Analiza trga kriptovalut s postopki slepega ločevanja izvorov

magistrsko delo

Jan Mikolič (Author), Aleš Holobar (Mentor)

Abstract

V magistrskem delu izvedemo analizo trga kriptovalut z metodami slepega ločevanja izvorov. Osredotočimo se na algoritma FastICA in SOBI. Preizkusimo različne vrednosti vhodnih parametrov in stroškovnih funkcij. Ugotovimo, da je algoritem SOBI s številom zakasnitev 400 primernejši, saj izkorišča časovno strukturo zgodovinskih cen kriptovalut. Na podlagi mešalnega modela kriptovalute gručimo v skupine, na katere vplivajo podobni dejavniki. Predstavimo model za napovedovanje cen kriptovalut na podlagi izračunanih neodvisnih komponent. Zaključimo z ugotovitvijo, da napovedovanje cen kriptovalut zgolj na podlagi zgodovinskih podatkov o cenah najverjetneje ni možno ne glede na napovedovalni model in predhodne transformacije.

Keywords

kriptovalute;analiza neodvisnih komponent;slepo ločevanje izvorov;napovedovanje časovnih vrst;magistrske naloge;

Data

Language:	Slovenian
Year of publishing:	2020
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[J. Mikolič]
UDC:	004.421:004.422.635(043.2)
COBISS:	23070998
Views:	1306
Downloads:	166
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Cryptocurrency Market Analysis with Blind Source Separation Algorithms
Secondary abstract:	In this master's thesis we perform cryptocurrency market analysis with blind source separation algorithms. We focus on algorithms FastICA and SOBI. Different input parameters and cost functions are tested. Algorithm SOBI with number of lags 400 proves to be the best choice as it exploits the time coherence of the cryptocurrency historical price data. Given the mixing model, we perform clustering and identify groups of cryptocurrencies which are under the influence of similar factors or sources. Further on, a forecasting model, based on calculated independent components, is presented. We conclude that cryptocurrency time series forecasting based on historical price data alone is most likely not possible, regardless of forecasting model or previous transformations used.
Secondary keywords:	cryptocurrencies;independent component analysis;blind source separation;time series forecasting;FastICA;SOBI;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages:	VIII, 49 str.
ID:	11373627