基于粗糙集优化支持向量机的泥石流危险度预测模型

为准确预测泥石流危险度,提出了基于粗糙集理论(RS)的粒子群算法(PSO)优化支持向量机(SVM)模型。首先离散化泥石流样本数据形成初始决策表,利用粗糙集理论对10个泥石流危险度影响指标进行属性约简,将约简后的泥石流指标数据归一化处理作为支持向量机的学习样本,通过粒子群算法寻优获得最佳支持向量机模型参数,最终建立基于粗糙集的泥石流危险度预测的优化支持向量机模型。并将构建的RS-PSO-SVM模型用于对测试样本的预测,结果表明：在相同训练样本的条件下,RS-PSO-SVM模型、PSO-SVM模型及RS-PSO-BP模型三者的预测准确率分别为：87.5%,87.5%,75%,说明RS-PSO-SVM模型和PSO-SVM模型具有比RS-PSO-BP模型更高的精度。此外,尽管RS-PSO-SVM模型和PSO-SVM模型具有相同的预测精度,但是由于进行了属性约简,RS-PSO-SVM模型可以有效提高运行效率,降低模型复杂度。

Prediction Model of Debris Flow Danger Degree Based on Support Vector Machine Optimized by Rough Set

In order to predict debris flow danger degree accurately, the support vector machine (SVM) model optimized by particle swarm optimization algorithm (PSO) was proposed based on rough set (RS). First, RS theory was introduced to make up initial decision table by the discretion data of debris flow samples and to make attribute reduction of 10 influence indexes of debris flow danger degree. Then, the normalized reduction data of debris flow indexes was used as the study samples and the optimal SVM parameters was found by PSO. Finally, the debris flow danger degree based on RS-PSO-SVM was established and this model was used to predict the test samples. The results show that for the same training samples, the prediction accuracy of three models of RS-PSO-SVM, PSO-SVM and RS-PSO-BP were 87.5%, 87.5%, 75% respectively, which illustrates that RS-PSO-SVM has higher accuracy than that of RS-PSO-BP model. Moreover, although RS-PSO-SVM model and PSO-SVM model have the same prediction accuracy, RS-PSO-SVM model can improve the operation efficiency and decrease model complexity.