Modelling and estimation for multivariate demand forecasting with censored data
Catalina Stefanescu
(London Business School)
26/02/08
Demand forecasting is a crucial component of revenue management systems,
particularly in the airline industry. Current practice
focuses on univariate demand forecasting, however in many industries there
is empirical evidence of correlated demand for related products. In addition,
product demand is observed at several time points over a long booking horizon,
and it may be censored due to inventory constraints so that in practice only
censored sales data are recorded. In this paper we propose two classes of models
for multivariate demand forecasting, when observations of demand at different
times during the booking horizon are correlated. We investigate model estimation
from censored sales data, using the EM algorithm and a pseudolikelihood approach.
The methodology is exemplified with the analysis of airline flight booking data.
Back to the seminar homepage
LSE Home Page |
Departmental Home Page |
Baurdoux Home Page
[Last modified: Feb. 18th 2008 by Erik Baurdoux]
