Back to Projects

Stock Price Forecasting

Interactive demo of an LSTM-based deep learning model for time series forecasting. Predicts stock prices with 94% R² score using historical data and technical indicators.

LSTMDeep LearningTime SeriesYahoo Finance API
94%

R² Accuracy Score

$1.82

Mean Absolute Error

60-Day

Sequence Length

<50ms

Inference Time

The Challenge

Traditional stock analysis relies on manual technical indicator interpretation and lagging signals, often missing short-term price movements. Traders need predictive insights that account for complex temporal patterns across multiple indicators simultaneously.

The Solution

Built a 3-layer LSTM neural network trained on 60-day sequences with 15+ technical indicators. The model captures long-term dependencies and momentum patterns, providing actionable forecasts with confidence intervals for risk management.

Forecast Parameters
Select stock and prediction timeframe

Select parameters to generate stock price forecast

Technical Implementation

Model Architecture

  • • 3-layer LSTM network with dropout regularization
  • • Sequence length of 60 days for temporal pattern learning
  • • Adam optimizer with learning rate scheduling
  • • Early stopping to prevent overfitting

Features & Data

  • • Historical price data from Yahoo Finance API
  • • Technical indicators: RSI, MACD, Bollinger Bands
  • • Moving averages (7, 21, 50-day)
  • • Volume-weighted average price (VWAP)