Review Sentiment Analyzer

This project is part of a broader initiative to explore both supervised and unsupervised learning approaches in machine learning. This repository contains the implementation of a Sentiment Analysis pipeline inside a Jupyter Notebook and the final trained models.

📌 Project Overview

The goal of this project is to classify customer reviews into one of three sentiment categories:

• Positive
• Negative
• Neutral

The analysis was conducted in two phases: an unsupervised exploration and a supervised training pipeline.

🧠 Learning Scope

🔍 Phase 1: Unsupervised Learning

We began with an unlabeled dataset to uncover hidden structures and patterns using clustering techniques. This helped us better understand the distribution and variety of sentiments present.

• Data Cleaning and Preprocessing
• Vectorization using TF-IDF
• Dimensionality Reduction
• K-Means Clustering to identify possible sentiment groups

🧠 Phase 2: Supervised Learning

Once labels were prepared, we trained multiple supervised models to identify the best-performing one for the sentiment classification task.

• Data preprocessing (text normalization, cleaning)

• TF-IDF vectorization

• Model experimentation with:

  • Logistic Regression
  • Random Forest
  • Naive Bayes
  • Linear Support Vector Classifier (SVC)

• Linear SVC was selected as the best model based on performance metrics

🔗 Dataset

The dataset used is publicly available:

• Amazon Product Reviews Dataset

🧪 Model Artifacts

The final repository includes:

• A trained Linear SVC model (model.pkl)
• A TF-IDF Vectorizer (vectorizer.pkl)

📁 Repository Structure

review-sentiment-analyzer/
├── Sentiment_Analysis_Notebook.ipynb  # Complete ML pipeline notebook
├── linear_svc_model.pkl                          # Final Linear SVC model
├── tfidf_vectorizer.pkl                     # TF-IDF vectorizer
├── README.md                          # This file

💡 Skills Demonstrated

• Text preprocessing (cleaning, vectorization)
• Clustering (unsupervised learning)
• Supervised ML model training and evaluation
• Model selection and comparison

🛠️ Tools & Libraries

• Python
• Scikit-learn
• Pandas
• Numpy
• Matplotlib / Seaborn (for EDA)

👨‍💻 Author

Rohit Gomes

Check out my deployed model: My Model

---

This project was developed to strengthen core ML skills by combining exploratory analysis with practical unsupervised and supervised learning implementation.