Prediction of cold region dew volume based on an ECOA-BiTCN-BiLSTM hybrid model

This paper presents a hybrid prediction model, ECOA-BiTCN-BiLSTM, for predicting dew in cold areas. The model integrates BiTCN and BiLSTM neural networks to enhance performance. An enhanced Crayfish optimization algorithm (ECOA) with four mixed strategies was employed to optimize the model's hyperparameters and reduce the impact of arbitrary selection. The proposed ECOA-BiTCN-BiLSTM model was validated using dew data from farmland in a northeastern Chinese city. Comparative experiments were conducted against the BiTCN model, the BiLSTM model, the original BiTCN-BiLSTM model, and other models optimized with advanced swarm intelligence algorithms. The experimental results demonstrate that the proposed model achieved a mean absolute error (MAE) of 0.002424, a root mean square error (RMSE) of 0.003984, and a mean absolute percentage error (MAPE) of 0.123050, with a coefficient of determination R2 of 0.999840. These results indicate that the ECOA-BiTCN-BiLSTM model outperforms the other prediction models across all evaluated metrics, offering higher prediction accuracy and highly effective prediction models.