Summary
In this chapter, we looked at how to prepare our features through feature engineering and how to prepare our labels through labeling.
In the first section, we learned that feature engineering includes creating new and missing features, transforming existing features, extracting features from a high-dimensional dataset, and using methods to select the most predictive feature for ML training.
In the second section, we learned that labeling is essential and tedious. Therefore, tooling such as Azure Machine Learning data labeling can be a blessing to alleviate this time-consuming task.
The key takeaway from this chapter is that creating, transforming, and selecting predictive features has the biggest impact on the quality of the ML model. No other step in the ML pipeline will have more influence on its outcome.
To pull off quality feature engineering, you must have intimate knowledge of the domain (or you must know someone with that knowledge) and a clear grasp of...