Rekurzivna delitev modelov linearne regresije za oceno zanimivosti asociativnih pravil v različnih časovnih obdobjih

Hrovat, Goran

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Rekurzivna delitev modelov linearne regresije za oceno zanimivosti asociativnih pravil v različnih časovnih obdobjih
Authors:	ID Hrovat, Goran (Author) ID Štiglic, Gregor (Mentor) More about this mentor...
Files:	DOK_Hrovat_Goran_2018.pdf (17,46 MB) MD5: 96612E1B3371E81446C685219FA20198 PID: 20.500.12556/dkum/c4a4eb72-a398-4cdd-95bf-10088f694a8d
Language:	Slovenian
Work type:	Doctoral dissertation
Typology:	2.08 - Doctoral Dissertation
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	Zanimivosti asociativnih pravil ali pogostih množic postavk se lahko skozi čas spreminjajo. Prav tako je lahko njihova zanimivost različna za različne skupine (npr. skupine ljudi). V doktorski disertaciji je predstavljen nov algoritem za določanje zanimivosti asociativnih pravil in množic postavk v različnih časov¬nih obdobjih. Algoritem odkriva skupine pacientov, ki se glede na trend zanimivosti asociativnega pravila statistično značilno razlikujejo. Rezultat algoritma je odločitveno regresijsko drevo, katerega povezave predstavljajo različne skupine pacientov, listi pa prikazujejo grafe z zanimivostmi asociativnega pravila ali množice postavk v različnih časovnih obdobjih. Pokazali smo tudi, da podpora pogoste množice postavk vpliva na kompleksnost zgrajenega regresijskega drevesa. Za demonstracijo smo uporabili elektronske zdravstvene zapise, zbrane v obdobju 15 let, ki predstavljajo odpuste pacientov iz različnih bolnišnic v Združenih državah Amerike. Predstavljeni algoritem predstavlja v tem primeru uporabno vrednost za bolnišnice, zavarovalnice, raziskovalne in ostale institucije, saj jih lahko odkrito znanje vodi do novih spoznanj in optimizacije poslovanja.
Keywords:	podatkovno rudarjenje, mere zanimivosti, asociativna pravila, podpora odločanju, regresijsko dre-vo, linearna regresija, elektronski zdravstveni zapis
Place of publishing:	Maribor
Publisher:	[G. Hrovat]
Year of publishing:	2018
PID:	20.500.12556/DKUM-70954
UDC:	005.31:519.816(043.3)
COBISS.SI-ID:	21567254
NUK URN:	URN:SI:UM:DK:YDW2N6VZ
Publication date in DKUM:	12.07.2018
Views:	2387
Downloads:	185
Metadata:
Categories:	KTFMB - FERI
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

Licences

License:	CC BY-NC-ND 4.0, Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International

Link:	http://creativecommons.org/licenses/by-nc-nd/4.0/
Description:	The most restrictive Creative Commons license. This only allows people to download and share the work for no commercial gain and for no other purposes.
Licensing start date:	22.06.2018

Secondary language

Language:	English
Title:	Recursive partitioning of linear regression models to assess interestingness of association rules in different time periods
Abstract:	Interestingness measures of association rules and frequent itemsets can change through time. Moreover their interestingness measure can also be different for different groups (e.g. groups of people). The doctoral dissertation presents a new algorithm for assessing the interestingness of association rules and itemsets in different time periods. The algorithm discovers groups of patients which statistically significantly differ depending on the trend of the interestingness of the association rule. The result of the algorithm is a decision regression tree where branches represent different patient groups with leaves representing interestingness of the association rule or itemset in different time periods. We have also shown that support of the itemset affects the complexity of the built regression tree. For the demonstration, we used electronic health records collected over a period of 15 years, representing patient discharges from various hospitals in the United States of America. In this case, the presented algorithm shows useful value for hospitals, insurance companies, research and other institutions, where discovered knowledge may leads to new insights and business optimization.
Keywords:	data mining, interestingness measures, association rules, decision support, regression tree, linear regression, electronic health record

Comments

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back