Driver-based linear regression. MAPE 4.1% vs. 6.9% baseline. Synthetic data, 24 months history + 6 months forecast. Built with Python, pandas, NumPy, scikit-learn, Snowflake.
Model MAPE
4.1%
vs. 6.9% baseline (41% improvement)
WAPE
4.5%
Revenue-weighted accuracy
Forecast Bias
+0.3%
Slight over-forecast, within tolerance
CV MAPE Range
3.8-4.6%
5-fold time series cross-validation
Actual vs. Forecast with Confidence Interval
Scenario Analysis
1.001.00
Scenario forecast comparison
Scenario impact summary
Month
Base Forecast
Scenario Forecast
Impact
Feature Importance
Coefficient weights (normalized)
Feature details
Feature
Coefficient
Importance
Type
Model Comparison
Residual distribution
Cross-validation MAPE by fold
All data are synthetic. Source: github.com/nicholasjh-work/fpna-forecasting-model