nedelja, 28. februar 2021 Jamolbek MATTIEV: The effect of distance metric "directness" on class association rules
V ponedeljek, 1. marca 2021, bo ob 16.00 uri prek spletnih orodij na daljavo izvedeno predavanje v okviru PONEDELJKOVEGA SEMINARJA RAČUNALNIŠTVA IN INFORMATIKE Oddelkov za Informacijske znanosti in tehnologije UP FAMNIT in UP IAM.
ČAS/PROSTOR: 1. marec 2021 ob 16.00 na daljavo
----------------------------------------------
PREDAVATELJ: Jamolbek MATTIEV
----------------------------------------------
Jamolbek Mattiev obtained his PhD degree in Computer Science at UP FAMNIT. He got his Master degree in Computer Science from National University of Uzbekistan and was awarded with first degree diploma at “the best Master Dissertation Work of Uzbekistan” competition in his master studies. He also studied at Turin Polytechnic University (Italy) in his master study as an exchange student. He is doing his research in the field of Data Mining. His research fields include Artificial Intelligence, Data Mining, Machine Learning. In particular, the sub-fields of Supervised and Unsupervised Learning, Frequent Pattern Discovery, Association Rule Learning, Classification and Clustering.
----------------------------------------------------------------------------------------------------------
NASLOV: The effect of distance metric "directness" on class association rules
----------------------------------------------------------------------------------------------------------
POVZETEK:
Huge amounts of data are being collected and analyzed nowadays. By using the popular rule-learning algorithms, the number of rules discovered on those big datasets can easily exceed thousands. To produce compact and accurate classifiers, such rules have to be grouped and pruned. To group the rules, we need appropriate distance metrics, because distance metrics are very important in clustering applications. To solve the above-mentioned problem, we present new methods (use different distance metrics) that are able to reduce the number of class association rules produced by classical class association rule classifiers, while maintaining an accurate classification model that is comparable to the ones generated by state-of-the-art classification algorithms. We also present how distance metrics are affecting the clustering of class association rules.
Experimental results performed on selected datasets from the UCI ML repository show that our methods are able to learn classifiers containing significantly less rules than state-of-the-art rule learning algorithms on datasets with larger number of examples. On the other hand, classification accuracies of the newly developed methods are not significantly different from state-of-the-art rule-learners on most of the datasets.
---------------------------------------------------------------------------------------------------
Predavanje bo potekalo v angleškem jeziku prek spletnega orodja Zoom.
Do predavanja dostopate tako, da se povežete prek sledeče povezave:
https://upr-si.zoom.us/j/297328207
Vabljeni!
