Modeling Censored Mobility Demand through Quantile Regression Neural Networks
share
Shared mobility services require accurate demand models for effective service planning. On one hand, modeling the full probability distribution of demand is advantageous, because the full uncertainty structure preserves valuable information for decision making. On the other hand, demand is often observed through usage of the service itself, so that the observations are censored, as they are inherently limited by available supply. Since the 1980s, various works on Censored Quantile Regression models have shown them to perform well under such conditions, and in the last two decades, several works have proposed to implement them flexibly through Neural Networks (CQRNN). However, apparently no works have yet applied CQRNN in the Transport domain. We address this gap by applying CQRNN to datasets from two shared mobility providers in the Copenhagen metropolitan area in Denmark, as well as common synthetic baseline datasets. The results show that CQRNN can estimate the intended distributions better than both censorship-unaware models and parametric censored models.
READ FULL TEXT