基于组合预测模型的滑坡变形趋势研究OACSTPCD

To achieve high-precision prediction of landslide deformation,based on the monitoring results of landslide deformation,a combined prediction model for landslide deformation is constructed using support vector regression,RReliefF algorithm,whale optimization algorithm,and Arima model.Then,the multi-scale characteristics of landslide deformation are studied using multifractal elimination trend fluctuation analysis and Spearman rank test.The results show that,through continuous combination optimization processing,the accuracy of landslide deformation prediction can be significantly improved.Moreover,through extrapolation prediction,the existing rate of monitoring points D1-D3 is relatively larger,and the decrease in prediction rate compared to the existing rate is 21.61%,12.95%,and 6.54%,respectively.The prediction rate of monitoring point D4 is relatively larger,with an increase of 19.85%compared to the existing rate,indicating that the stability of the landslide front edge is further becoming unfavorable.The four monitoring points have different fractal spectral widths and proportions of large waveforms,but D4 monitoring point is the most unfavorable in terms of fractal spectral widths and proportions of large waveforms.The fractal spectrum width of the deformation at the four monitoring points shows an increasing trend,and only the proportion of large fluctuations at the D2 monitoring point is an increasing feature,and the other monitoring points show a decreasing trend,indicating that the fluctuation of landslide deformation will be more severe.However,the overall trend of large fluctuations tends to weaken and only increases locally.It is recommended to carry out landslide prevention and control measures as soon as possible.