Benchmarking Time Series Forecasting Models: From Statistical Techniques to Foundation Models in Real-World Applications
This work addresses time series forecasting for operational intelligence in the hospitality industry, but it is incremental as it benchmarks existing methods on new real-world data.
This study tackled the problem of forecasting hourly sales over a 14-day horizon for a network of thousands of restaurants in Germany, evaluating statistical, ML, deep learning, and foundation models; results showed strong performance from ML-based meta-models and competitive performance from foundation models like Chronos and TimesFM with minimal feature engineering.
Time series forecasting is essential for operational intelligence in the hospitality industry, and particularly challenging in large-scale, distributed systems. This study evaluates the performance of statistical, machine learning (ML), deep learning, and foundation models in forecasting hourly sales over a 14-day horizon using real-world data from a network of thousands of restaurants across Germany. The forecasting solution includes features such as weather conditions, calendar events, and time-of-day patterns. Results demonstrate the strong performance of ML-based meta-models and highlight the emerging potential of foundation models like Chronos and TimesFM, which deliver competitive performance with minimal feature engineering, leveraging only the pre-trained model (zero-shot inference). Additionally, a hybrid PySpark-Pandas approach proves to be a robust solution for achieving horizontal scalability in large-scale deployments.