PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Robustness analyses for repeated mobility surveys in outdoor advertising
Dirk Hecker, Christine Körner and Michael May
In: ICSDM 2011, June 29 2011-July 1 2011, Fuzhou, China.


A growing number of companies use mobility data in their day-to-day business. However, as the data grows older, new data has to be collected in order to keep applications up-to-date. Consequently, it is of great importance to know the impact that a different mobility sample may cause. This aspect of analysis has been largely neglected in mobility data mining research so far. In this paper we therefore analyze the robustness of performance measures with respect to a changed GPS sample in outdoor advertisement. The evaluation of outdoor advertising campaigns is a challenging application because it requires the evaluation of mobility data on a very fine spatial level. Thus, the application has a higher dependency on routes of individual test persons than classical mobility surveys. In our robustness analysis we apply bootstrapping and subsampling in order to measure the effect of a) a repeated mobility survey and b) a mobility survey of smaller size. We conduct our experiments on a real-world data set from Swiss outdoor advertising. Our results show that the effect is comparably small for a typical campaign and may be mitigated further by increasing the campaign size.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
ID Code:9057
Deposited By:Christine Körner
Deposited On:21 February 2012