/Customer-Sentiment-Analysis

This is a data science project to analyse the sentiments of customers from Amazon Alexa Reviews data.

Primary LanguageJupyter NotebookMIT LicenseMIT

Customer-Sentiment-Analysis

This is a data science project to analyse the sentiments of customers from Amazon Alexa Reviews data.

Introduction

  • Natural language processing can be used to build predictive models to perform sentiment analysis on social media posts and reviews and predict if customers are happy or not.
  • Natural language processors work by converting words into numbers and training a machine learning model to make predictions.
  • That way, you can automatically know if your customers are happy or not without manually going through massive number of tweets or reviews!

Dataset

https://www.kaggle.com/datasets/sid321axn/amazon-alexa-reviews/code

Findings

Imbalanced Dataset

image

image

Most of the reviews are positive

image

The reviews are mostly shorts less than 500 words

World Cloud

Of the dataset

image

Of the positive reviews

image

Of the negative reviews

image

Naive Bayes Classifier Results

image

Logistic Regression Classifier Results

image