The initial dataset, composed by about 2000 records, has been pre processed by purging outliers and non
valid entries, obtaining a final database composed by about 1800 records. The percentage of data referring to
the occurrence of under pickling defects and to the absence of pickling defects is of 75% and 25%, respectively.
The noticeable difference between the number of data belonging to the two classes makes the dataset quite
unbalanced and prevents the application of traditional classification methods as well as standard soft-computing
based classification approaches.