# Sentiment Analysis of Amazon Oral Care Product Reviews

Overview

This project aims to analyze customer sentiment from Amazon reviews of oral care products using multiple machine learning and NLP models. The objective is to classify customer reviews into positive, negative, or neutral sentiments and visualize the results in a Power BI dashboard.

Hackathon Problem Statement

E-commerce platforms receive vast amounts of customer feedback daily. Understanding customer sentiment helps brands:

• Improve products
• Address customer concerns
• Enhance user experience

Challenge

Develop an automated solution to process and analyze thousands of product reviews efficiently and generate meaningful insights for decision-making.

Dataset

Source: Amazon oral care product reviews dataset
Size: 1000+ reviews

Attributes

• Id: Unique identifier for each review
• ProductId: Unique product identifier
• UserId: Unique user identifier
• ProfileName: Reviewer's name
• HelpfulnessNumerator & HelpfulnessDenominator: Measures of review helpfulness
• Score: User-provided rating (1-5 stars)
• Time: Timestamp of the review
• Summary: Short description of the review
• Text: Full review text

Sentiment Analysis Models Used

• VADER (Valence Aware Dictionary and sEntiment Reasoner) – Rule-based sentiment analysis model.
• TextBlob – Lexicon-based approach for polarity detection.
• BERT (Bidirectional Encoder Representations from Transformers) – Deep learning-based NLP model for contextual sentiment analysis.
• Other ML models (if applicable) – Additional models like Logistic Regression, SVM, or LSTM-based classifiers.

Expected Results and Insights

• Sentiment Distribution: Percentage of positive, negative, and neutral reviews.
• Model Performance Comparison: Accuracy, precision, recall, and F1-score for each model.
• Product Sentiment Trends: Time-based sentiment trends for different oral care products.
• Common Keywords & Topics: Frequently occurring words and phrases in positive and negative reviews.
• Helpfulness Score Analysis: Relationship between sentiment and review helpfulness ratings.
• Power BI Visualization: Interactive dashboard to explore sentiment trends, word clouds, and review insights.

Implementation Steps

1. Data Preprocessing
   • Cleaning, tokenization, and feature extraction from review text.
2. Sentiment Classification
   • Applying multiple models to classify sentiment.
3. Result Aggregation
   • Consolidating outputs from different models for comparison.
4. Exporting Results
   • Saving analysis results into a CSV file for Power BI integration.
5. Visualization in Power BI
   • Creating graphs, charts, and dashboards for data-driven insights.

Technologies Used

• Python: Data processing and sentiment analysis
• NLTK, TextBlob, Transformers (Hugging Face): NLP libraries
• Pandas, NumPy: Data handling and manipulation
• Matplotlib, Seaborn: Data visualization in Python
• Power BI: Interactive dashboard creation

How to Use

1. Run the Jupyter Notebook or Colab script to perform sentiment analysis.
2. Generate the all_model_results.csv file.
3. Load the CSV file into Power BI.
4. Explore insights using the Power BI dashboard.

Conclusion

This project provides a scalable approach to analyze customer sentiment using multiple NLP models. The insights help businesses to:

• Understand user feedback
• Improve product quality
• Enhance customer satisfaction

Future Enhancements

• Advanced deep learning models
• Real-time sentiment tracking
• Multilingual sentiment analysis