Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Real-time Continuous Clustering for Moving Objects and their Visualization

: Alam, M. Saif
: Russig, Benjamin; Heckel, Marcel

Dresden, 2018, 62 pp.
Dresden, Univ., Master Thesis, 2018
Master Thesis
Fraunhofer IVI ()
clustering algorithm; data analysis; visualization

Data clustering is an essential technique for empirical data analysis, and has been studied for several years. Various clustering techniques are introduced time to time for exploratory analysis of very large data set to discover useful patterns and correlations among attributes. This master thesis focuses on the problems of moving objects clustering and their visualization. In case of maintaining a cluster consists of a set of data points that moves continuously in a two-dimensional euclidean space is always costly and uncertain. This uncertainty is considered to be one of the major problems. It considers the comparative study and analysis of different clustering algorithms, worked on road net-work data for continuous clustering. K-means and Balanced Iterative Reducing and Clustering using Hierarchies (BIRCH) clustering algorithms are used for comparative study. Several visualization possibilities are implemented to represent the clustering of moving object (vehicles). It refers the study of K-means clustering that monitors a set of continuously moving objects and reevaluating K-means every time there is an object update. It imposes a heavy burden for computing the centers from scratch (on the server) and for continuously sending location updates (for the client). It explains the study of BIRCH and its performance extensively in terms of memory requirements, running time, clustering quantity, stability and scalability. BIRCH is applied on the same data set and the comparative study goes on favor of it comparing to K-means. It also includes the study of visualization of moving objects, which is always considered to be difficult in terms of proper presentation, maintenance and stability. It is because the objects are hopping around within a certain space. The study shows several differences in terms of clustering of same data using different algorithms and their visualizations.