P300 waveform is triggered in response to precise stimuli whose order of presentation can be designed to control the brain-computer system. Accurately decoding However, P300 is generally an Oddball paradigm, that is, the event that excites P300 is a small-probability event. It is a binary problem to analyze target and non-target P300 signals, and the extreme class imbalance problem (CIP) of data is faced in target recognition. In this study, an integrated data augmentation method applicable to P300 signal datasets was proposed. Specifically, a data augmentation method that can realize P300 target and non-target signal classification was found out through reasonable combined oversampling and under-sampling. Finally, the effectiveness of the integrated data augmentation method was verified by the classification results acquired through the deep network method.