Repository of colleges and higher education institutions

Show document
A+ | A- | Help | SLO | ENG

Title:Izgradnja, vzdrževanje in optimiziranje napovednih modelov za namen diagnosticiranja zmanjševanja tveganj in lažjega odločanja : Magistrska naloga
Authors:ID Bojanec, Luka (Author)
ID Erman, Nuša (Mentor) More about this mentor... New window
Files:.pdf RAZ_Bojanec_Luka_i2020.pdf (3,83 MB)
MD5: B7B6D0DA1B45A4F23492C25B462B7EDE
 
Language:Slovenian
Work type:Master's thesis/paper
Typology:2.09 - Master's Thesis
Organization:FIŠ - Faculty of Information Studies in Novo mesto
Abstract:Magistrska naloga govori o napovednih modelih in ugotavlja kako jih optimalno izdelati in uporabljati, ter kako z njimi sprejemati odločitve s čim manj tveganja. Predstavljamo praktične primere napovednih modelov iz različnih področij. V praktičnem delu se ukvarjamo z rakom dojk. S pomočjo prostodostopne podatkovne smo naredili 17 modelov, s katerimi želimo klasificirati bezgavke kot nerakave oziroma rakave. 8 modelov je narejenih na celotni in 9 na zmanjšani bazi podatkov. Najučinkovitejša modela sta bila narejena z metodo SVM. Ugotovili smo, da napovednega modela s 100 % natančnostjo ni mogoče izdelati, da se odgovorni za sprejemanje odločitev ne bodo vedno odločali na podlagi napovednega modela, čeprav jim bo le-ta na voljo, da sprejemanje odločitev na podlagi napovednih modelov prinaša manjša tveganja kot brez njih, da napovedni modeli brez vzdrževanja in posodabljanja ščasoma izgubljajo na napovedni natančnosti in so vedno manj uporabni, ter da so metode poglobljenega učenja po mnenju strokovnjakov pri gradnji napovednih modelov učinkovitejše od statističnih metod in metod strojnega učenja.
Keywords:napovedni model, metode strojnega učenja, podatkovno rudarjenje, sprejemanje odločitev, tveganja, uporabnost napovednih modelov, rak dojk
Place of publishing:Novo mesto
Place of performance:Novo mesto
Publisher:{L. Bojanec}
Year of publishing:2020
Year of performance:2020
Number of pages:XVII, str. 145
PID:20.500.12556/ReVIS-6936 New window
COBISS.SI-ID:38649091 New window
UDC:004.8(043.2)
Note:Na ov.: Magistrska naloga : študijskega programa druge stopnje;
Publication date in ReVIS:12.11.2020
Views:2140
Downloads:136
Metadata:XML DC-XML DC-RDF
:
Copy citation
  
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.

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:12.11.2020

Secondary language

Language:English
Abstract:Master thesis talk about predictions models. We want to know how to construct and use them to reduce risk in decision making. We present predictive models from different areas. In practical part we present breast cancer disease and some predictive models, which try to solve challenges related to breast cancer. With free available database we made 17 prediction models. Prediction models try to classificate lymph nodes as benign or malign. 8 of them are made based on whole and 9 on the reduced database. Master thesis main findings are that building 100 % accurate predictive model is not possible, not all people, who are responsible for decision making will always use prediction models, even if they can, decision making with prediction model usage is better option than the other, prediction models with no maintenance and no updating will lose predictive accuracy and usefulness with time passed by, deep learning methods are better than statistical and machine learning methods when building predictive models.
Keywords:prediction model, machine learning methods, data mining, decision making, risks, rrediction model usefulness, breast cancer


Back