A Hybrid Forecasting Model for Time Series Data Based on ARIMA and LSTM with Attention Mechanism

The paper proposes a hybrid forecasting model for time series data that combines the autoregressive integrated moving average (ARIMA) model and the long short-term memory (LSTM) model with an attention mechanism. The proposed model aims to overcome the limitations of the traditional ARIMA model in capturing long-term dependencies and the LSTM model in handling seasonal patterns.

The hybrid model consists of two stages: the ARIMA stage and the LSTM stage. In the ARIMA stage, the model is trained to capture the overall trend and seasonality of the time series data. In the LSTM stage, the model is trained to capture the residual error from the ARIMA stage and the long-term dependencies in the data. An attention mechanism is then applied to the LSTM stage to emphasize important features in the data.

The proposed model was evaluated on six real-world time series datasets and compared with other state-of-the-art models. The results showed that the hybrid model outperformed the other models in terms of forecasting accuracy and stability.

The authors conclude that the proposed hybrid model can effectively capture both short-term and long-term dependencies in time series data, making it a promising approach for forecasting applications.