Wednesday, May 25, 2016

Why Data Scientists Must Understand Normalization

My May post @All Analytics:

We are constantly told how data scientists must be “jacks of many skills”, but one of the most important is rarely included in the list.Very few databases are properly designed. Many SQL databases are denormalized inadvertently, or intentionally (and erroneously) "for performance". They require special constraints to control data redundancy and prevent inconsistencies, which are practically never enforced. Analysts cannot, therefore, take database consistency for granted. Furthermore, to issue sensible queries and ensure correct results and interpretation thereof, it’s not enough for analysts to know the types of fact represented in the database, but also whether and how the database designer has chosen to bundle -- nest or merge -- those facts and how to disentangle them for analysis.

Read it all. (Please comment there, not here)


No comments:

Post a Comment

View My Stats