Classical

• ARIMA, exponential smoothing and seasonal decomposition
• Hierarchical and reconciled forecasts across business units
• Establishing a baseline that machine learning must genuinely beat

Machine Learning

• Gradient boosting on lagged and calendar features
• Sequence models and where they justify their complexity
• Walk-forward validation and purged cross-validation to prevent leakage
• Probabilistic forecasting: intervals, quantiles and pinball loss

Application

• Revenue and demand forecasting feeding financial planning
• Driver-based scenario modelling and stress testing
• Alternative data and text signals from filings and earnings calls

Control

• Model risk management and independent validation expectations
• Explaining forecast changes to finance stakeholders and auditors
• Monitoring accuracy and retiring models that have quietly decayed