Predictive Models for Sales Forecasting: Methods and Tools

This thesis investigates the predictive models most commonly adopted by companies to forecast sales and demand, with the objective of identifying their relative strengths, weaknesses, and areas of application. The study is based on a comprehensive review of academic literature and applied case studies, examining parametric models, non-parametric approaches, and advanced hybrid solutions. Parametric models, while traditionally dominant for their interpretability and reliance on well-defined assumptions, show limitations in contexts characterized by high volatility or non-linear dynamics. Non-parametric models, including machine learning techniques, demonstrate higher flexibility and accuracy in capturing complex patterns, although often at the expense of transparency. Hybrid and advanced models, which integrate both statistical and machine learning approaches, emerge as a promising field, offering improved predictive performance, at the cost of increased complexity. The findings provide guidance for practitioners and highlight future challenges and opportunities in sales forecasting, analyzing real business cases.