Primerjava modelov LSTM, CNN in Transformer ter njihovih kombinacij pri napovedovanju časovnih vrst

| | SLO | ENG | Cookies and privacy

Bigger font | Smaller font

First page > Show document

Show document

Title:	Primerjava modelov LSTM, CNN in Transformer ter njihovih kombinacij pri napovedovanju časovnih vrst
Authors:	ID Koren, Tilen (Author) ID Strnad, Damjan (Mentor) More about this mentor... ID Kohek, Štefan (Comentor)
Files:	MAG_Koren_Tilen_2025.pdf (1,26 MB) MD5: DBF75ABCC4444A8180B0553808C3DC68
Language:	Slovenian
Work type:	Master's thesis/paper
Typology:	2.09 - Master's Thesis
Organization:	FERI - Faculty of Electrical Engineering and Computer Science
Abstract:	V magistrskem delu smo primerjali izbrane arhitekture nevronskih mrež (LSTM, CNN, časovni Transformer in hibridne CNN-LSTM) pri napovedovanju časovnih vrst s posamezno spremenljivko in z več spremenljivkami ter pri različnih časovnih horizontih. Rezultati kažejo, da TCN (različica CNN) in LSTM največkrat dosežeta najnižje vrednosti napak. TCN se je izkazal kot najboljša izbira, saj pri zelo majhnem številu parametrov dosega rezultate, primerljive z večjimi modeli.
Keywords:	Napovedovanje časovnih vrst, Nevronske mreže, Mreža z dolgim kratkoročnim spominom, Časovna konvolucijska mreža, Transformer
Place of publishing:	Maribor
Publisher:	[T. Koren]
Year of publishing:	2025
PID:	20.500.12556/DKUM-92528
UDC:	004.8.032.26:519.2(043.2)
COBISS.SI-ID:	244006147
Publication date in DKUM:	10.07.2025
Views:	354
Downloads:	53
Metadata:
Categories:	KTFMB - FERI
:	Copy citation

Average score:	(0 votes)
Your score:	Voting is allowed only for logged in users.
Share:

Hover the mouse pointer over a document title to show the abstract or click on the title to get all document metadata.

License:	CC BY 4.0, Creative Commons Attribution 4.0 International

Link:	http://creativecommons.org/licenses/by/4.0/
Description:	This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.
Licensing start date:	17.04.2025

Language:	English
Title:	Comparison of LSTM, CNN and Transformer models as well as their combinations at time series forecasting
Abstract:	In this master's thesis, we compared selected neural network architectures (LSTM, CNN, temporal Transformer, and hybrid CNN-LSTM) for forecasting time series with a single variable and with multiple variables and across different forecasting horizons. The results indicate that TCN (a variant of CNN) and LSTM frequently achieve the lowest error values. TCN proved to be the best choice, as it delivers performance comparable to that of larger models while using a very small number of parameters.
Keywords:	Time series forecasting, Neural networks, Long short-term memory, Temporal convolutional network, Transformer

Leave comment

You must log in to leave a comment.

Comments (0)

0 - 0 / 0

There are no comments!

Back