Thesis-2006-Mylonas.pdf (5.65 MB)
Technical analysis and artificial neural networks as prediction tools in the equity markets
thesis
posted on 2014-02-13, 12:52 authored by Petros MylonasThis study investigates the possible forecast power of a wide spectrum of technical
rules on the equity markets. Two equity indices (Nasdaq Composite and Athens
General Index) have been chosen as case studies. The results of the tests show that
there is evidence of forecast power for many of the technical strategies. The
optimization methodology improves substantially the achieved returns. The
performance is higher in the case of Nasdaq Composite which could be a paradox
since it is a much more developed and efficient market. The performance of technical
strategies deteriorates dramatically during the most recent period.
When transaction costs are taken into account in a realistic way, the technical trading
strategies fail in the majority of the cases to generate higher returns than a naIve buyand-
hold strategy. It is proved that transaction costs generate an inflated effect on the
total profits which can be more than double the nominal amount paid in these costs.
The importance of trading cost has been underestimated by many previous studies
mainly due to unrealistic ways for their calculation. The unsatisfactory performance
of technical analysis during different time horizons and data is mainly due to its static
nature that deprives the method from adjusting to the constantly changing market and
economic conditions. The proposed solution is a trading model that combines
artificial neural networks, genetic algorithms and technical analysis. The results are
very optimistic since there is evidence for significant forecast power and consistent
abnormal returns in a very difficult short term prediction as it is to predict next day's
market direction.
History
School
- Business and Economics
Department
- Economics
Publisher
© Petros MylonasPublication date
2006Notes
A Doctoral Thesis. Submitted in partial fulfilment of the requirements for the award of Doctor of Philosophy of Loughborough University.EThOS Persistent ID
uk.bl.ethos.432234Language
- en