DATA MINING ON F2 PEAK OVER SOUTH AFRICA

E. GULARTE; G. BAUME ; D. CARPINTERO

Nicosia

Congreso; International Reference Ionosphere 2019 Workshop; 2019

Frederick University

We have studied the data provided by several South African ionosondes stations at the solar cycle 24 to describe and predict the behavior of the F2 peak of the ionospheric region. Although F2 is a ionospheric layer of great interest in radio propagation and space weather studies, it is also the region which is the most variable, anomalous and difficult to predict.To achieve our goal, we employed a global process dubbed Knowledge Discovery inDatabases (KDD), which consists in searching for useful knowledge that may be hidden in a collection of data. This process includes several stages: data preparation, selection and cleansing, data mining, and interpretation and evaluation of the results.We focus here on the data mining step, which employs different techniques to recognize patterns in large data sets. We used supervised and unsupervised techniques to find the F2 peak dependences. In the former case, the ?decision trees? and ?classification rules? techniques were employed, whereas in the latter case we chose the ?association rules? and ?clustering? methodologies. One of the main advantages in using this kind of processes is their independence of any model, climatological or otherwise.Our results are consistent with the expected relations between the F ionospheric layer and the physical processes that sculpt it. We also compared our results with the IRI climatological model, in order to improve or complement it. In the near future, our results could be also useful for real time measurements.