prikaz prve stranice dokumenta Stabla odlučivanja i ansambl metode
  Download
PDF 15.1 MB
master's thesis
Stabla odlučivanja i ansambl metode
2023. urn:nbn:hr:126:389612
Robić, Pavao
Josip Juraj Strossmayer University of Osijek
School of Applied Mathematics and Informatics

Cite this document

Robić, P. (2023). Stabla odlučivanja i ansambl metode (Master's thesis). Osijek: Josip Juraj Strossmayer University of Osijek, School of Applied Mathematics and Informatics. Retrieved from https://urn.nsk.hr/urn:nbn:hr:126:389612

Robić, Pavao. "Stabla odlučivanja i ansambl metode." Master's thesis, Josip Juraj Strossmayer University of Osijek, School of Applied Mathematics and Informatics, 2023. https://urn.nsk.hr/urn:nbn:hr:126:389612

Robić, Pavao. "Stabla odlučivanja i ansambl metode." Master's thesis, Josip Juraj Strossmayer University of Osijek, School of Applied Mathematics and Informatics, 2023. https://urn.nsk.hr/urn:nbn:hr:126:389612

Robić, P. (2023). 'Stabla odlučivanja i ansambl metode', Master's thesis, Josip Juraj Strossmayer University of Osijek, School of Applied Mathematics and Informatics, accessed 24 December 2024, https://urn.nsk.hr/urn:nbn:hr:126:389612

Robić P. Stabla odlučivanja i ansambl metode [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, School of Applied Mathematics and Informatics; 2023 [cited 2024 December 24] Available at: https://urn.nsk.hr/urn:nbn:hr:126:389612

P. Robić, "Stabla odlučivanja i ansambl metode", Master's thesis, Josip Juraj Strossmayer University of Osijek, School of Applied Mathematics and Informatics, Osijek, 2023. Available at: https://urn.nsk.hr/urn:nbn:hr:126:389612

Please login to the repository to save this object to your list.
Metadata
TitleStabla odlučivanja i ansambl metode
Title (english)Decision Trees And Ensamble Methods
AuthorPavao Robić
MentorNenad Šuvak (mentor)
MentorNataša Šarlija (komentor)
Committee memberMirta Benšić (predsjednik povjerenstva)
Committee memberNenad Šuvak (član povjerenstva)
Committee memberNataša Šarlija (član povjerenstva)
GranterJosip Juraj Strossmayer University of Osijek
School of Applied Mathematics and Informatics
Osijek
Defense date and country2023-10-30, Croatia
Scientific / art field,
discipline and subdiscipline		NATURAL SCIENCES
Mathematics
Probability Theory and Statistics
Abstract
Stabla odlučivanja su jako oruđe u strojnom učenju i analizi podataka. Mogu se koristit i za klasifikaciju i regresiju i korisna su kad se bavimo s kompleksnijim skupovima podataka. Prednosti stabala su jako lagano objašnjiva. Moguće ih je prikazati grafički i lagana su za interpretaciju. Lako mogu raditi s kvalitativnim prediktorima bez potrebe kreiranja dummy varijable. Nažalost, nemaju isti nivo prediktivne točnosti kao neki drugi klasifikacijski/regresijski modeli. Dodatno, nisu robustna - mala promjena u podacima može voditi generiranju potpuno drugačijeg stabla. Metodom bagginga možemo poboljšati preciznost i stabilnost u modelima stabla odlučivanja. Kod bagginga, stabla nezavisno rastu na slučajnom uzorku promatranih podataka, te imaju tendenciju međusobno sličiti. Za slučajne šume, stabla rastu kao i kod bagginga, no svaka podjela podataka se radi koristeći slučajno izabran podskup prediktora, te tako dekoreliramo stabla, te dolazimo do temeljitije podjele prediktorskog prostora u odnosu na bagging. Boosting je još jedna metoda kojoj možemo poboljšati stabla, boosting kombinira mnoge slabe modele u jedan bolji tako da se fokusira na ispravljanje grešaka prethodnih modela - poput tima strušnjaka koji uče iz svojih grešaka kako bi donijeli bolje odluke.
Abstract (english)
Decision trees are a powerful tool in machine learning and data analysis. They can be used for both classification and regression and are useful when dealing with more complex data sets.The benefits of trees are very easily explained. They can be displayed graphically and are easy to interpret. They can easily work with qualitative predictors without the need to create a dummy variable. Unfortunately, they do not have the same level of predictive accuracy as some other classification/regression models. Additionally, they aren’t very robust - a small change in the data can lead to the generation of a completely different tree. Using the bagging method, we can improve precision and stability in stable decision-making models. In bagging, trees grow independently on a random sample of observed data, and tend to resemble each other. For random forests, the trees grow as in bagging, but each division of the data is done using a randomly selected subset of predictors, and thus we decorrelate the trees, and arrive at a more thorough division of the predictor space compared to bagging. Boosting is another method through which we can improve trees. Boosting combines many weak models into a strong one by focusing on fixing the errors of the previous models - like a team of experts who learn from their mistakes to make better decisions.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:126:389612
Study programmeTitle: Mathematics; specializations in: Financial and Statistical Mathematics, Mathematics and Computer Science, Industrial and Applied Mathematics
Course: Financial and Statistical Mathematics
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra matematike (magistar/magistra matematike)
Type of resourceText
File originBorn digital
Access conditionsOpen access
Terms of useLicence In copyright icon
Created on2023-11-10 10:35:03