Answer: To combine transformers and estimators.

In Spark's MLlib, pipelines are designed to streamline machine learning workflows by chaining together transformers and estimators. This allows for a structured approach to data preparation, feature engineering, and model training. The other options, while related to the functionalities within MLlib, do not represent the primary purpose of pipelines. For instance, training a model or implementing the .fit() method are actions performed by estimators, not pipelines themselves. Similarly, creating a new DataFrame or appending columns are outcomes of specific transformations within a pipeline, not its main objective.

Author: LeetQuiz Editorial Team