This paper presents a study on predicting discrete geomagnetic signals using various statistical models including autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) neural networks. The research is based on data from the Surlari geomagnetic observatory and includes methods like Fourier and wavelet analysis for better frequency localization of signals. The results aim to improve the forecasting of geomagnetic storm events through probabilistic models integrated with neural network predictions.