Application of deep reinforcement learning for extremely rare failure prediction in aircraft maintenance

•A novel deep reinforcement learning framework is developed for modelling rare failure prediction in a complex system.•The reward function, experience replay memory, and eligibility trace are introduced as key ingredients.•It handles model bias for extremely imbalanced classification problems.•An aircraft central maintenance system dataset is used to verify the performance of the proposed method.•The influence of extremely rare failure prediction on the proposed method is analyzed.