A New Methodology Based on Multicriteria Ordinal Classification for the Management of Financial Resources with Application to Real Data from the Stock Market

Stock selection is highly complex due to the high heterogeneity of its factors. The determination of the value of a stock depends to a large extent on the investor’s preferences towards such factors. This paper describes and evaluates a new methodology that uses investor decision policy to assign stocks to preferentially ordered classes. These classes can be of the “Don’t Buy”, “Doubt” or “Buy” type. The classes are identified by limiting profiles at the boundary of each pair of consecutive classes and are given a priori by the investor. A back-testing strategy is used to evaluate the proposal and its results are compared with those of some benchmark approaches. The primary findings highlight that the stocks classified within the best class not only yielded better average returns compared to the broader market but also exhibited significantly lower volatility, suggesting a more favorable risk-reward balance and outperforming conventional methods and market benchmarks in terms of both returns and risk management.