it is hard to see whether the available samples are limited compared to the complexity of the model, did anyone tried to run with different amount of samples? is performance going to change?
when data size is limited, training data may not be able to cover different types anomalies which makes the prediction and generalization much more difficult.
it is hard to see whether the available samples are limited compared to the complexity of the model, did anyone tried to run with different amount of samples? is performance going to change?
when data size is limited, training data may not be able to cover different types anomalies which makes the prediction and generalization much more difficult.