Blending Old and New: How Double Exponential Smoothing and Deep Learning Team Up to Predict Stock Prices

Financial markets are getting smarter with the blend of modern deep learning (DL) and classic time series forecasting methods. Researchers have created a unique hybrid model to predict stock prices for General Electric (GE), Microsoft (MSFT), and Amazon (AMZN). This model combines Double Exponential Smoothing (DES) with a Deep Learning (DL) approach called Dual Attention Encoder-Decoder based Bi-directional Gated Recurrent Unit (DA-ED-Bi-GRU). To top it off, the model uses Bayesian Optimization (BO) to fine-tune its parameters. DES is like an old friend that's really good at spotting trends and seasonal patterns in time series data. Meanwhile, the DA-ED-Bi-GRU part of the DL model digs deep into the complex patterns hidden within the stock data. By combining these two, the hybrid model becomes a powerful tool for understanding stock movements. To see if this model is any good, researchers used several measuring sticks like Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-Square (R²). Theil's U-Statistics (TUS) also joined the party to check the model's precision and accuracy. The results? The hybrid model did a decent job predicting stock prices for GE, MSFT, and AMZN. When teamwork meets technology, it looks like traders and investors can make better decisions.