# 🇸🇳 Senegal WhatsApp Disinformation Detector
## 📌 Overview
This project detects viral political disinformation on WhatsApp in Senegal using Machine Learning. It helps fact-checking teams prioritize dangerous messages during election periods.
**Live Demo:** [Link to your Streamlit app]
**Author:** Abdelaziz Abakar Tahir - AISIP Cohort 1 | Africa AI Hub
---
## 🎯 Problem Statement
During election periods in Senegal, deepfakes and AI-generated disinformation spread rapidly on WhatsApp. Citizens cannot distinguish real content from fake. No accessible detection tool exists in French or Wolof.
**Key statistics from my research:**
- 80% of survey respondents were exposed to political deepfakes
- 73% consider the problem "very serious"
- 90% would use a detection tool if available
---
## 🚀 Solution
A Machine Learning classifier that predicts whether a political WhatsApp message has high viral potential.
| Feature | Description |
|---------|-------------|
| Input | Political WhatsApp message (text) |
| Output | Viral probability score (0-100%) |
| Model | Random Forest (F1 Score: 0.86) |
| Deployment | Streamlit web application |
---
## 📊 Dataset
- **Size:** 2,000 synthetic messages
- **Label distribution:** 30% viral, 70% normal
- **Features:** message length, word count, capitals ratio, emoji count, political keywords, call-to-action, TF-IDF (50 features)
---
## 🤖 Models Trained
| Model | F1 Score |
|-------|----------|
| Logistic Regression | 0.80 |
| Random Forest | **0.86** 🏆 |
| Gradient Boosting | 0.85 |
| Neural Network (Keras) | 0.84 |
| Ensemble Voting | 0.86 |
---
## 🛠️ Technology Stack
| Component | Technology |
|-----------|------------|
| Data Processing | Python, Pandas, NumPy |
| Visualization | Matplotlib, Seaborn |
| ML Models | Scikit-learn |
| Neural Networks | TensorFlow / Keras |
| Deployment | Streamlit |
| Version Control | Git / GitHub |
| Environment | Google Col …