A full-stack system for cryptocurrency price prediction with real-time training monitoring and forecasting capabilities.

Architecture Overview

Server Stack (/server)

• Core Framework: Flask 2.3 + Gunicorn
• Machine Learning: TensorFlow 2.12, Keras
• Data Processing: Pandas, NumPy, yfinance
• Real-time Communication: Flask-SocketIO 5.3, Redis 4.5
• Visualization: Matplotlib, Seaborn
• Utilities: structlog, Pydantic, Joblib
• Containerization: Docker, docker-compose

Client Stack (/client)

• Framework: Next.js 15 (App Router)
• UI Components: Shadcn UI
• Visualization: Recharts
• Styling: Tailwind CSS + CSS Modules
• API Client: Axios
• WebSocket Communication: Socket.IO Client
• Form Handling: React Hook Form
• Type Safety: TypeScript 5.3

System Features

Server Capabilities

• Model training pipeline supporting CNN, LSTM, CNN-LSTM, and LSTM-CNN architectures
• Real-time WebSocket updates during training
• Historical price data fetching from Yahoo Finance
• Model persistence and versioning
• Prediction API with confidence intervals
• Comprehensive training metrics (MAE, RMSE, R²)

Client Features

• Real-time training progress dashboard
• Interactive price prediction charts
• Model configuration interface
• Historical performance comparison
• WebSocket connection management
• Responsive, mobile-first UI
• Dark/Light theme toggle

Development Setup

Prerequisites

• Docker 24.0+ and docker-compose
• Node.js 18.x + npm 9.x
• Python 3.9+ with pip

Installation

git clone https://github.com/mhanifiisik/crypto-predict.git
cd crypto-predict

# Server setup
cd server
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Client setup
cd ../client
bun install

Running with Docker

docker-compose up --build

Environment Configuration

Server (.env)

FLASK_ENV=production
REDIS_URL=redis://redis:6379/0
MODEL_SAVE_PATH=models/

Client (.env.local)

NEXT_PUBLIC_BASE_URL=http://localhost:5000

Project Structure

├── client/                # Next.js 15 Frontend
│   ├── app/               # App router
│   │   ├── dashboard/     # Main dashboard page
│   │   │   └── page.tsx
│   │   ├── layout.tsx     # Root layout
│   │   └── not-found.tsx  # 404 page
│   ├── components/        # UI Components
│   │   ├── charts/        # Recharts components
│   │   ├── training/      # Training progress visuals
│   │   ├── widgets/       # Dashboard widgets
│   │   └── ui/            # Shadcn/ui components
│   ├── hooks/             # Custom hooks
│   ├── lib/               # Utility functions
│   │   ├── api.ts         # Axios client
│   │   └── websocket.ts   # Socket.IO client
│   ├── schemas/           # Form schemas (Zod)
│   ├── types/             # TypeScript definitions
│   ├── styles/            # Tailwind CSS
│   └── public/            # Static assets
│
├── server/                # Flask Backend
│   ├── ml_app/            # Core ML Logic
│   │   ├── config.py      # Model configurations
│   │   ├── predictor.py   # Training pipeline
│   │   ├── models/        # Model architectures (CNN, LSTM, Hybrid)
│   │   ├── data/          # Data processing
│   │   └── visualization/ # Analytical visuals
│   ├── app.py             # Flask entry point
│   ├── docker-compose.yml # Redis configuration
│   ├── Dockerfile
│   └── requirements.txt   # Python dependencies
│
└── optuna/                # Hyperparameter optimization

License

Distributed under the MIT License. See LICENSE for more information.