marylegrand/Community-Future-Prediction-Sid

A deep‐learning toolkit to model and predict how diverse communities (African, Asian, Hispanic, Indigenous, Middle Eastern, European) appear across News, Social Media, and Entertainment (TV & Movies). # 📊 Forecasting Community Representation in Media A deep‐learning toolkit to model and predict how diverse communities (African, Asian, Hispanic, Indigenous, Middle Eastern, European) appear across **News**, **Social Media**, and **Entertainment** (TV & Movies). 🔗 **Live Demo**: sidmediadr.diversityatlas.io ## 🏆 Highlights * **Per‑Community LSTM Models** - Separate sequence models for each community, capturing long‑term trends in representation (2004–2024). * **Multi‑Domain Forecasts** 1. 📰 **News Channels** (BBC, CNN, Fox News, Al Jazeera, ABC, …) 2. 💬 **Social Media** (Twitter, Facebook, Instagram, YouTube, Reddit) 3. 🎬 **Entertainment** (Hollywood, Bollywood, K‑Dramas, Spanish TV, Nigerian Cinema) * **Interactive Web UI** - Flask with Jinja + Chart.js streams a responsive interface: * Dropdowns for community & platform * Textarea for entering 20 years of past data * Dynamic 2025–2035 line charts & plain‑English insights * **Benchmarking Accuracy** - LSTM vs ARIMA vs Prophet, measured by AMAPE & R² — achieving sub‑0.2 AMAPE and >0.95 R² in most domains. --- ## 📂 Repository Structure ``` ├── Deployment.py # Flask app entrypoint & model loader ├── templates/ # HTML templates (Flask_UI.html) ├── static/ # JS, CSS, Chart.js assets ├── models/ # Pretrained .keras LSTM files by community ├── data/ # Sample CSVs & preprocessing scripts ├── tests/ # Unit and integration tests ├── README.md # (You are here) └── requirements.txt # Python dependencies ``` --- ## 🚀 Quickstart 1. **Clone & install** ```bash git clone github.com cd Community-Future-Prediction-Sid python3 -m venv venv source venv/bin/activate pip install -r requirements.txt ``` 2. **Deploy on EC2** - **Instance**: Ubuntu 22.04 LTS on t3.medium - **Setup**: ```bash # Pull repo, create venv, install deps git pull origin main source v …