Daily Life Mobility of a Student: From Position Data to Human Mobility Model through Expectation Maximization Clustering

Abstract

There has been large number of research results to describe human mobility for various purposes. It has been researched that a person's mobility pattern can be predicted with the probability up to 93%, even though various factors and parameters can affect the human mobility pattern. In this research we tried to build a bridge between positioning data and human mobility pattern. Human mobility trails of a person can be presented in forms of positioning data sets. Positioning data from GPS or WPS and so on are somewhat accurate and usually in a tuple form of <time, latitude, longitude> while these form of data is barely interpreted by human perception. Humans can precept location information as street names, building names or shapes, etc. The error prone accuracy of positioning data leads a problem of clustering in order to figure out the point of frequent places for human mobility. These places and human mobility trails can be identified by clustering techniques, and we used Expectation Maximization clustering technique with the use of probability models derived from Levy Walk researches. We believe our research can be a starting point to model a human mobility pattern for further use.