基于Xgboost优化的KELM滑坡预报模型研究

Research on the Kelm Lirkeling Forecast Model Based on Xgboost

摘要:

针对极限学习机对滑坡预测准确性低及在训练过程中模型不稳定的问题，引入RBF高斯核函数并使用极限梯度提升树算法Xgboost对KELM进行优化，建立了Xgboost优化后的Xgboost-KELM预测模型；首先采用高斯核RBF作为极限学习机的核函数，解决隐藏节点随机映射问题，增加模型稳定性及适用性；其次将清洗后的监测数据作为模型输入，并使用Xgboost寻优算法对核函数中的超参数进行优化，通过4组测试集进行Xgboost-KELM建模，依据均方误差迭代曲线得出最佳超参数；最后使用两组10%样本集验证模型评价指标及稳定性，实验结果AUC均值对比模型至少提高3个百分点，Precision、Accuracy及Recall至少高于对比模型1.7个百分点，同时Xgboost-KELM模型的方差及偏差都较小，证明该模型稳定性较好，实验结果说明Xgboost-KELM模型具有较好的预测效果，在滑坡灾害预测中有较好的预测能力。

Abstract:

To solve the problems of low accuracy of extreme learning machine (ELM) in landslide prediction, and the instability of the model in the training process, RBF Gaussian kernel function is introduced and Xgboost algorithm is used to optimize KELM, and Xgboost KELM prediction model after Xgboost optimization is established; Firstly, the Gaussian kernel RBF is used as the kernel function of the limit learning machine to solve the problem of random mapping of hidden nodes and increase the stability and applicability of the model; Secondly, the cleaned monitoring data is used as the model input, and Xgboost optimization algorithm is used to optimize the super parameters in the kernel function. Xgboost KELM modeling is conducted through four groups of test sets, and the best super parameters are obtained according to the mean square error iteration curve; Finally, two groups of 10% sample sets were used to verify the model evaluation indicators and stability. The experimental results showed that the AUC mean increased by 3 percentage points compared with that before optimization, and the Precision, Accuracy and Recall were at least 1.7 percentage points higher than that of the comparison model. At the same time, the variance and deviation of Xgboost KELM model were small, which proved that the model was stable. The experimental results showed that Xgboost KELM model had a good prediction effect, It has good prediction ability in landslide disaster prediction.