Język Polski
English
Русский
Slovenčina
Hrvatski
AI in determining indicators for FX prediction models
Abstract
Decision support systems in FX trade has proven successfully has proven useful if not necessary. Their rate of prediction success depends on the model constructed and data selected. The basic information for constructing models comes from the interesting pair FX rate in the specific time periods. The rest of data comes from selected indicators stating what could be the reason for particular FX rates in the time. The indicators come from various sources like the economic theories, traders’ experiences, etc. Analysis of various methods revealed, that although authors and practitioners employ various methods including ML, ANN and others, they do not use any of the methods to verify or determine which indicators should be considered as an input. This approach to selection of indicators is insufficient. The best prediction models would be bases on the cause-effect relation, and not the coincidental one. To avoid a non causa pro causa and other fallacies the relation the indicators need to be determined. Yet in the complicated FX market, important indicators may be invisible at first glance, due to our computational and memory limitations. Therefore some methods of pattern
search including unsupervised learning are suggested due to their successes in areas like fraud and corruption detection. They have proven, that sometimes unexpected influential indicators may be found by methodologies like unsupervised learning. Also, a suggestion, that those indicators can be determined real-time and become a part of adaptive FX predictive system is made.
Keywords:
Forex, FX indicators prediction, FX, AI, Machine LearningMost read articles by the same author(s)
- Krzysztof Michalski, Rafał Lizut, GENERATIVE ARTIFICIAL INTELLIGENCE (GenAI) – THREATS, SECURITY REQUIREMENTS AND CHALLENGES , Fides, Ratio et Patria. Studia Toruńskie: No. 19 (2023): INNOWACYJNA MEDYCYNA W SŁUŻBIE ZDROWIA
Details
References
Statistics
Authors
Citation rules
Altmetric indicators
Licence
This work is licensed under a Creative Commons Attribution 4.0 International License.