FakeReviewML

A project to extract reviews from supported website links and analyze them to determine which reviews are computer generated and which are genuine, providing insights into review authenticity.

---

---

How to Run the Project

First, clone this repository.

Then follow these steps:

1. Start the Backend:

   cd ./backend
   python manage.py runserver

2. Start the Frontend:

   cd ./frontend
   python app.py

3. Access the Web Interface:

   • Open http://127.0.0.1:5000 in your browser.
   • Paste any Flipkart product link in the input field.
   • Wait for 1-2 minutes while the system scrapes reviews and processes them.
   • The system will display sentiments and the probability of reviews being AI-generated.

---

Steps Involved

1. Dataset Preparation

• The divide_dataset_in_parts.py script splits the dataset into multiple parts, ensuring a balance between real and fake reviews.
• The dataset is divided into training and testing sets.

2. Model Training & Hyperparameter Tuning

Each model undergoes hyperparameter tuning using GridSearchCV, followed by final training with optimized parameters.

Base Models

• SVM → Gridsearch/SVM.py (tuning) → Training/SVM.py (final training)
• Random Forest → Gridsearch/Random_forest.py → Training/Random_forest.py
• Logistic Regression → Gridsearch/Logistic_Regression.py → Training/Logistic_Regression.py

Stacking Model (XGBoost)

• Combines predictions from SVM, Random Forest, and Logistic Regression.
• XGBoost assigns weightage to each model’s predictions to improve accuracy.
• Gridsearch/XGBoost.py (tuning) → Training/XGBoost.py (final training).

3. Fake Review Detection & Sentiment Analysis

• The Scraping_Identifying.py script scrapes reviews from Flipkart.
• Uses trained models to predict authenticity of reviews.
• If a review has less than 60% authenticity probability, it undergoes sentiment analysis to understand its tone.

---

Project Structure

D:.
├───backend
│   ├───backend
│   │   └───__pycache__
│   ├───Models
│   └───process
│       ├───migrations
│       │   └───__pycache__
│       └───__pycache__
├───Dataset
├───frontend
│   ├───static
│   └───templates
├───Gridearch
├───Models
└───Training

---

Requirements

• Python 3.x
• pandas
• scikit-learn 1.2.2
• xgboost
• joblib
• BeautifulSoup
• vaderSentiment

Install the required packages using:

pip install pandas scikit-learn xgboost joblib beautifulsoup4 vaderSentiment

---

Conclusion

This project scrapes and analyzes online reviews using machine learning. It leverages ensemble learning (stacking) to improve accuracy, helping users identify authentic reviews before making purchase decisions. 🚀

---