Information Technology Reference
In-Depth Information
A Density-Based Clustering of Spatio-Temporal Data
Ehab Zaghlool 1 , Saleh ElKaffas 2 , and Amani Saad 1
1 College of Engineering and Technology, Arab Academy for Science,
Technology and Maritime Transport, Alexandria, Egypt
{ehab_zaghlool,amani.saad}@aast.edu
2 College of Computing and Information Technology, Arab Academy for Science,
Technology and Maritime Transport, Alexandria, Egypt
saleh.mesbah@gmail.com
Abstract. Moving objects are one of many topics that have large data sets
generated rapidly and continuously by spatial technologies. This paper focuses
on the data mining of an example of such large data sets, spatio-temporal data.
This research aims to predict future motion of moving objects regarding their
location and time of arrival. A spatio-temporal algorithm is developed and
presented which clusters sub-trajectories into similar groups taking into
consideration the time dimension; time-aware, using a density based clustering
technique. The presented algorithm partitions trajectories into smaller sub-
trajectories then groups these segments based on a density-based clustering
technique. Three different experiments are carried out, each one with a different
data set. The results of each experiment are analyzed and predictions are made
for the motion of each data set.
Keywords: Data Mining, Spatio-Temporal Data, Density Based Clustering.
1
Introduction
The quantity of data being generated daily from all sorts of different applications is
becoming a challenging issue. Wireless sensors, communication systems and all
different sorts of position tracking technologies such as GPS and RFID are flushing
complex data volumes. These technologies are continuously producing raw data, the
data captured from these devices, such as spatial data, time of movement, and even
contextual data [1]. This is exhaustive to many applications, since there is no need to
keep all this data without being able to understand it. This problem lead to an
important topic in research which is analyzing those large multidimensional data sets
to be able to obtain simplified information that could be useful when making
important decisions. The process of examining such large existing data sets stored in
databases is known as data mining.
Moving objects are one of many topics that produce data sets with large volumes,
generated rapidly and continuously by such technologies, thus lead to the creation of
what is today known as Trajectory Databases (TD) or Moving Object Databases
(MOD) [2]. A Moving Object Database (MOD) consists of spatial and temporal
information about objects whose location change over time (e.g. moving humans or
Search WWH ::




Custom Search