Time-series forecasting of solar activity (sunspot area) from historical observatory records.
This repository contains a selection of example models and notebooks taken from a larger MSc dissertation project in Advanced Aerospace Engineering at the University of Liverpool. The full project explored a wide range of architectures; here I’ve included some representative models for others to inspect and run.
The core objective is to forecast sunspot area using historical solar observations, under consistent preprocessing and evaluation settings.
More specifically, the project focuses on:
- Turning raw sunspot area measurements into supervised learning datasets (input–target sequences).
- Comparing different modelling approaches for time-series forecasting.
- Evaluating and visualising how well different models predict future solar activity.
These notebooks are intended as clean, readable examples, not a complete dump of every experiment run during the MSc project.
The notebooks are designed to work with historical solar / sunspot activity records, for example:
- Royal Greenwich Observatory (RGO) sunspot area data.
- Daily or aggregated sunspot area time series.
Depending on how the repository is configured, you may find data files (e.g. Combined_RGO.txt, daily_area.txt) inside a data/ folder
In all cases, the data is treated as a univariate time series, then transformed into supervised learning examples by building sequences of past values to predict future values.