Fraud Call Detection

This application is designed to detect fraudulent calls by analyzing audio recordings. Upload an audio file to see if it is classified as a fraud call or a normal call. This project was developed as part of the GSLab Hackathon 2024, showcasing our innovative approach to fraud detection using speech recognition and machine learning techniques.

Table of Contents

• Installation
• Usage
• Project Structure
• Technologies Used

Installation

Prerequisites

• Python 3.8 or higher
• Docker (optional, for containerized deployment)

Steps

1. Clone the repository:

   git clone https://github.com/yourusername/fraud-call-detection.git
   cd fraud-call-detection

2. Set up a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r requirements.txt

4. Download NLTK data:

   python -m nltk.downloader stopwords wordnet

Using Docker

1. Build the Docker image:

   docker build -t fraud-call-detection .

2. Run the Docker container:

   docker run -p 5000:5000 fraud-call-detection

Usage

1. Start the Flask application:

   flask run --host=0.0.0.0

2. Open your web browser and navigate to:

   http://localhost:5000
   

3. Upload a call recording to determine whether it is fraud or not.

Technologies Used

• Flask: A micro web framework for Python.
• SpeechRecognition: Library for performing speech recognition.
• NLTK: Natural Language Toolkit for text processing.
• pandas: Data manipulation and analysis library.
• scikit-learn: Machine learning library.
• Docker: Containerization platform.