| Resumen |
With the digitalization of financial markets, namely, stock markets the development of algorithms and computational techniques in order to determine trading strategies has gained relevance as much as in academia as in the industry. This article explores the use of pattern recognition techniques (Support Vector Machines, Multilayer Perceptron and C4.5) as tools for finding trading strategies in the Mexican Stock Market. Our results show that, statistically speaking, the methods proposed here, cannot outperform the so called Effcient Market Hypothesis in its weak version. Nonetheless, this paper presents a labeling method for financial time series, which permits future investigations using other supervised learning techniques. |
