降雨型滑坡灾害的约简和逻辑回归预测模型

Reduction and logistic regression model for prediction of rainfall landslides disaster

为实现秦巴山区域降雨诱发的滑坡灾害预警,收集处理不同时段滑坡灾害发生点最近区域站大量降雨数据,形成降雨特征滑坡灾害样本数据集;定义降雨型滑坡灾害知识表达系统,利用知识约简和预测结果反馈方法,选取影响滑坡灾害的降雨特征,进而确定发生滑坡灾害的诱发因子;改进逻辑回归方法,提高模型泛化能力,建立约简逻辑回归预测模型,并与不确定朴素贝叶斯模型和遗传神经网络模型对比。结果表明：约简逻辑回归预测模型能有效处理海量高维降雨数据,且预测准确率较高。

In order to realize the rainfall-induced landslide disaster warning in Qinba mountain area,data on rainfall at different time points of landslide disasters from the nearest regional stations were collected and processed. The data set on rainfall characteristic landslide disaster was formed. A rainfall landslide knowledge expression system was defined. Knowledge reduction and prediction results feedback were used to filter feature attributes of rainfall. And then inducing factors of landslide disasters were determined. The logistic regression method was improved,the model generalization ability was enhanced and a reduced logistic regression model was established and compared with the uncertain naive Bayes model and the genetic neural network model. Experimental result shows that the reduced logistic regression model can effectively deal with the high-dimensional rainfall data and has a higher prediction precision accuracy.