@misc{10481/106945,
year = {2025},
month = {9},
url = {https://hdl.handle.net/10481/106945},
abstract = {Cold-start forecasting is critical in dynamic scenarios where early-stage forecasting drives key
decisions, such as content prioritization, resource allocation, and demand estimation before
observable trends emerge. In this work, we explore the potential of multimodal forecasting
techniques for cold-start forecasting and offer insights into designing more scalable and adaptive models. In particular, we address context-driven cold-start web traffic forecasting that
includes textual content and historical web traffic of relevant web pages to generate forecasts
when no historical data is available for the target new web page. To advance research in this
area, we collect, clean, and align a high-dimensional, multimodal web traffic dataset. We
adopt a Retrieval-Augmented Generation framework, and propose the use of large language
models (LLMs) for this task. Our experiments demonstrate that the LLM-based strategy
consistently outperforms the statistical baseline across multiple forecasting horizons. The
best-performing LLM-based model reduces WRMSPE by 0.81% and WAPE by 4.5%, compared with other methods. Furthermore, LLM-based feature extraction enhances contextual
understanding, leading to greater stability in long-horizon forecasts.},
organization = {CAUL - Council of Australian University Librarians
(Open Access)},
organization = {Australian Research Council (ARC) -
(Discovery Early Career Researcher Award - DECRA DE250100032)},
organization = {Consejería de Universidad, Investigación e Innovación - Junta de Andalucía y European Union - FEDER Andalucía Program 2021-2027
(C-ING-250-UGR23)},
organization = {Ministry of Science, Innovation and Universities of Spain (MICIU) - 
(PID2023-149128NB-I00)},
organization = {Spanish Ministry of Universities y European Union - NextGenerationEU (María Zambrano Fellowship)},
publisher = {Springer},
keywords = {Web traffic analysis},
keywords = {Cold-start},
keywords = {Multimodal},
title = {Context-driven cold-start Web traffic forecasting},
doi = {10.1007/s11280-025-01359-7},
author = {Zhou, Xin and Wang, Weiqing and Buntine, Wray and Bergmeir, Christoph Norbert},
}