Kratkoročno predviđanje cijena kriptovaluta upotrebom dubokog učenja

master's thesis

• Title (english): Short-term Forecasting of Cryptocurrency Prices by Deep Learning
• Author: Andrea Bernat
• Mentor: Vladimir Čeperić (mentor)
• Committee member: Vladimir Čeperić (predsjednik povjerenstva)
• Committee member: Siniša Šegvić (član povjerenstva)
• Committee member: Ante Đerek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2021-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Ovaj rad predstavlja primjenu modela povratne neuronske mreže s mogućnošću kratkoročnih i dugotrajnih ovisnosti (engl. \textit{Long Short Memmory}, LSTM) na problemu predviđanja kretanja cijena kriptovaluta. LSTM mreža je snažan model za modeliranje podataka vremenskih serija jer ima definirano unutarnje stanje kako bi se mogli pratiti već viđeni podaci. Predviđanje cijena kriptovaluta smatra se vrlo izazovnim zadatkom zbog svoje kaotične prirode. U ovome radu pristupili smo tom problemu i... More pokazali kako odabrati značajke koje utječu na promjenu kretanja cijena kriptovaluta. Detaljno smo opisali arhitekturu modela dubokog učenja i pokazali kako pristupiti odabiru hiperparametara modela. Ovaj rad mogao bi pomoći financijskim stručnjacima u analizi i predviđanju kretanja cijena kriptovaluti, no pošto je predviđanje tržišta vrlo nestabilno pa time i rizično treba provesti realan sustav ulaganja. Less

Abstract (english)

This paper presents an example of a recurrent neural network model with the possibility of short-term and long-term dependencies (Long Short Memmory, LSTM) on the problems of forecasting the price movement of cryptocurrency. The LSTM network is a powerful model for modeling time series data because it has a defined internal state so that already seen data can be tracked. Predicting the price of cryptocurrency is considered a very challenging task, due to its chaotic nature. In this paper,... More we approached this problem and showed how to choose features that affect the change in the price movement of cryptocurrency. We described in detail the architecture of the deep learning model and showed how to approach hyperparameter tunning. This paper could help financial experts in analyzing and forecasting the price movement of cryptocurrencies. However, the market forecast is very volatile and risky over time so we need to implement a realistic investment system. Less

• Keywords: kriptovalute; duboko učenje; vremenske serije; financijske odluke
• Keywords (english): crypto currency; deep learning; time series; financial decisions
• Language: croatian
• URN:NBN: urn:nbn:hr:168:574284
• 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 11:48:15