For unsupervised classification purposes using several variables, I need to categorize one continuous variable into two classes, particularly, using Z-scores. The issue is that the continuous data need to be aggregated at a coarser scale (the unit of the classification). I understand the limitations of both categorization and aggregation, especially the high risk of information loss, but I am wondering what a better option can be if I have to choose between:

  1. First aggregate the data points (e.g. using the mean value), then categorize the spatially aggregated data into two classes; or
  2. First categorize the original data points into two classes, then aggregate the categorized data (in this case using the majority). I have applied both procedures and the difference is minor, however, this minor difference can significantly change the final classification results (when using all the other variables)



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.