One of the interesting problems in the analysis of electroencephalogram (EEG) signals is its possible use in epileptic seizure prediction. In recent years an important stream of research was directed to addressing this issue. This paper describes a technique based on wavelets to predict these seizures through the analysis of the EEG. The numerical method used is developed and the results obtained are shown. The EEGs analyzed were for patients for surgery and depth electrodes were used.