First of all, I should preface this by letting you know that I'm a SQL novice - I've never really used SQL Server before and what I'd like to do must be quite rare or challenging because I've been unable to find any relevant answers on StackOverflow or Google.
I'd really, really appreciate your help on this. In the meantime, I myself am currently trying to improve my SQL knowledge and unearth a way to tackle this - but let's get straight to the point
I'm currently in possession of a SQL Server (which I browse through SQL Server Management Studio) with 4 tables. Everything's in Greek so no point in writing the real names. The point is that each row in Table 1 is associated with multiple rows in Table 2, which in turn is associated with multiple rows in Table 3, which in turn is associated with multiple rows in Table 4
My task is to perform AI/Machine Learning on this multi-instance multi-label problem, but to do that, I have to make it so there is only 1 table containing all the information of all tables.
SQL Server database structure:
You use joins to get the information. Tehcnically, you can do something like
SELECT * FROM Table1 JOIN Table2 ON Table1.Table2Id = Table1.ID JOIN Table3 ON Table2.Table3Id = Table3.ID
Etc. But, you end up with repeats that can mess things up, so you are better to only select the columns you require. The joins here are one way, and will exclude nulls, so you might need other types of joins. The most information comes from a cross join, but it makes a Cartesian product of all of the tables involved, so you have the potential of getting more back than what you require.
Here is a link that explains joins in T-SQL: http://www.techonthenet.com/sql_server/joins.php
It is a good place to get started and may answer your question with a little bit of experimentation on your part.