A Flood Forecasting Model Based on Deep Learning Algorithm via Integrating Stacked Autoencoders with BP Neural Network

Abstract

Artificial neural network (ANN) has been widely applied in flood forecasting and got good results. However, it can still not go beyond one or two hidden layers for the problematic non-convex optimization. This paper proposes a deep learning approach by integrating stacked autoencoders (SAE) and back propagation neural networks (BPNN) for the prediction of stream flow, which simultaneously takes advantages of the powerful feature representation capability of SAE and superior predicting capacity of BPNN. To further improve the non-linearity simulation capability, we first classify all the data into several categories by the K-means clustering. Then, multiple SAE-BP modules are adopted to simulate their corresponding categories of data. The proposed approach is respectively compared with the support-vector-machine (SVM) model, the BP neural network model, the RBF neural network model and extreme learning machine (ELM) model. The experimental results show that the SAE-BP integrated algorithm performs much better than other benchmarks.