Multicollinearity
noun
say it: “mul-tee-koh-lin-ee-AIR-ih-tee” /ˌmʌltikoʊˌlɪniˈɛrəti/
In statistics, when two or more inputs to a model are closely related, making it hard to tell how much each one affects the result.
What is multicollinearity?
Multicollinearity is when features move together, such as a home's size and its number of bedrooms. The model can't tell their effects apart, so it splits the credit between them in ways a few rows can change, and each coefficient stops meaning much on its own. The guesses can still be good.
import pandas as pd
homes = pd.DataFrame({"sqft": [900, 1200, 1500, 1800, 2400], "bedrooms": [1, 2, 3, 3, 4]})
print(homes["sqft"].corr(homes["bedrooms"]).round(2)) # 0.96: they rise together
In a sentence
- “Size and bedrooms rise together, a classic case of multicollinearity.”
- “Because of multicollinearity, the bedrooms coefficient came out negative even though bigger homes cost more.”
Level: Freelancer · More AI Basics terms at this level
Updated