Model dubokog učenja za upravljanje novcem

master's thesis

• Title (english): Deep Learning Model for Money Management
• Author: Filip Šimek
• Mentor: Zvonko Kostanjčar (mentor)
• Committee member: Zvonko Kostanjčar (predsjednik povjerenstva)
• Committee member: Dražen Jurišić (član povjerenstva)
• Committee member: Stjepan Begušić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-15, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Široka primjena modela strojnog učenja u raznim znanstvenim disciplinama dovela je do sve češće upotrebe takvih modela u financijama, a osobito u algoritamskom trgovanju. Osim problema određivanja najboljih trenutaka za zauzimanje pozicije u financijskoj imovini, aktualan je i problem određivanja veličine pozicije. Jednostavni all-or-nothing pristup ne upravlja rizikom, dok je nedostatak često korištene Kelly strategije temeljenje na dobro procijenjenom očekivanom povratu investicije što može rezultirati sugeriranjem prevelikih i riskantnih pozicija. U ovom radu, razvijena je strategija trgovanja utemeljena na modelu ćelije s dugoročnim pamćenjem za određivanje dobrih trenutaka ulaska i veličine pozicije (utemeljena na vjerojatnosnoj interpretaciji izlaza modela). Rezultati pokazuju da takva strategija u nekim slučajevima ostvaruje najbolje rezultate prema Sharpe omjeru uz dobro upravljanje rizikom.

Abstract (english)

The widespread application of machine learning models in various scientific disciplines has led to the increasing use of such models in finance and, in particular, algorithmic trading. In addition to the problem of determining the best times to take positions in financial assets, the problem of determining the size of the positions is also important. A simple all-or-nothing approach cannot be used to manage risk, while the commonly used Kelly strategy has a disadvantage that it is based on a well – estimated expected return of investment which could result in oversizing and taking risks. In this paper, we developed a trading strategy based on a long short-term memory network model to determine good entry points and position sizes (based on a probabilistic interpretation of the model's output ). The results show that such a strategy can achieve the best results according to the Sharpe ratio along with good risk management.

• Keywords: model dubokog učenja; LSTM; strategija upravljanja novcem; klasifikacija trendova; strategija trgovanja; financijski vremenski nizovi; meta - označavanje
• Keywords (english): deep learning model; LSTM; money management strategy; trend classification; trading strategy; financial time series; meta - labeling
• Language: croatian
• URN:NBN: urn:nbn:hr:168:582084
• 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 16:16:19