Impact of Temperature Trends on RNN and LSTM Performance in some Stations in the Netherlands

This study investigates the impact of temperature trends on the performance of Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) networks in predicting temperature across five stations in the Netherlands from ۲۰۰۰ to ۲۰۲۴. The Mann-Kendall test was employed to analyze temperature trends, revealing that non-trending data tend to yield higher prediction accuracy. Optimally configured RNN and LSTM networks were used for temperature forecasting, with results indicating that the RNN network demonstrated more stable performance and a slightly lower average RMSE (۱.۶۲) compared to the LSTM network (۱.۶۴). This research highlights the significant effect of selecting an appropriate seed and precisely tuning parameters such as time steps, the number of neurons, and the number of epochs on temperature prediction accuracy. Future research suggestions include applying these findings to other geographical locations and examining network performance with trending data to further enhance temperature prediction models.

کلیدواژه ها:

Temperature prediction ، Recurrent Neural Network ، Long Short-Term Memory ، Mann-Kendall test