Financial time series are rich in information that can be extracted for knowledge. However, time series require special preprocessing in order to be analyzed by machine learning algorithms. Financial time series often have a multidimensional structure, where future movements are influenced not only by their trajectory but also by the movements of related financial time series. Furthermore, interpretability is frequently desired. In this study, we investigate a hybrid of feature engineering for financial time series, an evolutionary strategy for data processing, followed by a subsequent rule-mining step for interpretability. Simultaneously, we investigate different evolutionary strategy (ES) algorithmic configurations to further improve the overall discretization process outcome for profitable trading decisions.