Akimochkin Artem

Faculty: Computer science
Speciality: Economical cybernetics
Theme of master's work:
"Development and research of adaptive trading systems invariant to changes of a market situation"

Abstract ^*

of Thesis for a Master’s Degree in Computer Science

"Development and research of adaptive trading systems invariant to changes of a market situation"

Author: Akimochkin A.

Leader of work: Ph.D. of technical science Alexander Smirnov

Introdution

Recently Ukraine has entered the World Trade Organization. At many participants of the WTO external trading export-import transactions are organised through exchange trade, and 95 percent of all legal operations are carried out at stock exchanges. Ukraine only should make it. As in the USSR any stock exchanges and participation in the auctions at present in Ukraine there are no serious independent workings out in this sphere have been forbidden. Therefore the theme of exchange trade is actual. On the other hand in the technical analysis exists an order of two thousand various indicators and their updatings, but not one of them is not good for construction of computer trading system. The purpose of this work to develop algorithm of associations of several indicators for the purpose of increase of their quality on the basis of adaptability.

Problem and research objective statement

As initial data was the file of the prices of closing of day bars for 2 years of currency pair EUR/USD, from the beginning 2002 on the end is taken 2003
On historical data of the initial price schedule are subjectively placed turnaround points - points in which growth is replaced by falling and on the contrary. The moments of ideal turns of the market are designated by "units", and the moments of their absence in "zero". The received sequence of zero and units is basic. Further by means of technical indicators the sequence of signals of inputs in the market is formed. Thus, for us is available two time of numbers  "units" and "zero", basic and real.
The essence of the first method consists in a current estimation of size of mutual correlation between basic and real sequence. In an ideal this size should aspire to unit, but in practice the factor of mutual correlation is much less than unit. It is a consequence of poor quality of modern computer trading systems on the basis of one indicator. Current value is compared to in advance established special value, and on an exit the cleared sequence of trading signals turns out.
The following method uses the similar approach, but current value the relation quantity of the trends distinguished by the indicator and the general number of trends for the period pays off.
And last developed method, uses a current estimation of a dispersion of delay of signals of real sequence from signals of basic sequence.
Each of methods has special value which can vary, changing level of reliability and profitableness of trading system on the basis of this method.
As a result under more low resulted scheme trading systems have been constructed some.

Results of research

The received trading systems have been tested on historical data. Also have yielded following results.

Apparently from the table on a basis concerning the unreliable indicators, the developed methods have allowed to create reliable trading systems which pass "adverse" sites of the market, blocking proceeding signals. The positive effect was expressed in increase in a share of profitable transactions. Factor Sharp became more zero and total loss has decreased. Moreover, the prize was showed in essential increase in size PF to 4-5 times.

Conclusion

On the basis of the spent researches it has been revealed, that the technique of introduction of algorithms of increase of invariancy of MTS is the effective approach allowing from set of unreliable private indicators to realise adaptive and rather бездисковые MTS for conducting of stockjobbings.
It was possible to construct MTS with very small risk, taking some indicators with not correlated signals and having united them in one system it is possible to achieve that on all трендовых sites of the market the system will be more effective to work, than each indicator separately.
Use of the developed algorithms, essentially reduces investment risk of trading operations, stabilises economic indicators of trading systems, irrespective of change of a market situation and maximisation of a parity of the award over risk.

Literature