A machine-learning forecast for sunspot cycle 25

The dynamic activity of the Sun, governed by its cycle of sunspots modulate our solar system space environment creating space weather. Severe space weather leads to disruptions in satellite operations, telecommunications, electric power grids and air-traffic on polar routes. We use four different machine-learning algorithms, all of them belonging to a class called recurrent neural networks, for forecasting. We apply one of these algorithms -- reservoir computing -- to forecast the magnetic field generated by a stochastic dynamo model. We find that the forecast is inaccurate for times longer than the first cycle. We next apply the algorithms to forecast solar sunspot data. By comparing forecasts for cycles 22, 23, and 24 we conclude that a minor variation of reservoir computing algorithm performs the best. The standard reservoir computing forecasts that solar cycle 25 is going to last about ten years, the maxima is going to appear in the year 2024 and the maximum number of sunspots is going to be 113 (±15). A minor variation of the standard algorithm gives the forecast for duration and peak timing as that of the standard algorithm, but the forecast for the peak amplitude is 124 (±2) -- within the upper bound of the standard algorithm. We conclude that sunspot cycle 25 is likely to be a weak, lower than average solar cycle, somewhat similar in strength to sunspot cycle 24.