基于松鼠觅食算法优化LSSVM的泥石流预测

Prediction of Debris Flow Based on Squirrel Foraging Algorithm Optimized LSSVM

摘要:

针对山区村镇泥石流影响因素多元复杂、LSSVM算法参数随机导致的精度不佳及陷入局部最优问题，采用核主成分分析KPCA降维、SSA算法参数寻优的方法建立LSSVM泥石流灾害预测模型。以山阳县中村镇泥石流为例，分析泥石流全域地形地貌成灾因子，对数据预处理清洗规范，利用KPCA主成分贡献率选取出6个成灾因子作为LSSVM算法的输入数据，泥石流发生概率为输出，建立泥石流预报模型，并用SSA算法进行模型参数的优化。将SSA寻优后的LSSVM预测结果与GA、GC参数寻优模型预测结果比对，结果表明SSA-LSSVM准确率达到93.2%，相比其他模型提高[4.8%-1.4%]，且SSA算法优化的LSSVM模型的MAE、MSE和RMSE最小且接近于零，同时从泥石流发生的预报等级维度进行结果比对分析，结果进一步说明模型预测的精度及稳健性。本研究说明SSA-LSSVM算法可用于泥石流灾害发生概率的预测，为此类灾害预测提供了科学依据。

Abstract:

In order to solve the problem of poor accuracy and local optimal caused by multiple and complex influencing factors of debris flow in mountainous villages and towns and the random parameters of LSSVM algorithm, the LSSVM debris flow disaster prediction model was established by KPCA dimension reduction and SSA algorithm parameter optimization methods. Mudslides son duong district of villages and towns, for example, global topography by factor analysis of debris flow, wash specification for data preprocessing, 6 by using KPCA principal component contribution rate to select the factors as the input data of LSSVM algorithm, debris flow occurrence probability as output, debris flow forecast model is established, and model parameters are optimized with the SSA algorithm. By comparing the prediction results of LSSVM optimized by SSA with those of GA and GC parameter optimization models, the results show that the accuracy of SSA-LSSVM reaches 93.2%, which is higher than that of other models [4.8%-1.4%]. Moreover, MAE, MSE and RMSE of LSSVM optimized by SSA algorithm are minimum and close to zero. At the same time, the results are compared and analyzed from the prediction grade dimension of debris flow occurrence, and the results further illustrate the accuracy and robustness of the model prediction. This study shows that SSA-LSSVM algorithm can be used to predict the probability of debris flow disasters, and provides a scientific basis for the prediction of such disasters.