Geological disaster alarm method based on ensemble learning algorithm

Abstract

The invention discloses a geological disaster alarm method based on an ensemble learning algorithm, and aims to solve the problems that when an existing disaster alarm method faces complex and changeable geological environments and multi-source monitoring data, the situation of false alarm or missing alarm is likely to occur, and various factors influencing geological disaster occurrence cannot be comprehensively considered. Comprising the following steps: S1, data acquisition and preprocessing; s2, selecting base learners and training each type of base learners; s3, constructing an integrated learning model; s4, model training and optimization; and S5, geological disaster alarm is carried out. According to the method, the geological monitoring data are comprehensively analyzed by integrating a plurality of learning models, so that the accuracy and reliability of geological disaster alarm can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster monitoring and early warning, and specifically to a method for analyzing and processing geological monitoring data by utilizing an integrated learning algorithm to realize efficient and accurate geological disaster warning. Background Art

[0002] Geological disasters refer to geological actions or phenomena that are caused by natural or human factors and cause losses to human life and property and damage to the environment. The distribution and change patterns of geological disasters in time and space are subject to both the natural environment and human activities, and are often the result of the interaction between humans and nature. Geological disasters include collapse, landslides, mudslides, ground fissures, ground subsidence, ground collapse, rock bursts, tunnel water inrush, mud inrush, gas inrush, spontaneous combustion of coal seams, loess collapse, rock and soil expansion, sand liquefaction, land freezing and thawing, soil erosion, land desertification and swamping, soil salinization, as well as earthquakes, volcanoes, geothermal damage, etc. In order to avoid greater losses to human life and property caused by geological disasters, it is necessary to predict the trend of geological disasters, so as to facilitate the early deployment of geological disaster prevention and mitigation work, formulate prevention and control plans, and take precautions in advance.

[0003] The current geological disaster alarm methods have the following main shortcomings: the traditional threshold judgment-based method is too simple, and is prone to false alarms or missed alarms when faced with complex and changeable geological environments and multi-source monitoring data; some simple data processing methods are difficult to fully explore the inherent relationships in monitoring data and cannot comprehensively consider the various factors that affect the occurrence of geological disasters. Summary of the invention

[0004] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a geological disaster alarm method based on an integrated learning algorithm, which improves the accuracy and reliability of geological disaster alarms by integrating multiple learning models to conduct comprehensive analysis of geological monitoring data.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions: A geological disaster alarm method based on an integrated learning algorithm comprises the following steps: S1: Data collection and preprocessing; Set up various types of monitoring equipment in areas with potential geological hazards to collect various types of geological monitoring data in real time; and pre-process the collected data; S2: Select base learners and train each base learner; Decision tree, support vector machine and naive Bayes are selected as base learners; S3: Build an integrated learning model; The random forest ensemble learning algorithm is used to build an ensemble model; Random forests are composed of multiple decision trees. When constructing each decision tree, samples are randomly extracted from the original training data set with replacement, and some features are randomly selected for node splitting. The final prediction result is obtained by voting or averaging the results of multiple decision trees. S4: Model training and optimization; The preprocessed geological monitoring data is divided into training set, validation set and test set; the training set is used to train the ensemble learning model, and during the training process, the parameters of the model are adjusted and optimized through the validation set; the test set is used to evaluate the final performance of the model after the model training is completed; since the data of the test set is not used in the training process, it can provide an unbiased estimate of the generalization ability of the model; S5: Geological disaster alarm; The geological monitoring data collected and preprocessed in real time in step 1 is input into the trained integrated learning model; the model makes predictions based on the input data to determine whether there is a disaster risk in the current geological conditions; when the model prediction results show that the geological disaster risk exceeds the preset risk threshold, the alarm mechanism is triggered, a geological disaster alarm signal is issued, and the relevant monitoring data and prediction results are transmitted to the monitoring center.

[0006] Furthermore, the specific content of each base learner training phase in step S2 is as follows: Decision Tree: Feature selection: Select the best split feature, using information gain and Gini impurity as the criteria; Tree construction: recursively split the data until the stopping condition is met; Pruning: To prevent overfitting, prune the tree and remove branches that do not contribute much to the overall performance; Parameter tuning: adjust parameters such as tree depth and minimum number of samples required for splitting to optimize model performance; Support Vector Machines: Kernel function selection: Select an appropriate kernel function to process nonlinearly separable data; Hyperplane determination: Find an optimal hyperplane that maximizes the interval between different categories; Regularization parameters: Adjust the regularization parameters to control the penalty for misclassification and prevent overfitting; Kernel parameters: For nonlinear kernels, kernel parameters need to be adjusted; Model training: Use optimization algorithms to solve the optimization problem of SVM; Naive Bayes: Probabilistic model: Based on the assumption that features are independent of each other, the conditional probability of each feature under different categories is calculated; Parameter estimation: Use maximum likelihood estimation or Bayesian estimation to estimate parameters in probability models; Feature selection: Although naive Bayes usually does not perform feature selection, in practical applications, it is possible to select the features that have the greatest impact on classification; Model training: Calculate prior probabilities and conditional probabilities based on the training data set.

[0007] Furthermore, the specific content of step S3 is as follows: S3.1: Decision tree model construction, the formula is as follows: y = T (x) = ∑(n=1 to N) y n I(x ∈ R n ); Where: y is the predicted geological disaster risk; x is the input monitoring data feature vector; T (x) is the decision tree model; N is the number of leaf nodes in the decision tree; y n is the predicted value of the nth leaf node; R n is the feature space region corresponding to the nth leaf node, that is, the set of input data that satisfies a specific decision rule; I(x∈R n ) is the indicator function. If the input data x belongs to region R n , then it is 1, otherwise it is 0; The decision tree is represented as: y = T(x) = 1 · I(x_1>100) + 1 · I(x_1 ≤ 100 and x_2>50) + 0 · I(x_1 ≤ 100 and x_2 ≤ 50); S3.2: Support vector machine model construction to predict landslide occurrence; A classification model is constructed using a support vector machine. The goal of the support vector machine is to find a hyperplane that separates the positive and negative classes as much as possible; the positive class indicates that a landslide has occurred, y = +1; the negative class indicates that no landslide has occurred, y = -1; The specific steps to build a support vector machine model are as follows: S3.2.1: Define hyperplane; The hyperplane of the support vector machine is represented as: ; Where: w is the normal vector of the hyperplane, x is the feature vector of the sample point, including rainfall , soil moisture ,slope and geological types ; b is the bias term; S3.2.2: Optimization problem; The optimization problem of support vector machine is expressed as: ; Subject to the following constraints: ; in: is the target variable of the i-th sample (+1 or -1); is the feature vector of the i-th sample; S3.2.3: Kernel techniques; Due to the geological type is a categorical variable, and kernel techniques are used to handle nonlinear relationships. The radial basis function kernel is selected as the kernel function. The radial basis function kernel can map data to a high-dimensional space, making the data that is originally nonlinearly inseparable in the low-dimensional space linearly separable in the high-dimensional space. Radial basis function kernel: ; in, γ is the parameter of the kernel function, which controls the complexity and width of the decision boundary and is adjusted according to experimental data; S3.2.4: Lagrange multiplier method; In order to solve the above optimization problem, the Lagrange multiplier method is used; the Lagrange function is defined as: ; in is the Lagrange multiplier; S3.2.5: Dual problem; By solving the dual problem of the Lagrangian function, we can obtain: ; Subject to the following constraints: , ; S3.2.6. Solving and predicting; After solving the above dual problem, we can get the Lagrange multiplier for each sample: ; For a new sample point x, its predicted value can be calculated by the following formula: ; if , then the prediction is positive; otherwise, the prediction is negative; S3.3: Construct a naive Bayes model and obtain the target variable; In the landslide prediction problem, the basic formula for constructing the naive Bayes model is as follows: 1. Bayes’ Theorem: The core of the naive Bayes model is Bayes' theorem, which describes the probability of a category under given features; for the target variable y and the feature vector , Bayes’ theorem is expressed as: ; in, is the conditional probability of the feature vector under category y, is the prior probability of class y, is the marginal probability of the eigenvector; 2. Naive Bayes' independence assumption: The naive Bayes model assumes that the features are independent of each other, so the joint probability can be Expressed as the product of the conditional probabilities of each feature: ; 3. Calculation of posterior probability: Substituting the independence hypothesis into Bayes' theorem, we get the formula for calculating the posterior probability: ; This formula shows that the value of the target variable Y can be predicted by comparing the posterior probabilities of different categories y; 4. Classification decision: The final classification decision function It is expressed as: ; This means that the choice makes the posterior probability The largest category As a prediction result; S3.4: The ensemble learning algorithm uses the Stacking model (in ensemble learning, decision tree models, support vector machine models, and naive Bayes models are used as base learners to build a more complex ensemble model. The Stacking model is a method of ensemble learning that improves the overall prediction performance by combining the prediction results of multiple models). The construction of this model involves base classifiers and meta classifiers; the prediction results of the base classifiers are used as input features of the meta classifier. The formula is: Meta-features=clf1(x),clf2(x),…,clfn(x); Final prediction=meta_clf(Meta-features); Among them: clf1, clf2, …, clfn are base classifiers; meta_clf is the meta classifier.

[0008] The beneficial effects of the present invention are: the method of the present invention can effectively improve the accuracy and reliability of geological disaster alarms by integrating multiple learning models to conduct comprehensive analysis of geological monitoring data. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, 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 work.

[0010] Figure 1 The present invention is a flow chart of a geological disaster alarm method based on an integrated learning algorithm. DETAILED DESCRIPTION

[0011] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0012] like Figure 1 As shown, a geological disaster alarm method based on an integrated learning algorithm includes the following steps: Step 1: Data collection and preprocessing; Set up earthquake monitors, mountain displacement sensors and groundwater level monitors in the corresponding monitoring areas. Collect data every 10 minutes during the set monitoring period (for example, within a month). For example, if the collected seismic wave amplitude data has some abnormally large values, these abnormal values ​​can be removed through data cleaning methods (such as the 3σ principle). Then, use the Min-Max normalization method to normalize all monitoring data to the [0,1] interval.

[0013] Step 2: Model construction and training; Decision trees and SVMs were selected as base learners, and random forests were used to build an ensemble learning model. 70% of the data was used as a training set, 20% of the data was used as a validation set, and 10% of the data was used as a test set. When training the random forest model, the number of trees was set to 100, the maximum depth of the tree was set to 10, the model parameters were adjusted through the validation set, and the accuracy and loss function value of the model on the validation set were observed.

[0014] Step 3: Geological disaster alarm; At a certain moment, the real-time collected and pre-processed seismic wave amplitude, mountain displacement and groundwater level data are input into the trained random forest model. The model predicts that the probability of landslide risk at this moment is 0.7 (assuming the risk threshold is 0.6), then the alarm mechanism is triggered and a landslide alarm signal is sent to the monitoring center in the mountain area, with the current monitoring data and predicted risk probability attached, so that relevant personnel can take countermeasures.

[0015] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A geological disaster alarm method based on an integrated learning algorithm, characterized in that: The following steps are involved: S1: Data collection and preprocessing; Set up various types of monitoring equipment in areas with potential geological hazards to collect various types of geological monitoring data in real time; and pre-process the collected data; S2: Select base learners and train each base learner; Decision tree, support vector machine and naive Bayes are selected as base learners; S3: Build an integrated learning model; The random forest ensemble learning algorithm is used to build an ensemble model; Random forests are composed of multiple decision trees. When constructing each decision tree, samples are randomly extracted from the original training data set with replacement, and some features are randomly selected for node splitting. The final prediction result is obtained by voting or averaging the results of multiple decision trees. S4: Model training and optimization; The preprocessed geological monitoring data is divided into training set, validation set and test set; the training set is used to train the ensemble learning model, and during the training process, the parameters of the model are adjusted and optimized through the validation set; the test set is used to evaluate the final performance of the model after the model training is completed; since the data of the test set is not used in the training process, it can provide an unbiased estimate of the generalization ability of the model; S5: Geological disaster alarm; Input the geological monitoring data collected and preprocessed in real time in step 1 into the trained integrated learning model; The model makes predictions based on the input data to determine whether there is a disaster risk in the current geological conditions; when the model prediction results show that the risk of geological disasters exceeds the preset risk threshold, the alarm mechanism is triggered, a geological disaster alarm signal is issued, and the relevant monitoring data and prediction results are transmitted to the monitoring center.

2. A geological disaster alarm method based on an integrated learning algorithm according to claim 1, characterized in that: The specific content of each base learner training phase in step S2 is as follows: Decision Tree: Feature selection: Select the best split feature, using information gain and Gini impurity as the criteria; Tree construction: recursively split the data until the stopping condition is met; Pruning: To prevent overfitting, prune the tree and remove branches that do not contribute much to the overall performance; Parameter tuning: adjust parameters such as tree depth and minimum number of samples required for splitting to optimize model performance; Support Vector Machines: Kernel function selection: Select an appropriate kernel function to process nonlinearly separable data; Hyperplane determination: Find an optimal hyperplane that maximizes the interval between different categories; Regularization parameters: Adjust the regularization parameters to control the penalty for misclassification and prevent overfitting; Kernel parameters: For nonlinear kernels, kernel parameters need to be adjusted; Model training: Use optimization algorithms to solve the optimization problem of SVM; Naive Bayes: Probabilistic model: Based on the assumption that features are independent of each other, the conditional probability of each feature under different categories is calculated; Parameter estimation: Use maximum likelihood estimation or Bayesian estimation to estimate parameters in probability models; Feature selection: Although naive Bayes usually does not perform feature selection, in practical applications, it is possible to select the features that have the greatest impact on classification; Model training: Calculate prior probabilities and conditional probabilities based on the training data set.

3. A geological disaster alarm method based on an integrated learning algorithm according to claim 1 or 2, characterized in that: The specific content of step S3 is as follows: S3.1: Decision tree model construction, the formula is as follows: y = T (x) = ∑(n=1 to N) y n ·I(x ∈ R n ); Where: y is the predicted geological disaster risk; x is the input monitoring data feature vector; T (x) is the decision tree model; N is the number of leaf nodes in the decision tree; y n is the predicted value of the nth leaf node; R n is the feature space region corresponding to the nth leaf node, that is, the set of input data that satisfies a specific decision rule; I(x∈R n ) is the indicator function. If the input data x belongs to region R n , then it is 1, otherwise it is 0; The decision tree is represented as: y = T(x) = 1 · I(x_1 > 100) + 1 · I(x_1 ≤ 100 and x_2 > 50) + 0 · I(x_1 ≤ 100 and x_2 ≤ 50); S3.2: Support vector machine model construction to predict landslide occurrence; A classification model is constructed using a support vector machine. The goal of the support vector machine is to find a hyperplane that separates the positive and negative classes as much as possible; the positive class indicates that a landslide has occurred, y = +1; the negative class indicates that no landslide has occurred, y = -1; The specific steps to build a support vector machine model are as follows: S3.2.1: Define hyperplane; The hyperplane of the support vector machine is represented as: ; Where: w is the normal vector of the hyperplane, x is the feature vector of the sample point, including rainfall , soil moisture ,slope and geological types ; b is the bias term; S3.2.2: Optimization problem; The optimization problem of support vector machine is expressed as: ; Subject to the following constraints: ; in: is the target variable of the i-th sample (+1 or -1); is the feature vector of the i-th sample; S3.2.3: Kernel techniques; Due to the geological type is a categorical variable, and kernel techniques are used to handle nonlinear relationships. The radial basis function kernel is selected as the kernel function. The radial basis function kernel can map data to a high-dimensional space, making the data that is originally nonlinearly inseparable in the low-dimensional space linearly separable in the high-dimensional space. Radial basis function kernel: ; in, γ is the parameter of the kernel function, which controls the complexity and width of the decision boundary and is adjusted according to experimental data; S3.2.4: Lagrange multiplier method; In order to solve the above optimization problem, the Lagrange multiplier method is used; the Lagrange function is defined as: ； in is the Lagrange multiplier; S3.2.5: Dual problem; By solving the dual problem of the Lagrangian function, we can obtain: ； Subject to the following constraints: , ; S3.2.

6. Solving and predicting; After solving the above dual problem, we can get the Lagrange multiplier of each sample: ; For a new sample point x, its predicted value can be calculated by the following formula: ； if , then the prediction is positive; otherwise, the prediction is negative; S3.3: Construct a naive Bayes model and obtain the target variable; In the landslide prediction problem, the basic formula for constructing the naive Bayes model is as follows:

1. Bayes’ Theorem: The core of the naive Bayes model is Bayes' theorem, which describes the probability of a category under given features; for the target variable y and the feature vector , Bayes’ theorem is expressed as: ； in, is the conditional probability of the feature vector under category y, is the prior probability of class y, is the marginal probability of the eigenvector; 2. Naive Bayes' independence assumption: The naive Bayes model assumes that the features are independent of each other, so the joint probability can be Expressed as the product of the conditional probabilities of each feature: ； 3. Calculation of posterior probability: Substituting the independence hypothesis into Bayes' theorem, we get the formula for calculating the posterior probability: ； This formula shows that the value of the target variable Y can be predicted by comparing the posterior probabilities of different categories y; 4. Classification decision: The final classification decision function It is expressed as: ； This means that the choice makes the posterior probability The largest category As a prediction result; S3.4: The ensemble learning algorithm uses the Stacking model. The construction of this model involves base classifiers and meta-classifiers. The prediction results of the base classifiers are used as input features of the meta-classifier. The formula is: Meta-features=clf1(x),clf2(x),…,clfn(x); Final prediction=meta_clf(Meta-features); Among them: clf1, clf2, …, clfn are base classifiers; meta_clf is the meta classifier.

Images