Multi-Step-Ahead Time Series Prediction Method with Stacking LSTM Neural Network
Xiaofeng Wang, Ying Zhang
Abstract
The issue of multi-step-ahead time series prediction is a daunting challenge of predictive modeling. In this work, we propose a multi-output iterative prediction model with stacking LSTM neural network (MO-LSTMs). In the proposed model, we utilize a stacking LSTM network that consists of multiple hidden layers to learn the features of time series data, and use the dropout algorithm to improve the generalization ability and robustness of the deep learning method. In our stacking LSTM neural network, each hidden layer contains different neural units and the memory state of the cells in each layer are reset. The proposed method solves the problem that the single LSTM network structure is difficult to maintain the time-sequence characteristics between samples in the training process. Additional, in the prediction stage, we utilize the strategy of multi-output iterative prediction to reduce the errors accumulation and errors propagation for long-term time series prediction. It also reduces the computational complexity of the iterative strategy. The simulation experiments are conducted on actual engineering datasets, and the results show that the proposed method is provided with better prediction performance.