基于PCA-GEP算法的边坡稳定性预测

Slope stability prediction based on PCA-GEP algorithm

提出一种基于主成分分析的基因表达式程序设计算法,并将其用于边坡稳定性预测。该算法先采用主成分分析法对样本数据进行预处理,有效地减少预测模型的输入量,消除输入数据间的相关性,再将得到的新样本数据输入基因表达式,构建边坡稳定性的预测模型。利用该预测模型对82个危险圆弧破坏边坡实例中的71个实例进行学习,对另外11个实例进行预测,取得了较好的效果。在保留传统的以误差值作为评判模型优劣标准的同时,引入AIC信息准则法,分别对v-SVR算法和GA-BP网络算法和PCA-GEP算法三种预测模型进行比较分析,结果表明,运用该算法可以获得更优的预测模型,其预测结果比v-SVR算法和GA-BP网络等其他算法得到的结果具有更高的预测精度。工程实例计算表明,该方法是合理、可行的。

A novel gene expression programming （GEP） algorithm based on principal component analysis （PCA） is proposed and applied to predict slope stability. The PCA technology is utilized to preprocess the sample data and to reduce the input of prediction model, which thus improves the input factors and eliminates the correlation among the inputs. Then new sample data are input into gene expression to construct the prediction model. Applying the prediction model to predict the safety factors of 11 slope samples after learning other 71 samples, satisfactory result is obtained. Reserving the criteria that model judged by error value, Akaike＇s information criterion （AIC） is introduced; v-SVR algorithm, GA-BP network algorithm and PCA-GEP algorithm are compared respectively. The experimental results show the PCA-GEP algorithm is more accurate than v-SVR algorithm and GA-BP network algorithm. The engineering example indicates that the method is reasonable and feasible; and it provides a new idea for slope assessment. It can acquire slope safety factors quickly and accurately and evaluate the slope stability. It provides the decision basis for selecting economical slope design scheme.