In this paper, we introduce an improved feature selection algorithm TSSF designed to enhance computational efficiency in phylogenetic tree construction by evaluating feature subsets and identifying those that contribute positively or negatively to tree topology. Building on the symbolic aspects of pattern systems, which encompass symbols (data), syntax (relations between data points), and layout rules, our approach systematically assesses the effectiveness of feature subsets. This is illustrated using essential phylogenetic metrics, including the Consistency Index, Retention Index, and Rescaled Consistency Index. In these metrics, higher values suggest high-quality feature subsets, whereas lower scores on the Homoplasy Index indicate minimized noise within the feature set. A comprehensive analysis reveals that the algorithm efficiently handles non-linear complexities, enabling it to distinguish between good and bad feature sets while maintaining computational efficiency. The results underscore the importance of informed feature selection in improving computational performance and the accurate reconstruction of evolutionary history. Given its flexibility, our algorithm serves as a robust tool for analyzing the evolution of varied pattern systems, making a significant contribution to the emerging field of Scriptinformatics.