Predicting expressway subsidence based on niching genetic algorithm and Holt-Winters model

摘要

Mining subsidence prediction based on monitoring data is a vital task in engineering construction over underground mines. To improve the accuracy of predicting mining subsidence, a method based on the niching genetic algorithm (NGA) and the Holt-Winters model is proposed here. The NGA including niche selection methodology is chosen due to the defects of the genetic algorithm. The NGA is applied to optimize the parameters of the Holt-Winters model. The NGA-Holt-Winters model was applied to mining subsidence prediction on two sides of the Siping-Changchun expressway. The results show that the NGA enhances the convergence speed and precision of the algorithm, with both the convergence speed and forecast accuracy being superior to those of the grey model and the support vector machine model. The relative errors of the NGA-Holt-Winters model are less than 2% and the mean error is - 0.18%. The proposed model has better long-term prediction accuracy for mining subsidence than the support vector machine model, showing lower mean errors of between - 0.49 and - 0.79%.