Bitcoin Price Predictor🔍 Overview
This application uses machine learning models (Random Forest and GRU) to forecast Bitcoin prices across short- and long-term horizons. Tree-based models handle near-term fluctuations, while GRUs capture long-term patterns.
🧠 Learning & Tech Stack
This project helped me apply machine learning to financial time series, combining traditional and deep learning models in a production-ready pipeline.
- 💻 Python – data processing, modeling, and automation
- 📊 Scikit-learn & Keras – for training and experimenting with multiples ML and DL models (RF, XGBoost, LSTM, GRU)
- 📈 Feature Engineering – lagged prices, technical indicators (MACD, EMA, Bollinger Bands, etc.)
- 🌐 Multi-source Data – macroeconomic, on-chain, and sentiment features
- 🧪 Model Evaluation – test sets, rolling predictions, performance metrics
- 📤 Streamlit Deployment – interactive prediction interface
- 🗂️ GitHub & Versioning – code management and reproducibility
✨ Features
- 📆 Predict BTC prices for both short- and long-term horizons
- 🛠️ Toggle between GRU and Random Forest models
- 📈 Visual forecast charts with actual vs. predicted prices
- 📊 Feature importance breakdown (Random Forest)
- 🔁 Recursive prediction logic with simulated lag features
- 💬 Streamlit app for input, explanation, and download
🧠 Model Architecture Preview
Below is a preview of the model training and prediction interface:
