Which issue arises when applying Yeo-Johnson transformation to a feature containing a significant number of zero values?

Question

Data Science — Hard

Which issue arises when applying Yeo-Johnson transformation to a feature containing a significant number of zero values?

Accepted Answer

Unlike the Box-Cox transformation, which requires strictly positive data, the Yeo-Johnson transformation is specifically designed to handle zero and negative values. It uses different power transformation formulas for positive and negative values, ensuring continuity and differentiability at zero. Therefore, it does not fail or require manual shifting for zero-valued data.