Creating a novel approach for mobile positioning based on CDR data

Abstract

User geographical positioning is important for many fields that rely on passive geolocation analytics, like targeted marketing, urban and rural transportation planning, public health, etc. A new popular type of data that is commonly used for passive mobility analysis is mobile data or the so-called Call Detail Records (CDR). The CDR events are stored by mobile operators for the primary purpose of billing. They are generated every time we use SMS, call, or internet services. CDR data events are becoming more frequent due to the lower costs of using mobile services and smartphones becoming a necessary tool in our daily life. However, CDR data has two major drawbacks: temporal and spatial uncertainties. Although the first problem is widely covered by trajectory reconstruction techniques, the second problem still remains challenging. Hence, in this thesis, we propose the usage of a new method based on the Sequential Monte Carlo algorithms called particle filtering. The particle filtering application implemented in this thesis models the trajectory movement to predict the user’s position in a given area. This method uses CDR data and solely the information related to the area of the coverage from mobile towers. Our goal is to evaluate if this nonlinear method can out-perform the existent linear methods like Switching Kalman Filter. Therefore, the model performance and the effects of the parameters on accuracy were evaluated in controlled experimental settings. Additionally, experiments were performed on a dataset from a real case study and compared with the results achieved by existing methods. Finally, the usability of the method and future work is discussed.