Research On Landslide Displacement Prediction Based On Improved Harris Hawk Optimization Algorithm And Machine Learning

Landslides have become one of the main geological disasters in China,causing serious harm to the safety of people’s lives and property as well as the sustainable and healthy development of the regional economy and society.For the purpose of preventing and warning of landslide disasters,the high-precision prediction of landslide displacement based on landslide monitoring information has significant reference value.The research area for this paper is the Heifangtai Dangchuan landslide,a typical loess landslide location in China.A combined prediction model for landslide displacement using IHHO-IBP,which combines the improved Harris Hawk Optimization algorithm and genetic programming,was created using high-precision Beidou/GNSS monitoring data from multiple periods in the region and after in-depth analysis of the relationship between landslide deformation characteristics and induced factor response.This was done to address the issues with traditional BP neural networks being prone to local optima and difficult to determine initial weights and thresholds reasonably.Further,an IHHO-IFLN displacement combination prediction model was developed to solve the difficulties of significant initial parameter effect and poor network stability in the traditional Fast Learning Network.In the case study of the Dangchuan landslide in Heifangtai,the two types of combined prediction models produced positive prediction findings,offering technical assistance and a theoretical benchmark for landslide displacement prediction research.The following are the paper’s primary contributions to innovation and accomplishments:(1)The activation function of BP neural network is improved using genetic programming technique to address the issue that traditional BP neural network is easily susceptible to local optimization,improving the model’s suitability for solving realworld issues(IBP).In addition,the improved Harris Hawk Optimization algorithm(IHHO)was used to optimize the initial input weights and hidden layer thresholds of the IBP model,and an IHHO-IBP landslide displacement combination prediction model was created in response to the challenge of determining the initial input weights and hidden layer thresholds of the BP neural network.The IHHO-IBP model has a high level of prediction accuracy and generalizability,according to real measurements of landslide displacement prediction based on Heifangtai.(2)The hidden layer activation function of the Fast Learning Network is improved using the genetic programming,which increases the model’s stability in light of the persistent issues with randomness and instability(IFLN).Additionally,an IHHOIFLN combined prediction model for landslide displacement was created in order to address the problem of uncertainty in the random generation of input weights and hidden layer thresholds in fast learning networks.The improved Harris Hawk algorithm(IHHO)was used in place of the random algorithm to obtain input weights and hidden layer thresholds for IFLNs.In the testing of the Heifangtai landslide displacement prediction example,the combined prediction model produced good prediction results.(3)The primary influencing elements on landslide deformation were chosen based on the driving mechanism and induced environmental factors of typical loess landslides in the Heifangtai area of China.The IHHO-IBP and IHHO-IFLN models developed in this research were used to forecast and assess the displacement of typical landslide bodies based on Beidou/GNSS monitoring data.In comparison to traditional and single-enhanced models,the findings show that the two types of combined models have increased prediction performance.Comparing the cumulative displacement prediction RMSE of the IHHO-IBP and IHHO-IFLN models at monitoring points HF08,HF05,and HF09 reveals that the IHHO-IBP combined displacement prediction model has a higher degree of displacement prediction accuracy and has promising prospects for use and advancement in landslide displacement prediction research.