Prompt

Build time series forecasting model.

<time_series_data>
- Frequency: {hourly/daily/monthly}
- Length: {number of periods}
- Seasonality: {patterns observed}
- Trend: {increasing/decreasing/stable}
</time_series_data>

<external_factors>
{holidays, events, weather, etc}
</external_factors>

<forecast_requirements>
- Horizon: {how far ahead}
- Accuracy needs: {error tolerance}
- Update frequency: {how often retrained}
</forecast_requirements>

Build forecast:
1. Data preparation
   - Missing value handling
   - Outlier treatment
   - Stationarity testing
   - Decomposition
   
2. Model selection
   - ARIMA variations
   - Exponential smoothing
   - Prophet/Neural Prophet
   - LSTM if appropriate
   
3. Feature engineering
   - Lag features
   - Rolling statistics
   - External regressors
   - Calendar features
   
4. Validation strategy
   - Time series split
   - Walk-forward analysis
   - Prediction intervals
   
Include code implementation.