华南台风灾害的风险熵-极限学习机预测模型研究

结合广义模糊熵原理和模糊c均值聚类方法构建华南台风灾害风险熵模型,对华南台风灾害进行风险分析,讨论其分布情况;利用灰色关联分析法,探讨华南台风灾害的灾情因子、致灾源因子分别和灾害风险熵之间的关系以及二者对风险熵的影响程度;建立基于极限学习机的非线性回归模型,以多元线性回归和BP神经网络两种方法作为对照组,进一步探讨风险熵与灾情因子和致灾源因子关系。结果表明,华南台风灾害风险熵值呈正态分布,与灾情因子和致灾源因子的灰色关联度分别为0.7162和0.7949,受灾情因子和致灾源因子的影响较大;利用构建的极限学习机模型预测的华南台风灾害风险熵值平均绝对误差为0.059,拟合优度为92.82%,将预测结果与常规的多元线性回归和BP神经网络方法的预测结果进行对比分析,结果表明,用构建的极限学习机模型预测华南台风灾害风险熵值,其性能比常规多元线性回归和BP神经网络方法有明显的改进。

Research on Extreme Learning Machine-Risk Entropy Prediction Model of Typhoon Disaster in South China

Combined with the generalized fuzzy entropy principle and fuzzy c-means clustering method, the risk entropy model of typhoon disaster in South China is constructed. The risk analysis of typhoon disaster in South China is carried out and its distribution is discussed. The relationship between disaster condition factor, disaster source factor and disaster risk entropy of typhoon disaster in South China and the degree of their influence on risk entropy are discussed by means of grey relational analysis. A nonlinear regression model based on extreme learning machine was established. The relationship between risk entropy and disaster condition factor and disaster source factor was further discussed by using multiple linear regression and BP neural network as the compared group. The results show that the entropy value of typhoon disaster risk in South China is normal distribution, which is related to the disaster situation. The grey relation degree of the factor and the disaster-causing source factor is 0.7162 and 0.7949, respectively, and the influence of the disaster factor and the disaster-causing source factor is large;the average absolute error of the risk entropy value of the South China typhoon disaster, which is predicted by the built limit learning machine model, is 0.059, and the fitting goodness is 92.82%.The results of this paper are compared with the predicted results of the conventional multiple linear regression and the BP neural network method. The results show that the risk entropy of the South China typhoon is predicted by the built limit learning machine model, the performance of the method is better than that of the conventional multiple linear regression and the BP neural network.