For focal epilepsy patients, correctly identifying the seizure onset zone (SOZ) is essential for surgical treatment. In automated realistic SOZ identification, it is necessary to identify the SOZ of an unknown patient using another patient’s electroencephalogram (EEG). However, in such cases, the influence of individual differences in EEG becomes a bottleneck. In this paper, we propose the method with domain adaptation and source patient selection to address the issue of individual differences in EEG and improve performance. The proposed method was evaluated on intracranial EEG data from 11 patients with epilepsy caused by focal cortical dysplasia. The results showed that the proposed method significantly improved SOZ identification performance compared to existing methods without domain adaptation and source patient selection. In addition, it was suggested that data from residual-seizure patients may have adversely affected estimation performance. Visualization of the prediction on MRI images showed that the proposed method might detect SOZs missed by epileptologists.