26th International Association for Management of Technology Conference - IAMOT 2017

Spatiotemporal data is everywhere, being gathered from different devices such as Earth Observation and GPS satellites, sensor networks and mobile gadgets. Data collected from those devices might contain valuable information about different subjects, such as weather monitoring or mobility. Among these themes, moving object trajectory data has a particular interest in this work. Moving object trajectory is an example of Big Spatial Data (BSD). It meets the classical three features of Big Data: Volume, Velocity and Variety. Trajectory data sets are quickly becoming available for a larger collection of vehicles due to rapid proliferation of cell phones, in-vehicle navigation devices and other GPS data-logging tools. Such data sets are collected from different kinds of sensors with distinct spatial and temporal resolutions. In order to process this kind of data, there is a need for high-level programming environments that allow users to access big trajectory data sets and to develop new algorithms to analyze them. R is a software tool widely used for data analysis. It provides a broad variety of statistical methods (time-series analysis, classification and clustering) and a high-level programming environment and language suitable for fast developing of new algorithms. Although there are many packages for spatial and spatiotemporal data analysis, there are few R packages that work with trajectory data. In this work, we propose a framework that extends the R environment for big trajectory data handling. We present an R package that can access big trajectory data from different types of sources. In this work, we present existing tools for trajectory analysis, highlight their advantages and disadvantages and point out the need for a high-level programming environment that allows users to access big trajectory data sets and develop new algorithms over them. We implemented an R package for accessing big trajectory data by parts from different types of sources, as a solution to work in memory limited environments. Finally, we demonstrate this framework in a case study.