Linear Regression Explained 101

Machine Learning has become ‘the’ hottest topic nowadays, but I found everyone focus on fancy algorithms, and no one cares about where it all begins. When I looked back at my academic path, I found even though algorithms are getting more and more advanced, the fundamental concept remains unchanged:

Steps to understand a model

These steps are what I learned through the most classic machine learning algorithm — Linear Regression. I want to briefly explain what regression is, and how it influence my entire view on Machine Learning.

What is regression:

Regression is to use statistical tools to find relationship between X and Y. There are two types of relationship: correlation and causation. Regression can only reveal correlation. It can be served as a tool to model causation, but that needs strict assumptions.

Why do we need regression analysis:

1. Description: People want to link everything. Regression helps to describe the association between X and Y (positive, negative, strong, weak etc.).
2. Prediction: We can predict future Y with future observation of X. For example, with a model of smoking behavior and health problems. We can now what will be a person’s health status if he or she continues to smoke.
3. A tool for causal inference…