Tachyarrhythmias, such as ventricular fibrillation and atrial fibrillation, are caused by abnormal and complex electrical excitation waves in the heart. Ventricular fibrillation is a fatal condition that can lead to sudden cardiac death, and atrial fibrillation increases the risk of stroke due to thrombosis. Today, ablation therapy is widely used to treat fibrillation by ablating the abnormally excited area. However, although various ablation strategies have been proposed, the optimal ablation strategy has not been established. To establish an objective and effective fibrillation ablation strategy, we attempted to construct a machine learning model that selects the optimal ablation target based on the excitation pattern during fibrillation. We report the results of training a deep neural network model that selects the best ablation target based on the time series of the membrane potential distribution, which represents the excitation state of each cell, using a two-dimensional electrophysiological simulation model.