Crypto Prediction
is an automated system designed to fetch cryptocurrency market data, preprocess it, and make predictions using a trained machine learning model. The pipeline operates on a scheduled basis and can be triggered manually via GitHub Actions.
Key Features
1. Automated Data Retrieval The system fetches historical market data for selected cryptocurrency pairs (e.g., BTC-USDT). Supports different time intervals (e.g., 1-hour, 4-hour, daily) to allow for flexible analysis. Data is pulled from an exchange API (e.g., KuCoin) to ensure real-time and historical accuracy.
2. Data Preprocessing & Feature Engineering Cleans and structures raw market data before feeding it into the model. Calculates important technical indicators, such as: Relative Strength Index (RSI) – Measures momentum and potential overbought/oversold conditions. Moving Average Convergence Divergence (MACD) – Captures trend strength and direction. Volume Analysis – Assesses trading volume trends for additional predictive power. Applies normalization (e.g., MinMaxScaler) to standardize input features, improving model accuracy.
3. Machine Learning-Based Predictions Uses a deep learning model (built with TensorFlow/Keras) trained on historical price movements and technical indicators. The model learns patterns in past price action and attempts to forecast short-term trends. Prediction results are outputted for further analysis or integration into trading strategies.
4. Database Integration Processed market data and prediction outputs are stored in a structured MySQL database. Ensures easy access for further analysis or visualization in external tools. Database credentials are securely passed via environment variables.
5. Automated Execution with GitHub Actions The entire pipeline can be triggered automatically via Cron shedule or run manually. Runs in a controlled cloud environment (Ubuntu-based), ensuring consistency. Dependencies (e.g., TensorFlow, scikit-learn, pandas) are installed dynamically.
Workflow Overview
1.Fetch Market Data – Retrieves historical price and volume data for the selected cryptocurrency pair.
2.Preprocess Data – Cleans, transforms, and applies technical indicators.
3.Run Machine Learning Model – Predicts price trends using a trained deep learning model.
4.Store Results – Saves outputs in a local database for further analysis.
🚨Disclaimer
This project is purely a hobby experiment in machine learning and cryptocurrency market analysis. The predictions generated by the model are based on historical data and technical indicators, but they are not financial advice. The model's accuracy is not guaranteed, and cryptocurrency markets are highly volatile. Users should conduct their own research and not rely on this tool for investment decisions. 🚀