基于 C4．5决策树算法的土质边坡稳定性评价研究

摘要

When the soil slope stability is evaluated by neural network model,varieties of training samples always make the evaluation result unsatisfactory.In order to solve the problem,we introduce the C4.5 decision tree algo-rithm,build an evaluation model of soil slope stability based on decision tree classifier,and prune the tree structure established.Furthermore,we adopt measured data in several soil slope projects and select classification attributes ac-cording to gain ratio of information in this model.Compared with BP neural network and LVQ(Learning Vector Quantization)neural network,the result shows that decision tree algorithm has the highest accuracy for classifica-tion,up to 90%,and the computation time of this model is 2.24 seconds.Finally,it is feasible to introduce decision tree algorithm for stability evaluation in soil slope.%采用神经网络进行土质边坡稳定性评价时，差异性较大的训练样本往往会使评价结果不太理想。针对这一问题引入 C4．5决策树算法，采用多个土质边坡工程的实测数据，运用信息增益率进行分类属性的选择，并对建立好的树体结构进行剪枝操作，建立基于决策树的土质边坡稳定性评价模型。将该模型与 BP 神经网络和 LVQ （Learning Vector Quantization，学习向量量化）神经网络进行对比分析，结果显示决策树模型分类正确率最高，达到90％，模型所用时间为2．24 s，表明把决策树用于土质边坡稳定性评价是合理的。