Most Reviewed Products Analysis
This view highlights top-performing products based on review volume. It combines review count, rating, and sentiment score to identify high-engagement products and support product prioritization.
Sentiment-Driven Product Recommendation & Sales Insight System
This project follows a complete analytics workflow, starting from raw review data and ending with interactive dashboard insights for product evaluation and business decision-making. This project combines machine learning, sentiment analysis, and interactive visualization to analyze Amazon product reviews and generate actionable insights for product recommendation and business decision-making.
Overview
The system transforms unstructured customer reviews into structured insights by applying sentiment analysis and combining it with features such as rating, price, discount, and verified purchase status. The results are presented through an interactive dashboard for real-time exploration.
Business Problem
E-commerce platforms contain massive amounts of review data that are difficult to interpret. Customers struggle to identify high-performing products, while businesses lack clear insights into sentiment trends and product performance. This project addresses these challenges through automated sentiment analysis and dashboard visualization.
Objectives
- Apply machine learning models for sentiment classification
- Analyze review sentiment and product performance
- Extract insights from rating, price, and review data
- Build an interactive dashboard for visualization
- Evaluate model performance using accuracy and F1-score
Technical Stack
Python (Flask)
Pandas
NumPy
Machine Learning Alagorithms
TextBlob
HTML / CSS / JavaScript
Chart.js / D3.js
Web development
Machine Learning Models Evaluation
- Logistic Regression (Best Performance)
- Naive Bayes
- Support Vector Machine (SVM)
Key Results
Logistic Regression achieved the best performance after hyperparameter tuning, reaching 90.15% accuracy and 0.9008 F1-score, making it the most suitable model for sentiment classification.
Business Impact
- Improves product discoverability for customers
- Helps identify top-performing products
- Supports data-driven business decisions
- Transforms raw review data into actionable insights
Product-Level Sentiment Analysis
This page provides detailed insights into a selected product, including review text, sentiment classification, rating, and metadata, enabling deeper understanding of customer feedback.
Recommendations
- Prioritize high-sentiment, high-rating products in product promotion and recommendation strategies.
- Monitor products with high review volume but weaker sentiment to identify potential quality or customer satisfaction issues.
- Use verified-purchase-based analysis to improve trust in product evaluation and recommendation logic.
- Combine sentiment with price, rating, and helpful votes when assessing product performance rather than relying on one metric.
- Extend the system with real-time review monitoring and advanced NLP models such as BERT or RoBERTa for future improvements.