Probabilistic forecasting of train ticket booking demand with hierarchical correlations

利用订票系统的数据预测乘客数量对铁路公司管理收入和分配资源至关重要。预订需求数据具有独特的层次关联结构，无法用传统的时间序列方法适当地建模。因此，本项目旨在开发列车网络订票需求的最先进的概率预测模型。该项目将把分层数据相关性、训练能力和取消影响纳入模型，以提高预测准确性。与ExPretio合作，该项目的研究成果将有利于客运公司增加收入和减少资源浪费。此外，乘客出行需求也可以更好地与运输能力的优化配置相匹配。

Forecasting the number of passengers using the data from ticket booking systems is critical for railway companies to manage revenue and allocate resources. The booking demand data have a unique hierarchical correlation structure that cannot be appropriately modeled by conventional time-series approaches. Therefore, this project aims to develop state-of-the-art probabilistic forecasting models for booking demand in train networks. The project will incorporate the hierarchical data correlations, train capacity, and the impact of cancellation into the model to improve forecasting accuracy. In partnership with ExPretio, the research outcomes of this project will benefit passenger transport companies in increasing revenue and reducing waste of resources. In addition, passenger travel demand can also be better matched with the optimized allocation of transport capacities.