I have observed the mean accuracy after applying Stratifiedkfold is higher in comparison with the accuracy on the holdout data. I wonder if this can be a sign of over fitting in this case and if so, can someone explain.
Yhe accuracy on the holdout model is around 95.7% and on the Stratifiedkfold set is 96.3%.
I don't understand your second paragraph, but yes. When the training accuracy is higher than the validation accuracy, it indicates overfitting. However, that is always the case (unless lucky). A small amount of overfitting is completely normal, like in your case.
BTW, this question is better suited for datascience.stackexchange or crossvalidated.stackexchange.
