SD-LSTM: A Novel Semi–Decentralized LSTM Model for Time Series Forecasting

Time series forecasting is a crucial task in various fields such as finance, weather forecasting, and energy management. Traditional centralized models have been widely used for time series forecasting, but they often suffer from the curse of dimensionality and the scalability issue. To address these challenges, a novel semi-decentralized LSTM (SD-LSTM) model is proposed in this paper.

Key Contributions:
Propose a novel semi-decentralized LSTM model (SD-LSTM) for time series forecasting.
Design a hierarchical architecture with a central node and multiple decentralized nodes.
Develop a novel decentralized LSTM module for each node.
Evaluate the proposed model on several real-world time series datasets.

“The proposed SD-LSTM model can effectively capture the complex temporal and spatial relationships in time series data, and achieve state-of-the-art performance in various scenarios.”

Methodology

The SD-LSTM model consists of a central node and multiple decentralized nodes. Each decentralized node is equipped with a novel decentralized LSTM module, which is designed to capture the local temporal relationships in the time series data. The central node is responsible for fusing the information from all decentralized nodes to generate the final forecasting results.

Results

The proposed SD-LSTM model is evaluated on several real-world time series datasets, including the Electricity Demand dataset, the Weather Forecasting dataset, and the Stock Market dataset. The experimental results demonstrate that the proposed model outperforms the state-of-the-art centralized models in terms of forecasting accuracy and computational efficiency.

Conclusion

SD-LSTM is a novel semi-decentralized LSTM model for time series forecasting, which can effectively capture the complex temporal and spatial relationships in time series data. The proposed model can be applied to various fields such as finance, weather forecasting, and energy management, where accurate time series forecasting is crucial.

Author's Summary: This paper presents a novel semi-decentralized LSTM model for time series forecasting, which can effectively capture complex temporal and spatial relationships in time series data. The proposed model outperforms state-of-the-art centralized models in various scenarios, making it a promising solution for real-world time series forecasting applications.

springerprofessional.de — 2025-11-22