Korektor meteorološkog prognostičkog niza temeljen na strojnom učenju

master's thesis

• Title (english): Weather forecast sequence corrector based on machine learning
• Author: Danijel Bošnjak
• Mentor: Mario Vašak (mentor)
• Committee member: Mario Vašak (predsjednik povjerenstva)
• Committee member: Branimir Novoselnik (član povjerenstva)
• Committee member: Anita Banjac (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-09-22, Croatia

Abstract

Ovaj rad istražuje i definira korektor za predviđanje vremenskih varijabli. U pitanju su vremenske varijable temperatura, direktno i difuzno zračenje. Objašnjavaju se izvori podataka, a potom i pretprocesiranje istih podataka, što uključuje uzorkovanje, uklanjanje nepostojećih vrijednosti i spajanje podataka. Zatim su predstavljeni modeli korišteni u eksperimentu, kao i metrike procjene (model je ocijenjen s R-kvadratom i RMSE). Na kraju se modeli treniraju i testiraju, a nakon toga uspoređuju. Pronađen je model s najboljom metrikom procjene i dobili smo uvid u to zašto su neki modeli imali bolje rezultate od drugih.

Abstract (english)

This thesis explores and defines a corrector for predicting weather variables. Weather variables in question are temperature, direct and diffuse irradiance. Data sources are explained, and then preprocessing of the same data, which includes sampling, removing missing values and merging data. Then, models used for the experiment were introduced, as well as evaluation metrics (model was evaluated with R-squared and RMSE). Finally, models are trained and tested, and after that, compared. Model with the best evaluation metrics was found and we got an insight in why some models were performing better than others.

• Keywords: strojno učenje; model; linearna regresija; LightGBM; neuronska mreza; set za treniranje i testiranje; srednja kvadratna greška; skaliranje
• Keywords (english): machine learning; model; linear regression; LightGBM; neural network; training and test set; root mean square error; scaling
• Language: english
• URN:NBN: urn:nbn:hr:168:924587
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-10-11 08:30:28