A Deep Learning System for Multi-Interval Financial Market Forecasting
This project is a full-stack machine learning application that predicts future prices of financial assets such as stocks, cryptocurrencies, forex pairs, and commodities using LSTM (Long Short-Term Memory) neural networks. The system integrates real-time data ingestion, technical analysis, risk analytics, and interactive visualization into a unified Streamlit dashboard.
- ✔ Develop a robust LSTM model for multi-step financial time-series forecasting
- ✔ Support multiple assets and intervals (1h, 4h, 1d)
- ✔ Incorporate technical indicators for more stable predictions
- ✔ Provide real-time market risk analysis (Volatility, VaR, Sharpe Ratio, Drawdown)
- ✔ Build a user-friendly dashboard for traders, analysts, and researchers
- ✔ Maintain a modular, scalable codebase suitable for deployment and extension
- ✔ Cache data efficiently using SQLite to avoid redundant API calls
- ✔ Enable live visualization of past data, predictions, and confidence intervals
-
Pulls the latest historical OHLCV data using yfinance
-
Automatically updates based on the chosen asset and interval
-
Uses SQLite caching to reduce network calls
-
Trains a dedicated LSTM model per asset + time interval
-
Predicts the next price and provides confidence bounds
-
Computes evaluation metrics:
-
R² Score
-
MAE
-
RMSE
-
Integrated indicators include:
-
SMA / EMA (Simple/ Exponential Moving Average)
-
RSI (Relative Strength Index)
-
MACD ((Moving Average Convergence Divergence)
-
Bollinger Bands
-
Volatility
-
Daily Returns
These features improve the model’s predictive capability beyond raw prices.
Real-time risk evaluation: Volatility (%), Sharpe Ratio, Maximum Drawdown, Value at Risk (VaR 95%)
-
Built using Plotly + Streamlit, including:
-
Historical price chart
-
Future price predictions
-
Confidence bands
-
Indicator overlays
-
Real-time metric cards

