038] Furthermore, the method predicts the effect as Figure 5 As shown in the figure, the ordinate is the possibility of forest fire (%), the abscissa is time (days), the asterisk indicates an actual forest fire, and the black straight line indicates the predicted possibility of forest fire. It can be seen from the figure that when a real fire occurs, the predicted value obtained by this method is around 80%, which can fit the real data well, proving the effectiveness of the method.

[0039] The above are only preferred embodiments of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A forest fire risk prediction method, characterized in that: The following steps are involved: S1. Collect 9 types of monitoring data in each forest area, including temperature, precipitation, wind speed, wind direction, relative humidity, vegetation type, accumulation of combustible materials on the ground, terrain orientation and smoke content, and clean the data, including filling missing values, removing irrelevant data, and scaling some data to improve data quality and build a forest fire risk assessment dataset; S2, embed each type of data sequence generated by the forest area within a certain time interval into a feature space, and concatenate them into a feature vector of this type of data; S3. Build a data feature information interaction module, model the complex correlation between different data types through a multi-head attention mechanism, so that the feature information of the current data interacts and merges with the features of other data, thereby generating richer and more representative feature information embedding; S4. Construct a spatial neighborhood feature interaction module to learn the correlation weights between the current forest area and other adjacent areas through a graph convolutional network. These weights are used to aggregate the feature information of adjacent areas, thereby enhancing the spatial neighborhood information represented by the current forest area feature, so that it contains richer spatial comprehensive features. S5, extracting data information at each moment in the data sequence in chronological order, and splicing the information to form time series data, in which each time step contains all relevant data information of a specific forest area at the same time point, thereby retaining the data characteristics in the time dimension; S6. Build a time series feature capture module to extract and process complex dependencies in long time series data through LSTM to enhance the model's ability to understand and express time dynamic features; S7. Build a fully connected neural network and further calculate and process the final feature representation to generate the final forest fire risk prediction value.

2. A forest fire risk prediction method according to claim 1, characterized in that: In step S2, the forest area monitoring data input is ,in is the number of data types, is the time series length of the forest area related monitoring data, For the The data in The monitoring value sequence of this data type obtained within a certain length of time, where for The monitoring value obtained at any time will be After all the data in the table are obtained through the lookup table function to obtain the corresponding feature vectors, they are spliced ​​in chronological order. The specific process is as follows: ; ; In the formula is a lookup table function, for The new feature vector obtained by looking up the table, For the Type of data in length time The unified feature representation of the monitoring value sequence obtained in is the data dimension after the lookup table, symbol For splicing operation.

3. A forest fire risk prediction method according to claim 2, characterized in that: In step S3, first, the input feature vector matrix Generate the query matrix through three linear transformations respectively , key matrix Sum Matrix , using the number of attention heads as The multi-head attention mechanism captures the correlation between different types of data. The specific calculation process is as follows: ; ; ; In the formula , and are trainable weight matrices used to generate queries, keys, and values, respectively. , and are query matrix, key matrix and value matrix respectively, is the dimension of the vector. Then, the dot product between the query and the key is calculated to obtain the attention score matrix. The specific calculation process is as follows: ; In the formula For the The attention score matrix of the attention heads, for function, , and For the The query, key, and value of each attention head, is the dimension of the key vector, is the scaling factor, For the transposition operation, the outputs of all heads after calculating the attention are concatenated to obtain the output of the multi-head attention mechanism. The specific calculation process is as follows: ; In the formula is the final feature output, where For the The feature representation of class data is obtained through the multi-head attention mechanism. is the weight matrix of the output transformation after splicing, For splicing operation.

4. A forest fire risk prediction method according to claim 3, characterized in that: In step S4, the graph convolution network is used to learn the correlation weights between the current forest area and its adjacent areas, and these weights are used to aggregate the feature information of the adjacent areas. The specific calculation process is as follows: ; ; In the formula is the feature matrix obtained after the graph convolutional network, is the adjacency matrix between forest regions, where is the number of forest areas divided, is the adjacency matrix with self-loops added, for The degree matrix of The sum of each row of elements, is the identity matrix, is the trainable weight matrix.

5. A forest fire risk prediction method according to claim 4, characterized in that: In step S5, Extract the feature representation of various types of data at the same time, and concatenate these features to generate a time series feature representation ,in For the Feature representation of the moment.

6. A forest fire risk prediction method according to claim 5, characterized in that: In step S6, LSTM is used to capture long-term dependencies and obtain feature representations containing historical information. The specific calculation process is as follows: ; ; ; ; ; ; In the formula is the learnable weight parameter, is the bias parameter, for function, Timestamp The input gate, forget gate and output gate at and Timestamp The candidate memory cells and memory cells at Timestamp The hidden state of is element-wise multiplication.

7. A forest fire risk prediction method according to claim 6, characterized in that: Step S7, taking the output of the LSTM at the last moment in step S6 As the final feature vector representation, the feature vector is passed through a fully connected neural network to calculate the final forest fire risk prediction value: ; In the formula is a fully connected neural network, for function, is the output of LSTM at the last moment, is the final forest fire risk prediction value.

Images