MySQL how to rank objects by similarity of multiple property rows - mysql

Question

Hello all and a Happy New Year
SITUATION:
I have some tables in MySQL db:
Scores:
(Unique ID, unique (objectID, metricID))
| ID | ObjectID | MetricID | Score |
|--------+----------+----------+----------|
|0 | 1 | 7 | 0 |
|1 | 5 | 3 | 13 |
|2 | 7 | 2 | 78 |
|3 | 7 | 3 | 22 |
|.....
|--------+----------+----------+----------|
Objects:
(unique ID, unique ObjectName)
| ID | ObjectName |
|--------+------------|
|0 | Ook |
|1 | Oop |
|2 | Oww |
|3 | Oat |
|.....
|--------+------------|
Metrics:
(unique ID, unique MetricName)
| ID | MetricName |
|--------+------------|
|0 | Moo |
|1 | Mar |
|2 | Mee |
|3 | Meep |
|.....
|--------+------------|
For a given object ID:
There will be a number of scores between '0' and 'one per metric'
REQUIREMENT:
For a given ObjectID, I want to return a sorted list based on the following criteria:
Returned rows ranked in order of similarity to the provided object
Returned rows not to include provided object
(this is the hard bit I think) Order of similarity is determined by an object's "score distance" from the provided object based on the numeric offset/difference of its score from the provided object's score for any metric for which there is an entry for both the provided and the currently-examined objects
Contains objectID, Object name, score difference (or something similar)
PROBLEM STATEMENT:
I don't know the correct SQL syntax to use for this, and my experiments so far have failed. I would like to do as much of this work in the DB as possible and have little or none of this work done in nasty for-loops in the code or similar.
ADDITIONAL NON-FUNCTIONALS
At present there are only 200 rows in the Scores table. My calculations show that ultimately there may be up to around 2,000,000 rows, but probably no more.
The Objects table will only ever have up to around 5000 rows
The Metrics table will only ever have up to around 400 rows

Answer 1

Here's an approach to order objects based on their similarity to object 1:
select other.ObjectID
, avg(abs(target.Score - other.Score)) as Delta
from Scores target
join Scores other
on other.MetricID = target.MetricID
and other.ObjectID <> target.ObjectID
where target.ObjectID = 1
group by
other.ObjectID
order by
Delta
Similarity is defined as the average difference in common metrics. Objects that do not share at least one metric with object 1 are not listed. If this answer makes wrong assumptions, feel free to clarify your question :)
Live example at SQL Fiddle.