Implementation of an OpenCL-based data-mining algorithm for signal processing in healthcare applications

OData support
Supervisor:
Fekete Tamás
Department of Automation and Applied Informatics

Nowadays data-mining algorithms are used in many cases to extract new knowledge from large-scale data sets. These algorithms often run on huge data sets, so their computing complexity is big too. Executing them could use significant time. One solution to reduce the executing time is to execute the programs parallel, as much as possible. The ECG signals in the healthcare have important healing and research purpose. Systematic measurements of these signals results big data sets.

This thesis gives a solution for pattern searching in data sets with the OpenCL framework on data created by ECG measurements. Data mining and parallel computing as well as the OpenCL framework will be introduced in it too. Based on this knowledge it will present the developing process of the final application. During this process more and more pattern search related tasks with increasing complexity were implemented with the help of the OpenCL. Based on the gained experiences an application is completed, that can search parallel in data sets of ECG measurements, for a user specified pattern or patterns, which are unique for every data set, with a user specified tolerance. The searched pattern or patterns could be defined with pictures too.

The presented knowledge could be used not only in this area of application. But also in such areas where the goal is achieving big computing performance while processing big data sets too, for example other data mining areas different from pattern searching.

Downloads

Please sign in to download the files of this thesis.