Machine Learning: 2 Books in 1: Machine Learning for Beginners, Machine Learning Mathematics. An Introduction Guide to Understand Data Science Through the Business Application

Ridge Regression. This is a technique commonly used to analyze data that
suffer from multicollinearity. Using ridge regression properly can reduce
standard errors and make your model more accurate, depending on the
characteristics of your data.
Ridge regression can be useful when your data contains independent
variables with a high correlation. If you could make a prediction about an
independent variable by using another independent variable, then your
model is at risk of multicollinearity. For example, if you are using variables

---

that measure a person’s height and weight; these variables are likely to
create multicollinearity in the model.
Multicollinearity can affect the accuracy of your predictions. To avoid
multicollinearity, be cognizant of the type of predictive variables you are
using. Type of data you are using, as well as the method of collection, 
could be the cause of multicollinearity. There is a chance you may not have
selected a broad enough range of independent variables. Your data points
could be too similar because your choice of independent variables is
limited.
Multicollinearity can also be caused by having a model that is too specific.
You have more variables than you have data points. If you have decided to
use a linear model, and that has made multicollinearity worse, then you can
try to apply a ridge regression technique.
Ridge regression works by allowing a little bit of bias into the model in
order to make your predictions more accurate. This technique is also known
as regularization.
Another method is to improve the model's accuracy by standardizing
independent variables. The simplest way is to change the coefficients of
some independent variables to zero to reduce complexity. However, we
won't just change them to zero, but standardization implements a system
that rewards coefficients that are nearer to zero. This will cause coefficients
to shrink, so the complexity of the model is less, but the independent
variables remain in the model. This will give the model more bias, but it's a
trade-off for more accurate predictions.

---