Centralizirani modeli prostorno-vremenski ponderirane i grupirane regresije

master's thesis

• Title (english): Centralized Models of Spatio-Temporal Weighted and Clustered Regression
• Author: Kristijan Čačković
• Mentor: Krešimir Pripužić (mentor)
• Committee member: Krešimir Pripužić (predsjednik povjerenstva)
• Committee member: Siniša Sovilj (član povjerenstva)
• Committee member: Jurica Babić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-09, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Cilj ovog rada je implementacija centraliziranih modela prostorno-vremenski ponderirane i grupirane regresije te optimizacija istih za rad s velikim podacima koristeći programski jezik Python. Oba razvijena modela optimizirani su te nude podršku za višeprocesni rad kako bi bolje iskoristili računalne resurse i poboljšali performanse. Razvijeni modeli temelje se na modelu linearne regresije. Algoritam linearne regresije korišten prilikom izrade ovog rada dio je programske knjižnice scikit-learn. Tijekom implementacije ranije navedenih modela korištena je programska knjižnica pandas koja je olakšala manipulaciju podacima te programska knjižnica NumPy pomoću koje su implementirani ključni matematički dijelovi modela. Razvijeni modeli eksperimentalno su evaluirani nad velikim skupom podataka. Oba modela bolje adresiraju problem lokalnih prostornih i vremenskih varijacija u podacima u odnosu na model linearne regresije. Model prostorno-vremenski grupirane regresije nudi bolje performanse u radu s velikim podacima u odnosu na model prostorno-vremenski ponderirane regresije uz neprimjetne razlike u točnosti modela.

Abstract (english)

The goal of this paper is an implementation of centralized models of spatio-temporal weighted and clustered regression and their optimization for working with big data using the Python programming language. Both developed models are optimized and offer support for multiprocessing to make better use of computing resources and improve performance. The developed models are based on the linear regression model. The linear regression algorithm used for the implementation of spatio-temporal models is part of the scikit-learn library. During the implementation of previously mentioned models, the pandas library was used to facilitate data manipulation. The NumPy library was also used to implement key mathematical parts of the models. The developed models were experimentally evaluated over a large data set. Both models better address the problem of local spatial and temporal variations in the data compared to the linear regression model. The model of spatio-temporal clustered regression offers better performance in working with big data compared to the model of spatio-temporal weighted regression with imperceptible differences in model accuracy.

• Keywords: GWR; GCR; TWR; GTWR; GTCR; linearna regresija; Python; scikit-learn; pandas; NumPy; strojno učenje; velika količina podataka; zagađenje zraka
• Keywords (english): GWR; GCR; TWR; GTWR; GTCR; linear regression; Python; scikit-learn; pandas; NumPy; machine learning; big data; air pollution
• Language: croatian
• URN:NBN: urn:nbn:hr:168:585633
• 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: 2023-01-13 14:43:28