Creating a feature set
The most important factor involved in successful machine learning is the quality of your input data. A good model with misleading, inappropriately normalized, or uninformative data will not see the same level of success anywhere near a model run over appropriately prepared data.
In some cases, you have the ability to specify data collection or have access to a useful, sizeable, and varied set of source data. With the right knowledge and skillset, you can use this data to create highly useful feature sets.
In general, having a strong knowledge as to how to construct good feature sets is very helpful as it enables you to audit and assess any new dataset for missed opportunities. In this chapter, we will introduce a design process and technique set that make it easier to create effective feature sets.
As such, we'll begin by discussing some techniques that we can use to extend or reinterpret existing features, potentially creating a large number of useful parameters...
