Question

You are building a Spark ML pipeline on a large distributed DataFrame that includes a categorical column 'region' with string values. You want to convert these strings into numeric indices and then one-hot encode them. Which sequence of pipeline stages correctly accomplishes this in Spark ML?

Accepted Answer

StringIndexer on 'region' → OneHotEncoder on the indexed output → VectorAssembler

Answer

OneHotEncoder on 'region' → StringIndexer on the encoded output → VectorAssembler

Answer

VectorAssembler on 'region' → StringIndexer → OneHotEncoder

Answer

OneHotEncoder on 'region' directly, then VectorAssembler (no StringIndexer needed)