Banknote Authentication Dataset (DEA-Oriented Features)

The raw dataset contains four wavelet‐based features extracted from banknote images—variance, skewness, kurtosis, and entropy. Following the fundamental DEA principle that inputs represent quantities for which lower values are preferable and outputs represent quantities for which higher values are preferable, skewness and kurtosis were modeled as inputs because higher values indicate undesirable irregularity and extreme deviations in the signal, whereas variance and entropy were modeled as outputs since higher values reflect richer and more informative image structures.

For the purposes of applying Data Envelopment Analysis (DEA), the raw data were further preprocessed. Specifically, negative values in the original dataset were transformed into positive values using Min–Max scaling, mapping the data into a standardized positive range. This preprocessing step ensures compatibility with DEA efficiency measurement models.