prikaz prve stranice dokumenta Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka
No public access
scientific master's theses
Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka
Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing, 2000. urn:nbn:hr:168:659819
Ferišak, Marko
University of Zagreb
Faculty of Electrical Engineering and Computing
Department of Telecommunications

Institutional repository: FER Repository

Cite this document

Ferišak, M. (2000). Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka (Scientific master's theses). Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing. Retrieved from https://urn.nsk.hr/urn:nbn:hr:168:659819

Ferišak, Marko. "Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka." Scientific master's theses, University of Zagreb, Faculty of Electrical Engineering and Computing, 2000. https://urn.nsk.hr/urn:nbn:hr:168:659819

Ferišak, Marko. "Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka." Scientific master's theses, University of Zagreb, Faculty of Electrical Engineering and Computing, 2000. https://urn.nsk.hr/urn:nbn:hr:168:659819

Ferišak, M. (2000). 'Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka', Scientific master's theses, University of Zagreb, Faculty of Electrical Engineering and Computing, accessed 30 November 2024, https://urn.nsk.hr/urn:nbn:hr:168:659819

Ferišak M. Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka [Scientific master's theses]. Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing; 2000 [cited 2024 November 30] Available at: https://urn.nsk.hr/urn:nbn:hr:168:659819

M. Ferišak, "Primjena neuronskih mreža u planiranju razvoja umreženih baza podataka", Scientific master's theses, University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, 2000. Available at: https://urn.nsk.hr/urn:nbn:hr:168:659819

Metadata
TitlePrimjena neuronskih mreža u planiranju razvoja umreženih baza podataka
AuthorMarko Ferišak
MentorMladen Tkalić (mentor)
Committee memberMladen Tkalić (član povjerenstva)
Committee memberVladimir Brlić (član povjerenstva)
Committee memberZoran Skočir (član povjerenstva)
GranterUniversity of Zagreb
Faculty of Electrical Engineering and Computing
(Department of Telecommunications)
Zagreb
Defense date and country2000-02-25, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Universal decimal classification
(UDC)		004 - Computer science and technology. Computing. Data processing
Abstract
Neuralne mreže pokazale su se kao alat koji je davao dobre rezultate u područjima primjene u kojima su standardni pristupi imali poteškoća. U području telekomunikacija problemi adaptivnog usmjeravanja prometa definitivno mogu biti uspješnije rješavani primjenom logike usmjeravanja koja ima određenu "inteligenciju". Načini ostvarenja te "inteligencije" su upotreba učećih automata ili upotreba neuralnih mreža. Neuralne mreže i učeći automati se mogu kombinirati, tako da se neuralne mreže koriste kao adaptivna kritika za korekciju reakcija učećih automata. U tom slučaju radi se o specijalnom slučaju primjene neuralnih mreža za predviđanje vremenskih serija s on-line učenjem i reakcijama. Neuralne mreže se mogu implementirati hardverski ili kao softverska simulacija. Softverska simulacija razvijena u ovom radu obuhvaća definiciju topologije mreže, njeno funkcioniranje u normalnom režimu rada i učenje neuralne mreže po Back Propagation algoritmu. Područje baza podataka ima u današnje vrijeme sve veću i veću ulogu. U situaciji u kojoj se sve umrežuje i povezuje postojeće baze podataka ponekad su opasnosti da postanu izolirani "otoci" u inaće povezanim sistemima. Pri tome njihova primarna funkcija ostaje sačuvana, ali kod daljnjeg razvoja često je poželjno imati transparentan pristup do njihovih podataka. Radi toga razvijaju se skladišta podataka kao sistemi koji imaju uvid u podatke baze, ne ometaju i ne ugrožavaju njenu primarnu funkciju, a krajnjim korisnicima daju na uvid potrebne podatke. U ovom radu opisani su temeljni koncepti skladišta podataka. Ukratko su opisane metode potrebne za pristup bazama podataka i njihovo povezivanje sa ostatkom sistema. U samim skladištima podataka obično je dostupna ogromna količina podataka. Njihova organizacija se provodi na poseban način nazvan dimenzijski model podataka, a svrha upotrebe dimenzijskog modela je preglednost podataka i njihovo organiziranje na način koji ce biti razumljiv krajnjim korisnicima upoznatim s prikazanim procesom. Pokazano je da upotreba dimenzijskog modela podataka, iako se bazira na odvojenim zvjezdastim spojevima koji su još i raspodijeljeni po raznim odvojeno funkcionalnim komponentama skladišta podataka mora biti provedena tako da se poštuju usklađene dimenzije i standardne definicije činjenica, zato da bi se kasnije moglo kombinirati podatke iz raznih zvjezdastih spojeva. Jedan pristup rudarenju podataka je upotreba neuralne mreže za predviđanje vremenskih serija na osnovi povijesnih podataka o promatranom procesu dobivenih iz skladišta podataka.
Abstract (english)
Neural networks are known as a tool fit for application to areas where standard approaches experienced complications. In the telecommunications topic solutions of adaptive routing problems can certainly increase their effectiveness by the use of routing logic that possesses certain built - in "intelligence". This "intelligence" can be achieved by the use of learning automata or neural networks. Neural networks and learning automata can be combined in such a way that neural networks act as adaptive critic for correction of learning automata reactions. In such a scenario we have a special case of the application of neural networks to time - series prediction with on - line learning and reactions. Neural networks can be implemented in hardware or as a software simulation. Software simulation developed here enables the definition of network topology, its functioning in normal feed - forward mode and training by Back Propagation algorithm. The topic of databases has ever increasing role. In the situation where everything get networked and interconnected existing databases are sometimes exposed to danger of becoming isolated parts of systems. Although their primary function gets preserved it's often advantageous to have a transparent approach to their data. That's often the driving force behind the development of data warehouses as systems that have access to legacy databases but don't interfere with their primary functions, while providing end - users with a view of wanted data. In this paper basic concepts of data warehouses are described. In short lines methods necessary to access data bases and connect them to the rest of the system are described. Data warehouses usually have access to a huge amount of data. Their organization is implemented in a special way known as dimensional model. The purpose of dimensional model is to make data understandable and organize it in a way understandable to end users that are familiar with a displayed process. It's known that the use of dimensional model, although based on separate star joins, can be implemented with conformed dimensions and standard fact definitions. This approach enables later combining of data from various star joins. One of the possible approaches to data mining is the use of neural network for time - series prediction, where based on historic data form the data warehouse, information about the process of interest could be obtained.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:168:659819
Study programmeTitle: Postgraduate master programe in computing
Study programme type: university
Study level: postgraduate
Academic / professional title: magistar inženjer računarstva (magistar inženjer računarstva)
Type of resourceText
File originBorn digital
Access conditionsClosed access
Terms of useLicence In copyright icon
Created on2019-04-02 07:53:41