Optimizacija Sharpeovog omjera portfelja korištenjem modela dubokog učenja

master's thesis

• Title (english): Optimizing portfolio sharpe ratio using deep learning models
• Author: Luka Lukačević
• Mentor: Zvonko Kostanjčar (mentor)
• Committee member: Zvonko Kostanjčar (predsjednik povjerenstva)
• Committee member: Tomislav Petković (član povjerenstva)
• Committee member: Stjepan Begušić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2025-02-27, Croatia

Abstract

Ovaj rad predstavlja problem optimizacije portfelja, te rješenje tog problema kroz tra-
dicionalne pristupe i pristup temeljen na dubokom učenju. Najprije je rješenje problema
optimizacije portfelja predstavljeno kroz dvije temeljne metode: mean-variance optimi-
zacija, temeljena na Markowitzevom okviru moderne teorije portfelja, i maksimalna di-
verzifikacija, koja ima za cilj konstruirati portfelje s najvećim omjerom diverzifikacije.
Nakon detaljnog uvida u ove strategije, istražena je primjena dubokog učenja u optimi-
zaciji portfelja, usredotočujući se na LSTM mreže. Model temeljen na LSTM-u uči iz
povijesnih podataka kako bi ostvario optimalne udjele imovina unutar portfelja, s ciljem
maksimizacije Sharpeovog omjera. Implementirani model ispitan je na povijesnim tržiš-
nim podacima, uspoređujući njegovu izvedbu s referentnim pristupima mean-variance i
maksimalne diverzifikacije.

Abstract (english)

This paper presents the problem of portfolio optimization, and the solution to that
problem through traditional approaches and an approach based on deep learning. First
is the solution to the problem portfolio optimization presented through two basic meth-
ods: mean-variance optimization, based on Markowitz’s framework of modern portfo-
lio theory, and maximum diversification, which aims to construct portfolios with the
highest diversification ratio. After a detailed examination of these strategies, the appli-
cation of deep learning in portfolio optimization was explored, with a focus on LSTM
networks. An LSTM-based model learns from historical data to determine optimal asset
allocations within the portfolio, aiming to maximize the Sharpe ratio. The implemented
model was tested on the historical market data, comparing its performance with refer-
ence approaches of mean-variance and maximum diversification.

• Keywords: optimizacija portfelja; mean-variance optimizacija; maksimalna diverzifikacija; duboko učenje; LSTM
• Keywords (english): portfolio optimization; mean-variance optimization; maximum diversification; deep learning; LSTM
• Language: croatian
• URN:NBN: urn:nbn:hr:168:300317
• 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: Open access
• Created on: 2025-03-12 11:12:43