Feature Engineering

Source: Alteryx
Published: October 2019

Feature Engineering

Circulated: March 27, 2020

Machine learning uses features which are measurable properties of the object you’re trying to analyze. An example is using the number of bedrooms of a house (the feature) to predict the value of the house (the target).

Feature engineering is creating new features from your existing ones to improve predictive model accuracy. Here is an example of using the feature of flight date and time to predict the target of flight status.

Machine learning uses features which are measurable properties of the object you’re trying to analyze. An example is using the number of bedrooms of a house (the feature) to predict the value of the house (the target).

Feature engineering is creating new features from your existing ones to improve predictive model accuracy. Here is an example of using the feature of flight date and time to predict the target of flight status.

Before feature engineering:

After feature engineering:

The insight suggested from this example is that to predict a flight status, hour of day is likely to be a strong indicator of whether a flight is on time or delayed.