Numeric Outliers simple example

[ScottF]

Workflow originally developed for the thread below. Check it out for more context:

Identify Coding Outliers Using Unsupervised Method

I need an advice. For the file like: Specialty1, Provider1, Code1, Code_Frequency Specialty1, Provider1, Code2, Code_Frequency Specialty1, Provider1, Code3, Code_Frequency … Specialty1, Provider2, Code1, Code_Frequency … I’m looking for outliers who used too many Codes or too little codes for specified Specialty. Number of codes could be 100 per specialty and about 400 providers I used DBScan for some close task but just 10 codes. It would be nice to see an example with data preparation…