Is a snowflake pattern better than a stellar pattern for data mining?

Question

I know the main difference between a star-shaped scheme and a snowflake scheme - a snowflake scheme splits the measurement tables into several tables to normalize them, the star scheme has only one "level" of the measurement tables. But the Wikipedia article for the Snowflake scheme says

“Some users can send queries to the database, which, using conventional multidimensional reporting tools, cannot be expressed in a simple star scheme. This is especially true for data mining of customer databases, where the common requirement is to search for common factors that arise between customers, who buy products that meet complex criteria, usually requiring snow removal to allow simple query tools to create such a query, especially if providing these forms dew was not expected when the data warehouse was first designed. "

When would it be impossible to write a query in a star schema that could be written in a snowflake schema for the same underlying data? It seems that the star-shaped design always allows the same queries.

Answer 1

For data mining, you almost always need to prepare your data - basically like a “flat table”.

This can be a request, prepared viewing or export of CSV - it depends on the tool and your preferences.

Now, in order to correctly understand this article, one could probably smoke, drink the same thing as the author when he wrote it.

Answer 2

As you say, preparing a flat table for data mining, starting with a relational database, is not an easy task, and a snow or star scheme works only to a certain point.

However, there is software called Dataconda that automatically creates a flat table from the database.

Basically, you select a target table in a relational database, and dataconda “extends” it by adding thousands of new attributes to it; these attributes are obtained by performing complex queries using multiple tables.