Create a self-learning stock prediction system that automatically pulls live price data, runs ML models to forecast near-term movements, and tracks each model’s hit rate.
What you'll build
The Self-Learning Stock Prediction System is a sophisticated application that integrates live stock price data, machine learning models, and a user-friendly dashboard to predict near-term stock movements. The system automatically pulls live price data using yFinance, preprocesses the data using pandas, and trains LSTM models using TensorFlow/Keras to forecast stock prices. The system tracks the accuracy of each model and incorporates a feedback loop to retrain the models based on recent performance, ensuring continuous improvement. The dashboard, built using React and Tailwind CSS, displays a card for each stock showing its name, current prediction accuracy, and real-time updates. The application is containerized using Docker and deployed on Vercel/Render. The system also utilizes Celery for scheduled retraining of the models and PostgreSQL for storing stock data and model performance metrics. Chart.js is used for visualizing the stock prices and prediction accuracy. The project is designed to be scalable and maintainable, with a focus on expert-level technologies and concepts. This project has the potential to be a valuable addition to a portfolio and can be further developed into a business. Future enhancements can include integrating more data sources, improving the user interface, and exploring different machine learning models.
What you'll learn
Roadmap
5 steps · tasks unfold as you work
