Napovedni model gibanja tržnih cen s pomočjo analize sentimenta

magistrsko delo

David Pintarič (Author), Borko Bošković (Mentor), Janez Brest (Co-mentor)

Abstract

Na svetu obstajajo številni trgi, kjer lahko kupci in prodajalci trgujejo s finančnimi inštrumenti. Ker se cene na trgih neprestano spreminjajo, lahko to lastnost, ki jo imenujemo nestanovitnost, izkoristimo in, če imamo pravilno napoved, ustvarimo profit. V sklopu magistrskega dela se problema pravilne napovedi lotimo z uporabo analize sentimenta in jezikovnih tehnologij. S pomočjo objav uporabnikov na socialnem omrežju Twitter izdelamo model, ki napove gibanje tržne cene kriptovalute Bitcoin. Preizkusimo več različnih algoritmov za klasifikacijo sentimenta. Najboljše rezultate dosežemo z metodo podpornih vektorjev. Ugotovimo, da sta izdelan model in analiza sentimenta uporabna za napoved tržne cene, vendar sama po sebi nista dovolj natančna, da bi ju lahko uporabili kot edini kazalnik. Oba sta bolj primerna kot del večjega sistema za podporo pri odločanju.

Keywords

procesiranje naravnega jezika;analiza sentimenta;trgovanje;napovedni modeli;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:	[D. Pintarič]
UDC:	004.8.021(043.2)
COBISS:	139673347
Views:	132
Downloads:	35
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Prediction model for market price movement using sentiment analysis
Secondary abstract:	There are numerous markets in the world where buyers and sellers can trade financial instruments. Since prices in the markets are constantly changing, which is called volatility, we can take an advantage of it and, if we have the right prediction, make a profit. As part of the Master's thesis, we tackle the problem of correct prediction by using sentiment analysis and language technologies. Using the posts of users on the social network Twitter, we create a model that predicts the movement of the market price of the cryptocurrency Bitcoin. We test multiple different sentiment classification algorithms. The best results are achieved with a support vector machine. We find that the constructed model and sentiment analysis are useful for market price prediction, but they are not accurate enough to be used as a sole indicator. Both are better suited as part of a larger decision support system.
Secondary keywords:	natural language processing;sentiment analysis;trading;predictive model;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages:	1 spletni vir (1 datoteka PDF (X, 46 f.))
ID:	16432345