Teknik analiz ve derin pekiştirmeli öğrenme ile kriptopara alım-satımı
No Thumbnail Available
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Open Access Color
OpenAIRE Downloads
OpenAIRE Views
Abstract
Son yıllarda teknolojinin yıkıcı etkisi birçok alanda kendini göstermektedir. Finans sektörü de bu durumdan fazlasıyla etkilenmiş durumdadır. Finansal piyasalar, artan rekabet ve gelişen teknoloji ile kriptopara piyasası gibi yenilikçi piyasaların oluşmasına zemin hazırlamaktadır. Finansal piyasalardaki değişime paralel olarak, yapay zeka alanındaki çalışmalarda da çok önemli gelişmeler olmaktadır. Bu çalışmada Robotik üzerine başarılı sonuçlar veren modern Derin Pekiştirmeli Öğrenme yöntemlerinden Soft-Aktör-Kritik(Soft Actor Critic - SAC) yöntemi ile finansal piyasalarda sıklıkla tercih edilen Teknik Analiz yöntemlerini kullanarak alım-satım stratejileri geliştirilmiştir. Piyasa değeri en yüksek üç kriptopara (Bitcoin, Ethereum ve Ripple), hem USD hem de BTC paritesinde veri seti olarak kullanılmaktadır. Çalışma kapsamında OpenAI-Gym ile kriptopara alım-satım ortamı oluşturulmuş ve bu ortamda SAC etmeni öğrenme süreci gerçekleştirilmektedir. Teknik Analiz yöntemleri ve SAC yöntemiyle oluşturulan stratejilerin performansları geriye yönelik testler(Backtesting) yapılarak karşılaştırılmaktadır. Anahtar Kelimeler: Derin Pekiştirmeli Öğrenme, Kriptoparalar, Algoritmik Alım-Satım, OpenAI-Gym, Soft-Aktör-Kritik, Teknik Analiz, Geriye Yönelik Testler
In recent years, the destructive effect of technology is manifested in many areas. The financial sector has been also highly affected by this situation. Financial markets set the stage for the development of innovative markets such as increasing competition and developing technology and cryptographic market. Parallel to the change in financial markets, there are also important developments in artificial intelligence studies. In this study, trading strategies have been developed using Soft-Actor Critic (SAC) method, which is a state-of-the-art deep reinforcement learning method, which gives successful results on robotics, and the technical analysis methods which are frequently preferred in financial markets. Three cryptocurrencies (Bitcoin, Ethereum and Ripple) with the highest market value are used as the data set in both USD and BTC parity. Within the scope of this study, a cryptocurrency trading environment has been created with OpenAI-Gym and the process of SAC agent's learning is realized in this environment. Technical Analysis methods and SAC method's performances are compared by backtesting. Keywords: Deep Reinforcement Learning, Cryptocurrencies, Algorithmic Trading, Backtesting, OpenAI-Gym, Soft-Actor-Critic, Technical Analysis
In recent years, the destructive effect of technology is manifested in many areas. The financial sector has been also highly affected by this situation. Financial markets set the stage for the development of innovative markets such as increasing competition and developing technology and cryptographic market. Parallel to the change in financial markets, there are also important developments in artificial intelligence studies. In this study, trading strategies have been developed using Soft-Actor Critic (SAC) method, which is a state-of-the-art deep reinforcement learning method, which gives successful results on robotics, and the technical analysis methods which are frequently preferred in financial markets. Three cryptocurrencies (Bitcoin, Ethereum and Ripple) with the highest market value are used as the data set in both USD and BTC parity. Within the scope of this study, a cryptocurrency trading environment has been created with OpenAI-Gym and the process of SAC agent's learning is realized in this environment. Technical Analysis methods and SAC method's performances are compared by backtesting. Keywords: Deep Reinforcement Learning, Cryptocurrencies, Algorithmic Trading, Backtesting, OpenAI-Gym, Soft-Actor-Critic, Technical Analysis
Description
Keywords
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control
Turkish CoHE Thesis Center URL
Fields of Science
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
123