Time Series Forecasting Study Based on Decomposition-Prediction-Optimization Mode

Time series are ubiquitous in everyday life, and their predictive techniques are particularly crucial. However, the accuracy of a single model's prediction is often unsatisfactory, and due to the characteristics of different models, there are many limitations. To enhance prediction accuracy, this study proposes a third-order model of hybrid decomposition prediction optimization. It involves decomposing data using the Singular Spectrum Analysis signal decomposition algorithm, followed by prediction using Long Short-Term Memory neural networks, and subsequently, applying the Grey Wolf algorithm to optimize the weighted prediction results. Through comparative experiments, this hybrid approach in the study outperforms other hybrid models in various evaluation metrics, demonstrating higher precision.