We all know that PCA (Principal Component Analysis) is a popular statistical tool to reduce the dimensionality in a dataset. SMOTE (Synthetic Minority Over-sampling Technique) allows you to generate minority class data if there is an imbalance in the dataset. I know for a fact that we can use SMOTE on any dataset regardless of whether the dataset has undergone PCA and not.
But which one these methods better generalize the trends within the dataset. While PCA retains the most trends in the data(but not all), does application of SMOTE has any greater impact on the quality of data? Does it have an impact at all?
Here comes the question.
What is the difference between SMOTE before PCA and SMOTE after PCA?
if there is a difference, which practice is recommended/best?