Summary: The generalized linear model (GLM) extends from the general linear model to accommodate dependent variables that are not normally distributed. GLM is a methodology for modeling relationships between variables.
- Insurance and Loss Prediction
- Fraud Detection/Payment Default/Mortgage Default
- Medical Disease Prediction
Simple Explanation of GLM:
GLM Has Three Parts:
1.Response Component – Random or Dependent Variable
2.Predictor Component – Independent Variables
3.Link Function – A function that links predictor components to predicted mean of the response. Defines the noise or error around the response mean.
Common Family Functions:
GLM Mainly Supports Three Families:
1.Binomial or Linear Regression – Dependent Variable only has 2 possible values (T/F or 1/0)
2.Poisson Regression – Independent events that occur over a Time series/Space
3.Gaussian Regression – Data is grouped around a single mean (Bell Curve Distribution)