In English

Clustering Algorithms for Identifying Favourite Places

Jan Pauseback
Göteborg : Chalmers tekniska högskola, 2014. 102 s.
[Examensarbete på avancerad nivå]

Context: Vehicles are increasingly becoming connected devices that produce a lot of data -for example location data- which need to be processed and analysed. Cluster algorithms group objects in such a way that objects in the same cluster are more similar to each other than to those in other groups. They are essential for processing data and become especially useful when looking to identify important places from location data. Objectives: This thesis identifies and evaluates available approaches in the field of clustering algorithms for the place identification problem. Furthermore the algorithms OPTICS and DBSCAN are compared in terms of runtime performance and scalability. Methods: In the first part of the thesis a systematic literature review is used to identify and evaluate available clustering algorithms for the place finding problem. The second part of the thesis is an experiment that compares one of the found algorithms to an algorithm that is already in use at a research centre of a car manufacturer. Results: This thesis contributes in providing (1) an extensive list of clustering algorithms for identifying important places from location data, (2) an evaluation of the found algorithms and (3) a performance comparison of the algorithms OPTICS and DBSCAN. Conclusions: Most of the found approaches from the systematic literature review are density based and based on the algorithm DBSCAN. The algorithm OPTICS has the benefit of providing a hierarchical clustering structure as a result while working with the same density based approach as DBSCAN. The experiment in this thesis indicates that for geolocation data the difference in runtime performance between OPTICS and DBSCAN is considerably less than the previously reported difference.

Publikationen registrerades 2015-08-13.

CPL ID: 220572

Detta är en tjänst från Chalmers bibliotek