In this work, a new methodology is presented to analyze and predict the behavior of stocks of the Mexican Stock Market based on the synergistic concatenation of non-parametric statistical strategies and multi-objective optimization models. This methodology involves two phases. The first (filtering) leverages an automated process for the analysis, evaluation, and selection of the necessary and relevant information: for the characterization of the behavior of each action. The second (the model adjustment phase) involves adapting and solving a multi-objective model for the prediction of prices of the selected stocks. The database used in this work includes the behavior of twelve significant stocks in the Mexican stock exchange in the period 2006 to 2016, the source code used is available in “http://bit.ly/396h3J1”; the data was obtained from a specialized financial markets platform for Latin America. The numerical results show that the filtering phase can identify a compact set of relevant variables with a significant influence on the future price of each stock. In the second phase, the data from 2016 is used to predict the multi-objective model, that compared with the multiple linear regression model, provides a considerable improvement in the quality of the predicted observed data. The model generated from the second phase has reliability greater than 95%.

No abstract