Please use this identifier to cite or link to this item:
https://www.um.edu.mt/library/oar/handle/123456789/8627
Title: | Directionality prediction of currency exchange rates using gene expression programming |
Authors: | Dorato, Davide |
Keywords: | Artificial intelligence Algorithms Finance |
Issue Date: | 2015 |
Abstract: | system through the use of artificial intelligence algorithms applied to the financial markets. In particular, the studies performed focused on machine learning algorithms and evolutionary algorithms having as input technical analysis indicators of financial time series. The output of this research is a tool capable of simplifying the decision making activity in the financial trading, achieving the goal through an automated rules and criteria discovery algorithm. The system developed is auto-adapting to the different financial products analyzed, thus, outputting the structure of the best model found. Machine learning algorithms have been used as instruments to perform regression tasks, showing the capability to find patterns in financial time series. The patterns discovered are classified with a generalization principle that makes feasible the prediction of future values or decisions even on unseen data. The accuracy achieved in previous research in this field is very promising but the initial fine tuning processes which lead to the selection of the input for different financial products are still deeply lacking of re-usability of the logics implemented. Therefore, the literature is controversial on the application of these techniques. In fact even if previous research has produced good results, the level of uncertainty on how to apply proper configurations is still very high. On the other hand, evolutionary algorithms have been used to sort problems that have solutions lying in wide range of search spaces. In the present research, a hybrid system, built binding together machine learning algorithms with gene expression programming algorithm, has been applied to forecasting financial markets . The genetic approach shows promising outputs since the results significantly increased the prediction of future trends without performing complicated preprocessing activities. Therefore, the main goal of the present research has been reached thus, providing solid re-usable bases from which further analysis can be performed. |
Description: | M.SC.IT |
URI: | https://www.um.edu.mt/library/oar//handle/123456789/8627 |
Appears in Collections: | Dissertations - FacICT - 2015 |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
15MSCIT002.pdf Restricted Access | 4.84 MB | Adobe PDF | View/Open Request a copy |
Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.