A Hybrid SARIMA‐LSTM Model for Air Temperature Forecasting

There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

Abstract

In order to improve the prediction accuracy of air temperature forecasting, a temperature prediction model based on the hybrid SARIMA (seasonal autoregressive integrated moving average)‐LSTM (long short‐term memory) model is constructed. First, this method decomposes the temperature series into three series of trend, seasonal, and residual through seasonal‐trend decomposition procedure based on Loess decomposition method. It establishes SARIMA to predict the trend and seasonal series and extracts the linear information contained in the time series to the maximum extent. Then, the LSTM model is used to fit the residual series and the hidden nonlinear information is further extracted. Finally, the prediction results of two parts are added in series to obtain the prediction result of the final hybrid model. Three indexes, namely, root mean square error, mean absolute error, and mean absolute percentage error are evaluated to calculate the prediction accuracy about single models including ARIMA, SARIMA, and LSTM and the hybrid models ARIMA‐LSTM and SARIMA‐LSTM. Also the Kupiec index is used to show tail performance. The empirical results show that the SARIMA‐LSTM combination model is more accurate than the single prediction methods and other combination model. Its accuracy increases by 10.0–27.7%.

Related collections

Most cited references 14

Record: found
Abstract: found
Article: not found

Long Short-Term Memory

Jürgen Schmidhuber, Jürgen Schmidhuber (2002)

Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient-based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O(1). Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.