QLFHNet: quantum long short-term memory forward harmonic network for weather forecasting utilizing time series data

This paper proposes a Quantum Long Short-Term Memory Forward Harmonic Network (QLFHNet) to forecast weather using time series data. Initially, input time series data are considered, and next, technical indicators are extracted. Then, feature fusion is performed by a Siamese Convolutional Neural Network (SCNN) with a mean distance. Finally, weather is forecasted employing QLFHNet, which is derived by integrating Deep Quantum Neural Network (DQNN) with Long Short-Term Memory (LSTM) using Harmonic Analysis. QLFHNet demonstrated superior outcomes with values of 0.023 for Normalized Mean Squared Error (MSE), 0.151 for Normalized Root Mean Squared Error (RMSE), 0.085 for Normalized Mean Absolute Error (MAE), and 0.900 for R2. These outcomes indicate that the QLFHNet produces highly accurate predictions with minimal deviation from observed weather data.