Abstract
Recently expectations for the utilization of location-based services have been increasing. However, their practical use has been limited because of the difficulty of privacy protection. Although population statistics derived from operational data of mobile phone network have also attracted much attention, essential information about human mobility is eliminated by the aggregation process. Under these circumstances, this paper proposes a novel method to produce pseudo-data on human mobility and activity by integrating non-identifiable statistics. This data is composed of numerous virtual agents generated by a stochastic model and fulfills the given statistical features. The performance of the proposed model was examined by applying it to a real data set and comparing the results with Mobile Spatial Statistics which is estimated population using mobile phone data. The results show the hourly population estimated by the proposed model has good reproducibility for any age-groups and sex.
