Position Mining Algorithm Based on Massive Trajectory Data

Abstract

The important position refers to the main sites of activities in the daily life of people, such as the residence and workplace. The log automatically generated by the mobile phones throughaccess to the base station is also an important data source for the user behavior pattern mining, for example, the identification of the important position. However, the relevant work is faced with many challenges, including the massive volume of the trajectory data, the low precision of the positionand the diversity of the mobile phone users. For this purpose, a general solution framework is put forward to improve the availability of the trajectory data. The framework includes a filter module that is based on state, which has improved the availability of data, as well as an important position mining module. Based on this framework, two kinds of distributed mining algorithms have been designed: FQM and AXM, and optimization is carried out from three aspects: (1) Make use of the fusion technology for the multivariate data, so as to improve the precision of the results. (2) The identification algorithm for the non-working population is put forward; (3) The identification algorithm for the night working population is put forward; and the theoretical analysis and experimental results show that, the proposed algorithms are of relatively high execution efficiency and scalability, with higher precision.