Prediction of solar cycle 25 using ARIMA models and LSTM neural networks

Abstract

A study was conducted to predict the number of sunspots in Solar Cycle 25 using two models: a Long short-term memory (LSTM) recurrent neural network model and an autoregressive integrated moving average (ARIMA) model. The data used to train the models was obtained from the website of the World Data Center SILSO, Royal Observatory of Belgium in Brussels, from 1749 to 2018. Our LSTM model demonstrated exceptional performance (RMSE=3.6) compared to the best ARIMA model (RMSE=32.6). This showed that our LSTM model is significantly more accurate in terms of prediction, with an 89% improvement in RMSE reduction. According to our LSTM model, the maximum number of sunspots in Solar Cycle 25 is expected to occur in March 2025, reaching a maximum value of 182 sunspots. In contrast, the ARIMA model predicts that the maximum will be reached in December 2024, with a maximum value of 99 sunspots.