Sentiment Analysis of Real-time Flipkart Product Reviews

An end-to-end MLOps project that performs sentiment analysis on real Flipkart product reviews and deploys a trained model using Streamlit Community Cloud. The application classifies customer reviews as Positive or Negative and provides insights into customer satisfaction and pain points.

App Link: https://sentiment-analysis-flipkart-reviews.streamlit.app

---

Project Structure

Sentiment_Analysis_Project/
│
├── data/
│   └── data.csv
│
├── artifacts/
│   ├── vectorizer.pkl
│   ├── sentiment_model.pkl
│   └── model_metadata.json
│
├── src/
│   ├── config.py
│   ├── data_loader.py
│   ├── preprocessing.py
│   ├── feature_engineering.py
│   ├── train.py
│   ├── evaluate.py
│   ├── model_registry.py
│   └── inference.py
│
├── app.py
├── run_training.py
└── requirements.txt

---

Project Overview

Objective:
To analyze real-time Flipkart product reviews and:

• Classify reviews into Positive or Negative sentiment
• Understand common pain points from negative reviews
• Deploy a scalable, production-ready ML system using MLOps best practices

Product Analyzed:
YONEX MAVIS 350 Nylon Shuttle (Flipkart)

---

Dataset Description

• Source: Real-time scraped Flipkart reviews
• File: data.csv
• Total Reviews: 8,518

Key Features

• Reviewer Name
• Rating
• Review Title
• Review Text
• Place of Review
• Date of Review
• Up Votes / Down Votes

---

Data Preprocessing

• Text cleaning (lowercasing, special character & punctuation removal)
• Stopword removal
• Lemmatization
• Handling missing and noisy data
• Label creation using ratings (Positive / Negative)

---

Feature Engineering

Multiple text representation techniques were explored:

• Bag of Words (BoW)
• TF-IDF (final selected approach)
• (Design supports future extension to Word2Vec / BERT)

---

Model Training & Evaluation

• Multiple ML models trained and compared
• Best model selected using F1-Score
• Model artifacts and metadata saved for reproducibility

Evaluation Metric

• F1-Score (chosen due to class imbalance)

---

Streamlit Web Application

The deployed web app allows users to:

• Enter a custom product review
• Get real-time sentiment prediction
• View model information and performance (F1-score)

---

How to Run Locally

1. Clone the Repository

git clone https://github.com/Avik-Das-567/Sentiment-Analysis-Flipkart-Product-Reviews.git
cd Sentiment-Analysis-Flipkart-Product-Reviews

2. Create Virtual Environment & Install Dependencies

python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

3. Train the Model

python run_training.py

4. Run Streamlit App

streamlit run app.py

---

Deployment

• Platform: Streamlit Community Cloud

• Deployment via GitHub repository

• Artifacts committed for fast startup and inference-only runtime

---

Key Highlights

• End-to-end MLOps workflow

• Clean modular code structure

• Real-world dataset

• Production-style deployment

• Easily extensible to advanced NLP models (BERT, Transformers)