Pipeline components

Pipelines consist of a set of components joined together such that the DataFrame produced by one component is used as input for the next component. The components available are split into two classes: transformers and estimators.

Transformers

Transformers transform one DataFrame into another, normally by appending one or more columns.

The first step in our spam classification algorithm is to split each message into an array of words. This is called tokenization. We can use the Tokenizer transformer, provided by MLlib:

scala> import org.apache.spark.ml.feature._
import org.apache.spark.ml.feature._

scala> val tokenizer = new Tokenizer()
tokenizer: org.apache.spark.ml.feature.Tokenizer = tok_75559f60e8cf 

The behavior of transformers can be customized through getters and setters. The easiest way of obtaining a list of the parameters available is to call the .explainParams method:

scala> println(tokenizer.explainParams)
inputCol: input column name (undefined)
outputCol...