What is the reason for such low classification accuracy? How can I overcome this problem?

For context, I am working with a dataset for diagnosing AD patients using their EEG recordings. Using a train test split on different epochs produced very high accuracy. However, when computing the train test split by subject, my validation accuracy is incredibly low. Do you have any suggestions for how I can overcome this problem?

Answer 1

[yanqi liu]

yes，sir，may be it is overfitting，so add some batch normal or dropout in layers to retrain the model