Extreme Learning Machine Application in Flood Forecasting
Sun Mia · Shenyang Nongye Daxue xuebao · 2014
In order to mitigate the flood damage to humans, extreme learning machine(ELM) would be referenced to flood forecasting. Flood forecasting of Daling River Basin and the comparison between this model and traditional neural networks were conducted with the 44 observation precipitation data along Daling River Basin from 1960 to 2010(including 33stations in Daling River Basin, 11stations in Xiaoling River Basin). The prediction model based on ELM predictive value of annual precipitation in Daling River Basin had mean square error(MSE) of 0.003, the coefficient of determination(R2) of 0.927 and Xiaoling River Basin had mean square error(MSE) of 0.0037, the coefficient of determination(R2) of 0.8481. The error precision requirements were met. The results were better than those from BP neural network prediction model mean square error value and the decisive factor. ELM forecasting model for floodforecasting works well for adding a new flood forecasting method.