Return to search

Data Mining On Architecture Simulation

Data mining is the process of extracting patterns from huge data. One of the branches
in data mining is mining sequence data and here the data can be viewed as a
sequence of events and each event has an associated time of occurrence. Sequence
data is modelled using episodes and events are included in episodes.
The aim of this thesis work is analysing architecture simulation output data by
applying episode mining techniques, showing the previously known relationships
between the events in architecture and providing an environment to predict the
performance of a program in an architecture before executing the codes. One of the
most important points here is the application area of episode mining techniques.
Architecture simulation data is a new domain to apply these techniques and by using
the results of these techniques making predictions about the performance of
programs in an architecture before execution can be considered as a new approach.
For this purpose, by implementing three episode mining techniques which are
WINEPI approach, non-overlapping occurrence based approach and MINEPI
approach a data mining tool has been developed. This tool has three main
components. These are data pre-processor, episode miner and output analyser.

Identiferoai:union.ndltd.org:METU/oai:etd.lib.metu.edu.tr:http://etd.lib.metu.edu.tr/upload/2/12611635/index.pdf
Date01 March 2010
CreatorsMaden, Engin
ContributorsSenkul, Pinar
PublisherMETU
Source SetsMiddle East Technical Univ.
LanguageEnglish
Detected LanguageEnglish
TypeM.S. Thesis
Formattext/pdf
RightsTo liberate the content for METU campus

Page generated in 0.0022 seconds