Multinominal Logistic Regression

Contents 

1. Introduction to Regression……………………………………………………………………………
2. What is Multinominal Logistic Regression? ………………………………………………………………
3. Predicting Mobile Price range using Python……………………………………………………….. 
4. Conclusion…………………………………………………………………………………………………………….
5. Introduction to Regression  

 Regression analysis is a Predictive modelling technique used to analyse the  relationship between the one Dependent variable usually denoted as “Y” and multiple  independent variablesenoted as (X1, X2, etc.,).

There are multiple Regression Techniques as mentioned below, However Linear and  Logistic Regression analysis were widely used by data analysts. 

Types of Regression Techniques: – 

1. Linear Regression 
2. Logistics Regression 

o Binominal Logistics Regression 

o Multinominal Logistic Regression 

3. Ridge Regression 
4. Lasso Regression 
5. Polynomial Regression 
6. Bayesian Linear Regression 

2.What is Multinominal Logistic Regression? 

Multinomial Logistic Regression is an extension of Logistic Regression. By Default, Logistic is limited to Two-Class classification problem like (True or False, Yes or No,  Male or Female) etc., 

If more than two classes need to predict like one-vs-rest (Small, Medium, Big) can  allow logistic regression to be used for multi-class classification problems, although 

they require that the classification problem first be transformed into multiple binary  classification problems. 

Instead, Multinominal logistic Regression is an extension to the Logistic regression  model that involves changing the loss function to cross-entropy loss and predict  multinominal probability distribution to support native multi-class classification  problems.  

A logistic regression that is adapted to learn and predict multinomial probability  distribution is called as Multinominal Logistic Regression (predicts more than two  classes for each input). 

Similarly, by default or standard logistic regression is referred as Binominal Logistic  Regression (predicts two classes for each input). 

3. Predicting Mobile Price range using Python 

Problem Statement: – 

Market size of mobile phones if growing every day and so is the competition. To  capture the maximum market electronics companies, try to make improvements in  their products. However, sales of mobile phones depend on various factors like  demand, technology, marketing, brand, availability, user experience, service, price,  etc. As we understand that selling price makes a huge difference when it comes to  sales and profits. Estimating an optimal price for a new mobile phone can be a tricky  task especially when you are new in the business or when you want to launch a new  kind of mobile phone in the market. 

As part of this exercise, we will try to estimate the price range for a given mobile  phone using given feature information. These details are collected from various  similar companies. 

DATASET Attribute Information: –

With given different Independent variables (Xs) like battery_power, clock_speed,  dual_sim etc., we need to predict multi-class price range (Y) low cost – 0, medium  cost – 1, high cost – 2 and very high cost – 3. 

Download required Python libraries 

Perform Basic Data Analysis and Preparation

Perform Statistical analysis using describe () 

Check the Correlation within the variables

Building the Logistic Regression model, getting Dependent variable (Y ‘price_range’) and  Independent variables (Xs) 

Predict the y_test ( Price range) with given input (X_test) with defined model. 

Print the classification report to understand the model accuracy.

Predicting Mobile Price range using Python 

4. Conclusion 

Multinominal Logistic Regression is an extension of logistic regression, it is used to analysis  Multi-class classification problems for every single Input.

Author: – G. Boobalan 

Linkedin Profile: https://www.linkedin.com/in/gboobalan/