Agile combination of advanced booking models for short-term railway arrival forecasting11The author would like to show gratitude to National Science and Technology Council in Taiwan for financial support (MOST 109-2221-E−507-001) and also Taiwan Railway Corporation for data collection

Accurate forecasts are essential for allocating perishable resources. As arrivals are the accumulation of reservations, utilizing reservation records to construct forecasting models is novel but challenging. This study focuses on agilely combining individual alternatives via a complexity-based selec...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Engineering applications of artificial intelligence Ročník 141
Hlavní autor: Tsai, Tsung-Hsien
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Ltd 01.02.2025
Témata:
Advanced booking models
Booking curves
Combination scheme
Data complexity
Revenue management
Data complexity
Booking curves
Advanced booking models
Revenue management
Combination scheme
ISSN:0952-1976
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Accurate forecasts are essential for allocating perishable resources. As arrivals are the accumulation of reservations, utilizing reservation records to construct forecasting models is novel but challenging. This study focuses on agilely combining individual alternatives via a complexity-based selecting procedure. Real records over a year from a railway corporation are collected to form booking curves. Eleven complexity indicators are proposed to describe the characteristics of booking curves and five advanced booking models are constructed. Then five machine learning classifiers are applied to map the relationships between complexity and model suggestion. The empirical results show promising outcomes when a booking point is distant from the service day and cautiously selecting forecasting alternatives for combination is critical to predictive accuracy. More specifically, the Support Vector Machine combination achieves a 10% performance improvement compared with the all-inclusive average method. Additionally, the proposed scheme can achieve equal or better performance compared with individual models.
ISSN:0952-1976
DOI:10.1016/j.engappai.2024.109841