Bayesian optimization of hybrid quantum LSTM in a mixed model for precipitation forecasting

Abstract Precipitation forecasting has important applications in meteorological research. Accurate forecasting is of great significance for reducing the impact of floods, optimizing crop planting plans, rationally allocating water resources, and ensuring traffic safety. However, the factors affecting precipitation are complex and nonlinear, and have spatiotemporal variability, making rainfall forecasting extremely challenging. In response to these challenges, this paper proposes a hybrid model based on temporal convolutional network, quantum long short-term memory network (QLSTM), and random forest regression (RFR) to achieve more accurate rainfall forecasting. The hyperparameters of the model are optimized using the Bayesian optimization algorithm to obtain the best performance. Experiments are conducted on meteorological datasets from Seattle and Ukraine, and the results are verified using mean absolute error (MAE), root mean square error (RMSE), and bias evaluation indicators. The results show that the proposed hybrid model outperforms traditional models such as RFR, support vector machine, K-nearest neighbor, LSTM, and QLSTM in terms of MAE, RMSE, and bias. The proposed model achieves improvements of 1.89 % MAE, 2.65 % RMSE, and 31 % Bias, respectively. These results highlight the improved forecast accuracy and robustness of the proposed hybrid model. This research provides a new approach to weather forecasting and demonstrates the potential of combining quantum computing with traditional machine learning techniques.