| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

Show document Help

Title:Razvoj metode za izbiro klasifikatorja
Authors:ID Černezel, Aleš (Author)
ID Rozman, Ivan (Mentor) More about this mentor... New window
Files:.pdf DOK_Cernezel_Ales_2016.pdf (3,04 MB)
MD5: 4295DF0891474438F47F47C2E14346AD
 
Language:Slovenian
Work type:Dissertation
Typology:2.08 - Doctoral Dissertation
Organization:FERI - Faculty of Electrical Engineering and Computer Science
Abstract:V doktorski nalogi opišemo razvoj metode za izbiro klasifikatorja. Glavni prispevek omenjene metode je izbor najustreznejših kombinacij: metode za merjenje točnosti, klasifikacijskega algoritma in velikostjo učne množice; v okviru uporabniško definiranih kriterijev. Metoda je splošna in posledično tudi prilagodljiva ter razširljiva. Postopek izvajanja je formalno zapisan v obliki psevdokoda. Za potrebe zagotavljanja teoretične podlage izvedemo tudi več empiričnih raziskav, kjer dobljene rezultate analiziramo s serijo statističnih preizkusov. Izsledki raziskav doprinesejo naslednje prispevke k znanosti. Formalizacija odločitev in kriterijev za izbiro najustreznejše metode za merjenje točnosti. Formalizacija odločitev in kriterijev za izbiro najustreznejšega klasifikacijskega algoritma. Izbor matematičnega modela, ki v splošnem najbolje opiše obliko učnih krivulj. Formalizacija terminalnih kriterijev za določanje najustreznejše velikosti učne množice.
Keywords:Strojno učenje, klasifikacija, klasifikacijski algoritmi, podatkovne zbirke, primerjava algoritmov, zmogljivost klasifikacije, navzkrižna validacija, metoda bootstrap, učna krivulja, prileganje krivulj, aproksimacija krivulj, potenčni zakon, eksponentni zakon, terminalni kriteriji
Place of publishing:Maribor
Publisher:[A. Černezel]
Year of publishing:2016
PID:20.500.12556/DKUM-60794 New window
UDC:[004.8.021+519.23]:004.65(043.3)
COBISS.SI-ID:19688214 New window
NUK URN:URN:SI:UM:DK:DTRC5NAC
Publication date in DKUM:21.07.2016
Views:2093
Downloads:279
Metadata:XML DC-XML DC-RDF
Categories:KTFMB - FERI
:
Copy citation
  
Average score:(0 votes)
Your score:Voting is allowed only for logged in users.
Share:Bookmark and Share



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

Secondary language

Language:English
Title:Development of a classifier selection method
Abstract:In this dissertation we present the development of a classifier selection method. The main contribution of the method is to obtain the most appropriate combinations of: method for measuring accuracy, classification algorithm, and size of the training set --- all in accordance with user-defined criteria. The method is general and therefore adjustable and expandable. The method's procedure is formally defined in the form of pseudo-code. For the purpose of providing theoretical background, several experiments were conducted and their results were analysed with a series of statistical tests. Results of the research yielded the following contributions to science. Formalising decisions and criteria for choosing the most appropriate method for measuring accuracy. Formalising decisions and criteria for choosing the most appropriate classification algorithm. Selecting a best-fit learning curve model. Formalising terminal criteria for selecting the most appropriate train set size.
Keywords:Machine learning, classification, classification algorithm, datasets, algorithm comparison, classification performance, cross-validation, bootstrap, learning curve, curve fitting, curve approximation, power law, exponential law, terminal criteria


Comments

Leave comment

You must log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back
Logos of partners University of Maribor University of Ljubljana University of Primorska University of Nova Gorica