Application of Binary Classifiers to Filter Transactions on the Financial Market

Abstract

One of the key problems relating to concluding transactions on financial markets is the definition of whether or not, in given market conditions, we should conclude a specific transaction. This problem applies to all transactions irrespective of whether they are concluded in a discretionary manner by a person, or by the appropriate software. The subject of research presented in this document was practical verification of the hypothesis that the use of filters composed of classifiers independent of the main algorithm of the strategy may improve the characteristics of this strategy. This means constructing an additional filter, independent of the main logic and algorithm of the strategy, either allowing the transaction to be conducted or not. The strategy being examined is a real strategy used in trading in shares on the American market. Many various classifiers based on various algorithms, as well as sets of classifiers composed of them, were examined.As a consequence of this research interesting results were obtained, noticeably improving the various characteristics of an automatic strategy for which the filters had been created. The research confirmed the justification for using classifiers as a filter for transactions for the automatic strategy under examination.