next up previous contents
Next: Training a SOM Up: Process data Previous: Process data

Data preprocessing

This case study focused on a particular product grade. The data reflecting these products was filtered from the database. The data with missing values were removed. The data was scaled so that each component had variance one. The training set had approximately 3000 training vectors.



Jaakko Hollmen
Fri Mar 8 13:44:32 EET 1996