Linear regression models the relationship between one or more input variables and a real-valued output. For example, car characteristics such as build year, engine, and mileage can be used to estimate market price and judge whether a dealer’s asking price is fair. section 16.4.1
Begin by choosing a category of mathematical functions that can approximate the relationship in the dataset. A simple choice is a line, y = mx + b; the parameters m and b then determine the particular line learned from the data. section 16.4.1
The modeling progression is:
- Choose a model family, such as a constant function, a line, or a more complicated curve. section 16.4.1
- Fit the family’s parameters so that one specific function describes the observed data as well as possible. section 16.4.1
- Balance model complexity against available data: simpler models generally require fewer data points to learn. section 16.4.1
Estimating a car’s market price
Train on cars whose characteristics and prices are known, then provide the characteristics of a car at a dealership. The model’s real-valued output gives an estimated price for comparison with the asking price. section 16.4.1
