From 04823c75f47b438e8953facb439751db59706106 Mon Sep 17 00:00:00 2001
From: Kriti Subedi <kreetesubeede123@gmail.com>
Date: Tue, 14 Apr 2026 08:40:19 -0230
Subject: [PATCH]  fix readme file

---
 README.md | 193 ++++++++++++++++++++++++++++--------------------------
 1 file changed, 101 insertions(+), 92 deletions(-)

diff --git a/README.md b/README.md
index 87d1947..f49a91d 100644
--- a/README.md
+++ b/README.md
@@ -2,64 +2,58 @@
 
 ## 📌 Project Overview
 
-This project was developed by a three-person academic team over 10 weeks (part-time, $0 budget).
+This project was developed by a three-person academic team over 10 weeks on a part-time basis with a $0 budget.
 
-The **Review Guard System** is a locally runnable web application that uses **Natural Language Processing (NLP)** and **Machine Learning** to classify online reviews as:
+**Review Guard** is a locally runnable web application that uses **Natural Language Processing (NLP)** and **Machine Learning** to classify online reviews as:
 
-- ✅ Genuine  
-- ❌ Fake  
+* ✅ Genuine
+* ❌ Fake
 
 ## 🚀 Features
-- Single review prediction  
-- Batch CSV upload and processing 
-- Fake/Genuine classification  
-- Fake probability score output  
-- Download results as CSV  
-- Multiple model support (v1, v2, v3)  
-- Fully offline system  
 
+* Single review prediction
+* Batch CSV upload and processing
+* Fake/Genuine classification
+* Fake probability score output
+* Download results as CSV
+* Multiple model support (`v1`, `v2`, `v3`)
+* Fully offline system
 
-
-```markdown
 ## 📂 CSV Format
 
 The CSV file must contain the following columns:
 
+```text
 text,rating,helpful_vote,verified_purchase
-
+```
 
 ## 🎯 Project Goals
 
-- Achieve **greater than 80% classification accuracy** on unseen test data  
-- Develop a **fully offline, locally runnable web application**  
-- Support both **single review prediction and batch CSV processing**  
-- Provide **fake probability scores for each prediction** 
-- Follow an **Agile development methodology with iterative improvements**  
-- Ensure **fast response time and user-friendly interface**  
-
+* Achieve **greater than 80% classification accuracy** on unseen test data
+* Develop a **fully offline, locally runnable web application**
+* Support both **single review prediction and batch CSV processing**
+* Provide **fake probability scores for each prediction**
+* Follow an **Agile development methodology with iterative improvements**
+* Ensure **fast response time and user-friendly interface**
 
 ## 🛠️ Technologies Used
 
-- **Backend:** Python (Flask)  
-- **Frontend:** HTML, CSS, JavaScript  
-- **ML/NLP:** Scikit-learn, NLTK  
-- **Data Processing:** Pandas, NumPy  
-- **Model Storage:** Joblib  
-- **Testing:** Pytest  
-- **CI/CD:** GitHub Actions  
- 
+* **Backend:** Python, Flask
+* **Frontend:** HTML, CSS, JavaScript
+* **ML/NLP:** Scikit-learn, NLTK
+* **Data Processing:** Pandas, NumPy
+* **Model Storage:** Joblib
+* **Testing:** Pytest
+* **CI/CD:** GitHub Actions
 
 ## 📂 Dataset
 
 ### Label Definition
 
-- `1` → Fake review  
-- `0` → Genuine review  
-
-Note: The model uses binary classification where 1 represents fake and 0 represents genuine.
-
-
+* `1` → Fake review
+* `0` → Genuine review
 
+The system uses binary classification where `1` represents a fake review and `0` represents a genuine review.
 
 ## 📂 Project Structure
 
@@ -103,110 +97,125 @@ Review-1/
 │   ├── notebooks/
 │   └── src/
 └── images/
+```
 
 ## ⚙️ Installation
-- pip install -r requirements.txt
 
+```bash
+pip install -r requirements.txt
+```
 
-## ✅ RUN APPLICATION
+## ✅ Run Application
 
 ```bash
 python3 run.py
+```
 
+## 🏋️ Train Models
 
-🏋️ Train Models
-## Default Training (Recommended)
+### Default Training (Recommended)
 
-- python app/ml/training/train_model.py \
---input_csv dataset/amazon_labeled_fake_reviews/final_labeled_fake_reviews.csv \
---phase1_root app \
---random_seed 42
+```bash
+python app/ml/training/train_model.py \
+  --input_csv dataset/amazon_labeled_fake_reviews/final_labeled_fake_reviews.csv \
+  --phase1_root app \
+  --random_seed 42
+```
 
+### Train Model Versions
 
-## 🔹 Train Model Versions
-- **v1 (Hybrid):** Text + metadata  
-- **v2 (Text-only):** Only text features  
-- **v3 (Best Model):** Blended model combining v1 + v2  
+* **v1 (Hybrid):** Text + metadata
+* **v2 (Text-only):** Only text features
+* **v3 (Best Model):** Blended model combining v1 + v2
 
+#### v1
+
+```bash
+python app/ml/training/train_model.py --include_behavioral --model_version phase1-v1
+```
 
+#### v2
 
-## v1
-- python app/ml/training/train_model.py --include_behavioral --model_version phase1-v1
+```bash
+python app/ml/training/train_model.py --model_version phase1-v2
+```
 
+#### v3 (Best)
 
-## v2
-- python app/ml/training/train_model.py --model_version phase1-v2
+```bash
+python app/ml/training/train_model_v3.py
+```
 
+### Optional (Advanced)
 
-## v3 (Best)
-- python app/ml/training/train_model_v3.py
+```bash
+python app/ml/training/train_model.py --enable_xgboost
+```
 
 ## 🧪 Testing
 
-- 106 test cases implemented  
-- 100% pass rate  
-- Includes:
-  - Unit Testing  
-  - Integration Testing  
-  - Security Testing  
-  - Regression Testing  
+* 106 test cases implemented
+* 100% pass rate
+* Includes:
 
+  * Unit Testing
+  * Integration Testing
+  * Security Testing
+  * Regression Testing
 
+### Run Tests
 
-## Optional (Advanced)
+```bash
+python3 -m pytest -q app/tests
+```
 
-- python app/ml/training/train_model.py --enable_xgboost
+## 📊 Model Evaluation
 
-## Run Tests
-- python3 -m pytest -q app/tests
+The system is evaluated using:
 
+* Accuracy
+* Precision
+* Recall
+* F1-Score
 
-📊 Model Evaluation
-- Accuracy (>80%)
-- Precision
-- Recall
-- F1-Score
+## ⚠️ Risk and Validation
 
+To reduce data leakage and improve reliability, the project includes:
 
-⚠️ Risk & Validation
-- To avoid data leakage:
-- Text normalization + hashing
-- Duplicate-safe splitting
-- Near-duplicate audit
-- Remaining Risk
-- Paraphrased reviews may still exist → residual risk
+* Text normalization and hashing
+* Duplicate-safe splitting
+* Near-duplicate audit
 
-## ⚙️ CI/CD
+### Remaining Risk
 
-- GitHub Actions used  
-- Automated testing on every push  
+* Paraphrased reviews may still exist in the dataset, creating residual risk
 
+## ⚙️ CI/CD
 
+* GitHub Actions used
+* Automated testing on every push
 
 ## 📊 Project Management (Jira)
 
-The project was managed using Jira with sprint planning and task tracking.
-
-
+The project was managed using **Jira** with sprint planning and task tracking.
 
 ### 🗂️ Jira Board Overview
 
 ![Jira Board](images/jira-board.png)
 
 ## 📅 Agile Milestones
-- Planning & Dataset Preparation
-- Model Development & Evaluation
-- Web Application Integration
-- Testing & Final Delivery
-
 
+* Planning and dataset preparation
+* Model development and evaluation
+* Web application integration
+* Testing and final delivery
 
 ## 👥 Team
-- Kriti Subedi
-- Swapnali Kudale
-- Aditi Sharma
-
 
+* Kriti Subedi
+* Swapnali Kudale
+* Aditi Sharma
 
 ## 📜 License
-Developed for academic purposes only.
\ No newline at end of file
+
+Developed for academic purposes only.