In this contribution, a novel approach for the prediction of epileptic seizures is introduced using binary input-output patterns and Boolean CNN with linear weight functions. Two different algorithms are introduced and verified on invasive recordings of different patients.