Improved Swarm Intelligence Optimization And Machine Learning Algorithm Applied In Landslide High Precision Displacement Prediction

Landslides have become a major geological disaster that seriously endangers people's lives and property in our country,and loess landslides are the most typical and serious.Due to the special geological structure and susceptibility to external environmental influences,the damage caused by the easy deformation and instability of the loess landslide is extremely serious.Therefore,carrying out research on highprecision displacement prediction of loess landslides has important theoretical and practical significance for landslide disaster prevention and mitigation.This paper uses the typical loess landslide area in Miaodian,Shaanxi Province as the experimental area.Based on the high-precision displacement time series information obtained by the landslide monitoring network in the demonstration area,the development of the loess landslide body deformation characteristics and external influence factors(irrigation,rainfall,etc.).Two types of improved swarm intelligence optimization and fusion machine learning algorithms were developed,and they were successfully applied to the high-precision prediction of the displacement of the landslide body in the demonstration area.The research results have important theoretical reference value for landslide disaster prevention and early warning research.The main results and innovations obtained are as follows:(1)An improved swarm intelligence optimization algorithm--Improved Crow Algorithm(ICSA)is proposed.In view of the outstanding shortcomings of the traditional crow algorithm(CSA)that the parameter optimization process is slow and easy to fall into local optimum,which will lead to the weakening of the generalization ability of the predictive model,it is proposed to use crow individuals to perform Levi flight instead of random search.This improvement effectively reduces Compared with the traditional crow algorithm and existing optimization algorithms,it is easier to obtain the global optimal solution of the parameters,and it can greatly improve the efficiency of parameter optimization.(2)Two types of current popular models for landslide prediction are selected:Support Vector Machine(SVM)and Extreme Learning Machine(ELM).For the above two types of models,when dealing with small,high-dimensional,and non-linear sample data for landslide displacement prediction,there is a problem with the prediction for the shortcomings of insufficient risk control,an SVM model based on parameter interval optimization and a regularized extreme learning model(RELM)are proposed to improve the ability of the machine learning model to deal with risks,and the improved crow algorithm ICSA algorithm is further used to optimize the model parameters of SVM and RELM,two types of landslide prediction high-efficiency combined models based on swarm intelligence optimization and machine learning algorithms are constructed: ICSA-SVM and ICSA-RELM.(3)Taking full account of the main external factors(irrigation,rainfall,temperature,etc.)of the loess landslide in the demonstration area,based on the highprecision displacement time series information of the landslide,the two types of landslide prediction high-efficiency combined models ICSA-SVM and ICSA-RELM were successfully applied to the landslide prediction.In displacement prediction,the research results show that the two types of landslide prediction high-efficiency combined models are superior to SVM and RELM optimized by particle swarm optimization(PSO),genetic algorithm(GA)and standard crow algorithm(CSA)in terms of fitting accuracy and prediction accuracy.The improved intelligent combination algorithm proposed in this paper can meet the actual needs of landslide disaster prevention and early warning with high displacement prediction accuracy,and has very important reference value for other types of landslide disaster prevention and early warning research.