Deep Learning Experiments - Feature Extraction - Traditional techniques

I have a dataset, the data set is about labeled users' reviews. I want to apply this to preprocessing of the dataset. Split the dataset into testing data and actual data, then input the data into the next stage which uses traditional techniques for word embedding and traditional classifiers as below.

Apply on all Datasets
TF-IDF and Random Forest
Bag of Words and Random Forest
TF-IDF and naive bayes
Bag of Words and naive bayes