Method of modeling critical frequency data on the basis of combining the wavelet-transform and autoregression models

This paper is devoted to the development of new approaches and automated systems for the analysis of ionospheric data and also for extracting and classifying local abnormal features encountered in datasets. The main difficulty in dealing with these data is caused by lack of apriori knowledge of the internal structure of data and also lack of a formal model to represent them. The suggested method of multidimensional modeling and forecasting of ionospheric data allows one to distinguish the characteristic structure, perform the forecasting procedure, and automatically detect changes in the parameters of a model employed. The procedure of analyzing ionospheric data includes estimating errors contributed by different factors and minimization of their influence on data representation, decomposing data into components with the help of selected methods and extracting the most informative components together with their identification and classification of abnormal features.