DSC400: Amazon Reviews Analysis

Comprehensive analysis of Amazon product reviews from 2000–2018, combining ML, NLP, and dashboard visualization for sentiment & rating prediction.

GitHub Repo 💻 Project Page 🌐 Dashboard 📊 Dataset 📦

Project Overview

This project analyzes 129,794 Amazon reviews across 58,902 products to uncover patterns in review sentiment, product ratings, and reviewer behavior. Machine learning models and transformer-based NLP were applied to predict review ratings and extract actionable insights.

129,794

Reviews

58,902

Products

82.6%

Sentiment Accuracy

66%

5-Star Reviews

Key Insights & Research Questions

Positive sentiment correlates strongly with higher star ratings.
Low-star reviews tend to be longer and more detailed.
Verified purchases slightly influence rating patterns.
Hierarchical models improved mid-range rating prediction.

Dashboard Preview

Created by: Kirsten Livingston – DSC 400