S strojnim učenjem podprta analiza vzorcev vektorizirane porabe električne energije

magistrsko delo

Matic Pintarič (Author), Sašo Karakatič (Mentor)

Abstract

V magistrskem delu raziskujemo smiselnost pretvorbe električne porabe hišnih naprav v večdimenzionalno vektorsko predstavitev za uporabo pri nadaljnjih raziskavah na področju analiziranja električne energije s tehnikami obdelave naravnega jezika. Slednje storimo s pomočjo postopka vdelave besed, ki ga izvedemo z umetno inteligenčnima tehnikama za obdelavo naravnega jezika Word2Vec in Doc2Vec. V empiričnem delu predlagamo metodo pretvorbe električne porabe v znakovne kategorije, pridobljene vektorje pa nadaljnje analiziramo s pomočjo treh tehnik strojnega učenja. Podrobneje se osredotočimo na iskanje električnih naprav s podobnimi vzorci porabe, določanje tipa električne naprave in napovedovanje porabe električne naprave. Raziskovanje zaključimo z odgovori na zastavljena raziskovalna vprašanja in potrditvijo pripadajočih hipotez ter glavne teze z dejstvom, da vektorizirana poraba električnih naprav zajame specifične vzorce porabe.

Keywords

električna energija;vdelava besed;algoritem Word2Vec;algoritem Doc2Vec;strojno učenje;magistrske naloge;

Data

Language:	Slovenian
Year of publishing:	2022
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[M. Pintarič]
UDC:	004.65+004.6.057.6(043.2)
COBISS:	98515971
Views:	172
Downloads:	31
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Machine learning-supported analysis of vectorized electricity consumption patterns
Secondary abstract:	In the master's thesis we investigate the feasibility of converting the electrical consumption of household appliances into a multidimensional vector representation for use in further research in the field of electricity analysis with natural language processing techniques. The latter is done with the help of the word embedding process, which is performed with artificially intelligent techniques for processing the natural language Word2Vec and Doc2Vec. In the empirical part, we propose a method of converting electrical consumption into character categories, and the obtained vectors are further analyzed with the help of three machine learning techniques. We focus in more detail on finding electrical devices with similar consumption patterns, determining the type of electrical device, and forecasting the consumption of an electrical device. We conclude the research with answers to the research questions and confirmation of the associated hypotheses and the main thesis with the fact that the vectorized consumption of electrical devices captures specific consumption patterns.
Secondary keywords:	electricity;word embedding;Word2Vec;Doc2Vec;machine learning;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Informatika in tehnologije komuniciranja
Pages:	1 spletni vir (1 datoteka PDF (XV, 107 f.))
ID:	14126424