Best stage order for a Spark ML text classification Pipeline?

Question

Accepted Answer

Tokenizer → HashingTF/CountVectorizer → optional IDF → Classifier (plus label indexing) — Tokenize → vectorize → classify. Estimators come after their feature transformers.

Answer

Classifier → Tokenizer → HashingTF

Answer

VectorAssembler after the classifier only

Answer

StringIndexer on features after prediction only

Best stage order for a Spark ML text classification Pipeline?

Short Explanation

Full Explanation