[Supervised ML] Feature Engineering - 7

1. What is Feature Engineering?

---

Using intuition to design new features, by transforming or combinging original features.

 

Example) House price. 

If we are given the dimension of the house such as length and width, we can create a new variable to include for price prediction of a house. For example, we create a area variable with length and width.

 

2. Feature engineering in Polynomial Regression

---

We may want to create a curve line instead of a linear line. 

 

We can use feature engineering by multiplying x by exponents. But now x has a different scale of value than the original x. This is where feature scaling comes to play its job. We need to rescale the data for better gradient descent performance. 

 

Other than x^2, x^3, there are other choices of feature.