Author Information
Boyang Zhang, University of Turku, FinlandAbstract
Restaurant sales forecasting in data science education is significant for decision making, human resource management, and inventory control. This paper proposes a decision-centric framework which includes context layer, data layer, model layer, and decision layers for the design and implementation of context-aware forecasting model. The Kaggle's Restaurant Sales Report 2024–2025 dataset is used in this research, it provides comprehensive daily sales records across multiple restaurant categories, enriched with contextual variables such as weather, promotions, special events, and pricing information. The methodology of this paper implements a structured analytical pipeline consisting of data cleaning, feature construction, time-series decomposition (trend and seasonality), and exploratory regression analysis using contextual variables such as weather, promotions, events, and prices. These steps are mapped to the proposed context, data, model, and decision layers to demonstrate how analytical outputs inform forecasting-related decisions. The findings of this research suggest that the proposed framework is suitable for forecasting the restaurant sales to practical decision making, by explicitly various layers, the framework helps to understand how data preprocessing, feature construction and modeling choices affect downstream outcomes, the prediction of the sales for the future. The framework supports the implementation of forecasting models that are transparent, reproducible and adaptable to various restaurant settings.
Paper Information
Conference: ECE2026Stream: Design
This paper is part of the ECE2026 Conference Proceedings (View)
Full Paper
View / Download the full paper in a new tab/window
Comments
Powered by WP LinkPress