prikaz prve stranice dokumenta Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina
No public access
undergraduate thesis
Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina
Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing, 2021. urn:nbn:hr:168:738462
Rajnović, Marko
University of Zagreb
Faculty of Electrical Engineering and Computing

Institutional repository: FER Repository

Cite this document

Rajnović, M. (2021). Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina (Undergraduate thesis). Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing. Retrieved from https://urn.nsk.hr/urn:nbn:hr:168:738462

Rajnović, Marko. "Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina." Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2021. https://urn.nsk.hr/urn:nbn:hr:168:738462

Rajnović, Marko. "Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina." Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, 2021. https://urn.nsk.hr/urn:nbn:hr:168:738462

Rajnović, M. (2021). 'Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina', Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, accessed 14 January 2025, https://urn.nsk.hr/urn:nbn:hr:168:738462

Rajnović M. Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina [Undergraduate thesis]. Zagreb: University of Zagreb, Faculty of Electrical Engineering and Computing; 2021 [cited 2025 January 14] Available at: https://urn.nsk.hr/urn:nbn:hr:168:738462

M. Rajnović, "Modeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina", Undergraduate thesis, University of Zagreb, Faculty of Electrical Engineering and Computing, Zagreb, 2021. Available at: https://urn.nsk.hr/urn:nbn:hr:168:738462

Metadata
TitleModeli obrade prirodnog jezika za predviđanje i analizu cijena nekretnina
Title (english)Natural Language Processing Models for Real Estate Price Prediction and Analysis
AuthorMarko Rajnović
MentorJan Šnajder (mentor)
Committee memberJan Šnajder (predsjednik povjerenstva)
Committee memberMartin Tutek (član povjerenstva)
Committee memberZoran Medić (član povjerenstva)
GranterUniversity of Zagreb
Faculty of Electrical Engineering and Computing
Zagreb
Defense date and country2021-07-05, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
Zadatak ovog završnog rada bio je skupljanje podataka o reklamama sa stranice Njuškalo, pretprocesiranje podataka te predvid̄anje cijene. Poveći broj reklama je skinut sa stranice, podaci su skupljeni i očišćeni te pripremljeni za strojno učenje. Sveukupno, implementirana su tri modela: regresija potpornih vektora, pojačavanje gradijenta i umjetna neuronska mreža. Svaki od ova tri modela imali su tri varijante: varijantu bez teksta, samo tekst i kombiniranu varijantu. Modeli i varijante imali su vrlo slične rezultate. Na kraju, nisu postojale statistički signifikantne razlike izmed̄u varijanti bez teksta i kombiniranih varijanti za regresiju potpornih vektora i pojačavanje gradijenta, dok je kombinirani model bio bolji kod umjetnih neuronskih mreža. Izabrani model je bio pojačavanje gradijenata, koristeći varijantu bez teksta, jer mu je trebalo najmanje vremena za treniranje i jer se pokazao statistički najboljim za ovaj zadatak.
Abstract (english)
The task of this thesis was to gather data about advertisements from Njuškalo, preprocess it, then try and do price prediction. A sizable number of advertisements were scraped, the data gathered was cleaned, and prepared for machine learning. In total, three models were implemented: support vector regression, gradient boosting, and a feed-forward neural network. Each of these models had three variants: non-text variant, with only numeric and categorical features present, text-variant, with only the descriptions present, and the combined variant, joining the two last variants together. The models and variants had very similar results. Ultimately, there was no statistically significant difference observed between the non-text and combined variants for support vector regression and gradient boosting models. The combined variant of the feed-forward network model performed better than the non-text one. The model chosen as the best one was gradient boosting, using the non-text variant, as it took the least amount of time to train and it was shown that it was statistically the best model for this task.
Keywords
Keywords (english)
Languageenglish
URN:NBNurn:nbn:hr:168:738462
Study programmeTitle: Computing
Study programme type: university
Study level: undergraduate
Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsClosed access
Terms of useLicence In copyright icon
Public note
Created on2023-01-14 21:31:31