Transferring trading strategy knowledge to deep learning models
- PDF / 1,187,291 Bytes
- 18 Pages / 439.37 x 666.142 pts Page_size
- 63 Downloads / 234 Views
Transferring trading strategy knowledge to deep learning models Avraam Tsantekidis1
· Anastasios Tefas1
Received: 17 January 2019 / Revised: 26 August 2020 / Accepted: 30 August 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract Trading strategies are constantly being employed in the financial markets in order to increase consistency, reduce human errors of judgment and boost the probability of taking profitable market positions. In this work, we attempt to transfer the knowledge of several different types of trading strategies to deep learning models. The trading strategies are applied on price data of foreign exchange trading pairs and are actual strategies used in production trading environments. Along with our approach to transfer the strategy knowledge, we introduce a preprocessing method of the original price candles making it suitable for use with Neural Networks. Our results suggest that the deep models that are tested perform better than simpler models and they can accurately learn a variety of trading strategies. Keywords Trading strategy · LSTM · RNN · Deep learning
1 Introduction Financial exchanges are considered to be all the licensed hubs where financial institutes, investors and other entities submit their demands to buy or sell financial assets and speculate on their future values, among other financial activities. The most important function of financial markets is to efficiently allocate capital to businesses, so they can expand their activities. Investors that place their investment in successful companies can have those investments increase in value as the stock price of said companies rises. The price of an asset is determined by the price investors who are willing to pay for it at the time. Since this way of pricing can be volatile, the price of assets usually fluctuates through time. Another activity happening in the financial markets is the speculation of the true price of assets. Seasoned investors can, at their own discretion, take advantage of price fluctuations to profit. Financial firms such as hedge funds employ discretionary traders who use quantitative and qualitative analysis of price movements to decide where to invest capital.
B
Avraam Tsantekidis [email protected] Anastasios Tefas [email protected]
1
Department of Informatics, Aristotle University of Thessaloniki, Thessaloniki, Greece
123
A. Tsantekidis, A. Tefas
Humans undertaking such tasks have the disadvantage that an individual’s sentiment may greatly affect their judgment and thus the efficacy and consistency of their trading behavior. Even though humans are not always suitable for the task, the quantitative methods they use have met great success in the past with firms having considerable profits from utilizing them [10,15]. Due to the success of quantitative methods such as technical analysis and since they are relatively formulaic, many of them were implemented as automated computer programs. The trading algorithms that emerged from this process were more consistent, and a great deal
Data Loading...
