How tanishkraj2005/flipkart-fake-review-detection works
This is a Flipkart review-fraud analytics product: a batch pipeline cleans reviews, engineers sentiment and fraud signals, assigns rule-based labels, trains an ML comparison model, and exports analyzed data. Users then explore the results in a Streamlit dashboard or Power BI, and can analyze an individual summary in the app. Product-level fraud profiles and exact-duplicate reporting are additional analytical capabilities. The app excerpt is partial, and the Power BI implementation is not in the tree, so those links reflect only documented behavior and visible file paths.
Main entry points: main.py Batch Pipeline · app.py Streamlit App · feature_engineering.py Feature Engineering · .gitkeep Processed Reviews
Components 16
Batch Analysis
- Batch Pipeline
main.py
Detection Analytics
- Text Cleaning
src/clean_text.py - Feature Engineering
src/feature_engineering.py - Fraud Signals
src/flags.py - Sentiment Analysis
src/sentiment.py - Rule Classifier
src/fake_detector.py - ML Classifier
src/ml_model.py - Exact Duplicates
src/duplicate_detection.py - Product Profiles
src/reviewer_detection.py
Review Interfaces
- Processed Reviews
Data/.gitkeep - Power BI Export
main.py - Streamlit App
app/app.py
Other components
- Dataset Provider
- Raw Reviews
Data/.gitkeep - App User
- Power BI
Connections 19
- Dataset Provider → Raw Reviews: provides reviews
- Raw Reviews → Batch Pipeline: loads data
- Batch Pipeline → Text Cleaning: cleans text
- Batch Pipeline → Feature Engineering: builds features
- Feature Engineering → Fraud Signals: computes signals
- Feature Engineering → Sentiment Analysis: scores sentiment
- Batch Pipeline → Rule Classifier: labels reviews
- Batch Pipeline → ML Classifier: trains and predicts
- Batch Pipeline → Exact Duplicates: tags duplicates
- Batch Pipeline → Processed Reviews: writes results
- Batch Pipeline → Power BI Export: exports data
- Processed Reviews → Streamlit App: loads results
- App User → Streamlit App: explores and analyzes
- Streamlit App → Text Cleaning: cleans input
- Streamlit App → Fraud Signals: checks signals
- Streamlit App → Sentiment Analysis: scores input
- Streamlit App → Rule Classifier: calculates probability
- Streamlit App → Product Profiles: builds product profile
- Power BI Export → Power BI: feeds reporting