Developed and integrated a new variable scaling transformation into the root-project/root repository’s TMVA preprocessing pipeline, enabling linear scaling of features to the range [-1, 1] while preserving the original sign of input data. This enhancement extended the VariableNormalizeTransform class using C++ and improved the consistency of feature scaling for machine learning workflows. The update included comprehensive documentation and usage examples written in LaTeX, ensuring clarity for future users. By refining the data preprocessing stage, the work aimed to facilitate faster model convergence and reduce manual tuning, demonstrating a focused approach to improving machine learning infrastructure through robust C++ development.