Skip to main content

Evolving Directional Changes Trading Strategies with a New Event-based Indicator

Kampouridis, Michael and Adegboye, Adesola and Johnson, Colin (2017) Evolving Directional Changes Trading Strategies with a New Event-based Indicator. In: Shi, Yuhui and Tan, Kay Chen and Zhang, Mengjie and Tang, Ke and Li, Xiaodong and Zhang, Qingfu and Tan, Ying and Middendorf, Martin and Jin, Yaochu, eds. Simulated Evolution and Learning 11th International Conference. Lecture Notes in Computer Science . Springer, Cham, Switzerland, pp. 727-738. ISBN 978-3-319-68758-2. E-ISBN 978-3-319-68759-9. (doi:10.1007/978-3-319-68759-9_59) (KAR id:62608)

PDF Author's Accepted Manuscript
Language: English
Download (243kB) Preview
[img]
Preview
Official URL
http://dx.doi.org/10.1007/978-3-319-68759-9_59

Abstract

The majority of forecasting methods use a physical time scale for studying price fluctuations of financial markets, making the flow of physical time discontinuous. An alternative to this is event-based summaries. Directional changes (DC), which is a new event-based summary method, allows for new regularities in data to be discovered and exploited, as part of trading strategies. Under this paradigm, the timeline is divided in directional change events (upwards or downwards), and overshoot events, which follow exactly after a directional change has been identified. Previous work has shown that the duration of overshoot events is on average twice the duration of a DC event. However, this was empirically observed on the specific currency pairs DC was tested with, and only under the specific time periods the tests took place. Thus, this observation is not easily generalised. In this paper, we build on this regularity, by creating a new event-based indicator. We do this by calculating the average duration time of overshoot events on each training set of each individual dataset we experiment with. This allows us to have tailored duration values for each dataset. Such knowledge is important, because it allows us to more accurately anticipate trend reversal. In order to take advantage of this new indicator, we use a genetic algorithm to combine different DC trading strategies, which use our proposed indicator as part of their decision-making process. We experiment on 5 different foreign exchange currency pairs, for a total of 50 datasets. Our results show that the proposed algorithm is able to outperform its predecessor, as well as other well-known financial benchmarks, such as a technical analysis.

Item Type: Book section
DOI/Identification number: 10.1007/978-3-319-68759-9_59
Uncontrolled keywords: directional changes; algorithmic trading; financial forecasting; genetic algorithms
Divisions: Faculties > Sciences > School of Computing > Computational Intelligence Group
Faculties > Sciences > School of Computing > Data Science
Depositing User: Michael Kampouridis
Date Deposited: 10 Aug 2017 15:27 UTC
Last Modified: 14 Oct 2019 11:16 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/62608 (The current URI for this page, for reference purposes)
Kampouridis, Michael: https://orcid.org/0000-0003-0047-7565
Johnson, Colin: https://orcid.org/0000-0002-9236-6581
  • Depositors only (login required):

Downloads

Downloads per month over past year