Research And Implementation Of Landslide Prediction Based On Soil Slope Data Classification Model

As one of the common geological disasters,landslides have serious negative impacts on the safety of people's lives and property and the economic benefits of society.Forecasting landslides is an effective method to reduce the losses caused by landslide disasters,which has become a research hotspot and focus in recent years.Partial slope stability landslide prediction method uses displacement-time curve tangent angle value to predict landslide according to the three-phase spatiotemporal evolution characteristics of landslide disaster,which has the problem of assuming that the monitoring starting point is the constant velocity landslide stage and taking the landslide speed value results in lower prediction accuracy and the prediction algorithm efficiency needs to be improved.Aiming at this problem,this thesis proposes a landslide prediction algorithm based on a landslide data classification model.It does not require hypothetical conditions and has a good performance in solving the accuracy and operation efficiency of the prediction algorithm.This thesis uses data analysis,neural network,intelligent algorithm and other popular methods to process the landslide monitoring data,including data analysis,classification and prediction.Because the parameters that affect landslide disasters are numerous and complex,this thesis uses the gray correlation method to improve the orthogonal test analysis method to quantitatively analyze the correlation between the measurement parameters of each sensor and the slope stability coefficient,which is conducive to selective key placement of sensors,cost control of landslide monitoring and early warning and subsequent data classification and landslide prediction.The purpose of the landslide data classification model is to divide historical landslide monitoring data into safety data and dangerous data based on the time of landslide disaster occurrence.After comparing and experimenting with common classification algorithms,the Support Vector Machine classification algorithm with RBF function as the kernel function was selected among the naive Bayes algorithm,support vector machine algorithm,and decision tree algorithm.On this basis,the particle swarm optimization algorithm is used to improve the SVM classification algorithm to form the PSO-SVM model to classify the landslide data,and the data set whose classification result is dangerous data will continue the subsequent landslide prediction.Landslide prediction includes landslide displacement prediction and landslide time prediction.The Long Short Term Memory Network that is suitable for time series data prediction is adopted to predict the displacement of landslide data.Since there is a certain error between the prediction result and the actual monitoring displacement value,this thesis selects the Autoregressive Integrated Moving Average Model to correct the prediction error of LSTM.The combined LSTM-ARIMA algorithm predicts the landslide displacement based on the landslide monitoring data.The improved Verhulst inverse function model is used to forecast the time of landslide disaster,which is nonlinearly processed based on displacement monitoring data and prediction data to obtain the parameters for solving the time of landslide disaster occurrence,and the landslide time prediction is made according to the results of the solution.Using the parameter data obtained by the slopes actually built in the laboratory to verify the established model,the result is that the correlation analysis results of the landslide data in the thesis are consistent with the actual situation,and the accuracy rate of the established landslide data classification model has reached 95.89%.As for the landslide prediction model,the average root mean square error of the prediction results of landslide displacement is within2 mm,and the final predicted landslide time differed from the actual recorded landslide time by only five minutes.The results show that the landslide data classification model and landslide prediction algorithm built in the thesis are of great practicality,precision and efficiency.