Forecasts

Forecasts are the core output of the demand forecasting platform. They represent AI-generated predictions of future demand based on your historical data and selected forecasting models.

Creating Forecasts

Step 1: Select Your Data

1. Navigate to the Forecasts section in your workspace
2. Click "New Forecast" to start the creation process
3. Select your main dataset containing historical demand data
4. Optionally select categorical and external datasets for enhanced accuracy

Step 2: Configure Forecast Parameters

Forecast Horizon

• Definition: How far into the future to predict
• Range: 1-52 periods depending on frequency

Forecast Frequency

• Auto-detected: Based on your main dataset frequency
• Options: Daily, weekly, monthly, quarterly, yearly
• Consideration: Should match your business planning cycles

Model Selection

Choose from available forecasting algorithms:

Traditional Methods:

• ETS (Exponential Smoothing): Good for trend and seasonality
• ARIMA: Effective for stationary time series
• Holt-Winters: Handles trend and seasonal patterns

Machine Learning:

• LSTM: Deep learning for complex patterns
• LightGBM: Gradient boosting for structured data
• Linear Regression: Simple baseline model

Advanced Neural Networks:

• N-BEATS: Neural basis expansion analysis
• N-HITS: Hierarchical time series forecasting
• PatchTST: Transformer-based forecasting
• Autoformer: Attention-based forecasting

Ensemble Methods:

• Mean Ensemble: Averages predictions from all selected models

Step 3: Start the Forecast

1. Review your configuration
2. Click "Create Forecast" to start the process
3. Monitor progress in the forecast detail page
4. Access results when processing completes

Forecast Results

Interactive Charts

The forecast results page provides comprehensive visualizations:

Time Series Chart

• Historical Data: Actual demand values (blue line)
• Forecast Lines: Predicted values for each selected model
• Item Selection: Choose specific items to analyze
• Model Toggle: Show/hide individual models for comparison

Data Export

• CSV Download: Export forecast results for external analysis

Forecast Management

Viewing Forecasts

• List View: All forecasts with status and key information
• Details View: Comprehensive forecast information
• Results View: Interactive charts and analysis

Forecast Actions

• Edit: Modify forecast name and description
• Download: Export forecast results
• Delete: Remove forecast (cannot be undone)

Status Tracking

• Not Started: Forecast created but not yet processed
• Running: Currently being processed
• Completed: Successfully generated
• Failed: Error during processing

Troubleshooting

Common Issues

Forecast Creation Failures

• Insufficient Data: Ensure minimum 12-24 periods of historical data
• Data Quality: Check for missing values or outliers
• Model Selection: Verify at least one model is selected
• System Resources: Large datasets may require more processing time

Poor Forecast Accuracy

• Data Quality: Review and clean historical data
• Model Selection: Try different algorithms for your data type
• External Factors: Include relevant external data
• Forecast Horizon: Consider shorter horizons for better accuracy

Processing Delays

• Dataset Size: Larger datasets require more processing time
• Model Complexity: Advanced models take longer to train
• System Load: High system usage may slow processing
• Network Issues: Check internet connection for API calls

---

Forecasts provide valuable insights for business planning. Regular monitoring and refinement help maintain forecast accuracy over time.