School of duplicates - and how to deal with them

Dealing with duplicates is a constant theme with data scientist. And a lot of things can go wrong. The easienst ways to deal with them is GROUP BY or DISTINCT. Just get rid of them and be done. But as this examples might demonstrate this might not always be the best option. Even if your data provider swears your combined IDs are unique especially in Big Data scenarios there might still be lurking some muddy duplicates and you shoudl still be able to deal with them.

And you should be able to bring a messy dataset into a meaningful table with a nice unique ID without loosing too much information. And this workflow would like to encourage you to think about what to do with your duplicates and not to get caught off guard but to take control :-)

External resources

• use H2 to produce a Position / Rank number within a group variable (window functions)
• School of duplicates - and how to deal with them (corresponding article)
• School of Hive - with KNIME's local Big Data environment (SQL for Big Data)
• long forum debate about duplicates