09/18/2023

Based on the previous class of Simple linear regression class today we have discussed about Multiple linear regression which comes under Quadratic Model.

 

A Quadratic Model is a model which forms a linear regression in the form of U shaped or a Parabolic which might shape. It helps to classify data points into different categories.

Quadratic model will help to evolve new features from the existing features which enhances the performance of a model

.

 

 

Overfitting

It is a undesirable function occurs when a Machine Learning model gives accurate data to only well trained data but not to the new input data or unseen data which will be causing an error in prediction.

Overfitting is caused because of too much training of data is given to a Machine Learning model, where it creates a noisy data and model doesn’t categorize the data well mannered.

Main 3 reasons to cause this overfitting are:

1. High complex Machine Learning model
2. Size of training data

• More focus on training data set instead of unseen data

So, overfitting can be reduced by following these methods as followed by:

1. Keep ML model as simple as possible
2. Training data need to be increased.

• Monitor the training period of a model to prevent drop out of training.

1. Ridge Regularization (R²) which helps to find out whether model fits or not.

Multiple Regression

It is statistical method to find the relationship between single dependent variable and multiple independent variables.

 

y=β₀​+β₁​x₁​+β₂​x₂​+…. β_n​x_in+ϵ

where, y is Dependent variable

             x₁ is Independent/Explanatory  variable

          β₀  is y intercept

          β_n  slope coefficient of independent variable

            ϵ is errors or residuals

This will help to  predict for larger complex data model consideration based on the business case.