In the late 90-s of the previous century the American Society of Epileptologists defined a priority for research as "possibilities to predict a seizure, early determinate and reduct". A method, which would allow the prediction of epileptical seizure's onset based on the EEG data registered with the patient with an epilepsy disease, would also allow implementing the new approach to treatment. If it became reliably possible to predict a moment of seizure, based on the EEG dynamics, one could create an automated closed loop system to prevent a seizure. In the article a number of works regarding this subject were reviewed. Also own results were discussed which were derived from analyses of brain electrical activity of rats with absence epilepsy and with the use of own developed software. Moreover specifics of absence initiation and course were discussed, as well as formation mechanism of thalamus-cortical loop, existing abilities of reduction not only absences, but also cognitive and emotional dissociations. Also described results of analyses of the EEG time series, that were derived by computation of correlation dimension with own developed software.