kruts/Fake-Review-Detector
🔍 Detect fake product reviews using NLP techniques, TF-IDF, and Logistic Regression, with an interactive Streamlit app for real-time predictions.
What's novel
🔍 Detect fake product reviews using NLP techniques, TF-IDF, and Logistic Regression, with an interactive Streamlit app for real-time predictions.
Code Analysis
8 files read · 3 roundsA small TF-IDF + Logistic Regression classifier with 6 hand-crafted behavioral features (sentiment, exclamation count, all-caps, cliché phrases, length, word uniqueness) that labels product reviews as FAKE or REAL, wrapped in a minimal Streamlit UI.
Strengths
The code is clean, readable, and well-organized for its size; the feature engineering (behavioral signals like all-caps tokens, repeated phrases, unique word ratio) is a reasonable complement to TF-IDF and shows some thought beyond a pure bag-of-words approach.
Weaknesses
The sample dataset is only ~60 rows of 7 unique reviews repeated, making the perfect 1.0 AUC meaningless; there are zero tests, duplicated prediction logic across predict.py and streamlit_app.py, unused dependencies (plotly, nltk), and the README's installation instructions (download ZIP, double-click) don't match the actual repo structure.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
7
Contributors
2
Files
16
Active weeks
3
Repository
Language
Python
Stars
1
Forks
0
License
MIT