README.md

metadata

title: DataSynthis_ML_JobTask
emoji: 📈
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.49.0
app_file: app.py
pinned: false
license: mit
allow_internet: true

Stock Price Forecasting App

This application uses three different models (ARIMA, Prophet, and LSTM) to forecast stock prices.

================================================================================

FINAL RECOMMENDATIONS

================================================================================

Based on the comprehensive evaluation:

1. BEST PERFORMING MODEL: LSTM

   • Lowest RMSE: $5.39

2. KEY FINDINGS:

   • ARIMA Model:

     • Simpler and faster to train
     • Better for short-term forecasts
     • Assumes linear relationships
     • RMSE: $28.98
     • MAPE: 11.57%

   • Prophet Model:

     • Excellent at capturing seasonality and trends
     • Handles missing data and outliers well
     • Provides uncertainty intervals
     • RMSE: $16.29
     • MAPE: 6.97%

   • LSTM Model:

     • Captures non-linear patterns
     • Better for complex time series
     • Requires more data and computation
     • RMSE: $5.39
     • MAPE: 2.06%

3. RECOMMENDATIONS:

   • For production deployment, consider ensemble methods combining all three models
   • Prophet is excellent for interpretability and trend analysis
   • LSTM performs well when sufficient training data is available
   • ARIMA provides quick baseline forecasts
   • Regularly retrain models with new data
   • Monitor prediction intervals and confidence bounds
   • Consider external factors (news, market sentiment) for better predictions

4. MODEL SELECTION GUIDE:

   • Use ARIMA for: Quick forecasts, baseline comparisons, stationary data
   • Use Prophet for: Seasonal patterns, interpretable results, business forecasts
   • Use LSTM for: Complex patterns, non-linear relationships, large datasets

5. LIMITATIONS:

   • Stock prices are inherently unpredictable
   • Past performance doesn't guarantee future results
   • Models should be used as decision support tools, not sole decision makers
   • Consider risk management and diversification strategies
   • All models assume patterns will continue into the future

Features

• Real-time stock data fetching from Yahoo Finance
• Multiple forecasting models
• Interactive visualizations
• Customizable forecast periods

Models

1. ARIMA - Traditional statistical model
2. Prophet - Facebook's time series forecasting
3. LSTM - Deep learning neural network

Usage

1. Enter a stock ticker symbol (e.g., AAPL, GOOGL)
2. Select forecast period (1-90 days)
3. Choose which model(s) to use
4. Click "Generate Forecast"

⚠️ Disclaimer: For educational purposes only. Not financial advice.