An Automated System for Stock Market Trading Based on Logical Clustering

Abstract

In this paper a novel clustering-based system for automated stock market trading is introduced. It relies on interpolative Boolean algebra as underlying mathematical framework used to construct logical clustering method which is the central component of the system. The system uses fundamental analysis ratios, more precisely market valuation ratios, as clustering variables to differentiate between undervaluated and overvaluated stocks. To structure investment portfolio, the proposed system uses special weighting formulas which automatically diversify investment funds. Finally, a simple trading simulation engine is developed to test our system on real market data. The proposed system was tested on Belgrade Stock Exchange historical data and was able to achieve a high rate of return and to outperform the BelexLine market index as a benchmark variable. The paper has also provided in-depth analysis of the system's investment decision making process which reveals some exciting insights.