Abstract

Sales forecasting in e-commerce is essential for optimizing inventory management, enhancing customer satisfaction, and supporting strategic decision-making. This study investigates sales forecasting using exploratory data analysis (EDA) and advanced feature engineering techniques. Utilizing rich data from a prominent household items retailer in Saudi Arabia, key patterns and trends influencing sales performance are identified. Several predictive models are developed and their performances are evaluated, with a particular focus on the impact of domain- specific features and state-of-the-art machine learning algorithms on forecast accuracy. The findings demonstrate that incorporating domain-specific features and advanced machine learning techniques significantly improves sales forecasting precision. This research provides valuable insights and practical methodologies for practitioners and researchers aiming to enhance their e- commerce sales forecasting capabilities. The implications of the results are discussed, providing a solid foundation for future work in the field.