Build adjacency matrix from list of weighted edges in BigQuery

问题

Related issue: How to create dummy variable columns for thousands of categories in Google BigQuery

I have a table of list of weighted edges which is a list of user-item rating, it looks like this:

| userId | itemId | rating
| 001    | 001    | 5.0
| 001    | 002    | 4.0
| 002    | 001    | 4.5
| 002    | 002    | 3.0

I want to convert this weighted edge list into a adjacency matrix:

| userId | item001 | item002
| 001    | 5.0     | 4.0
| 002    | 4.5     | 3.0

According to this post, we can do it in two steps, the first step is to extract the matrix entry's value to generate a query, and second step is to run the query which is generated from 1st step.

But my question is how to extract the rating value and use the rating value in the IF() statement? My intuition is to put a nested query inside the IF() statement such like:

IF(itemId = blah, 
                 (select rating 
                  from mytable 
                  where 
                    userId = blahblah 
                    and itemId = blah), 
                 0)

But this query looks too expensive, can someone give me an example?

Thanks

回答1:

Unless I am missing something - it is quite similar to the post you referenced

Step 1 - generate query

SELECT 'SELECT userID, ' + 
   GROUP_CONCAT_UNQUOTED(
    'SUM(IF(itemId = "' + STRING(itemId) + '", rating, 0)) AS item' + STRING(itemId)
   ) 
   + ' FROM YourTable GROUP BY userId'
FROM (
  SELECT itemId 
  FROM YourTable  
  GROUP BY itemId
)

Step 2 - run generated query

SELECT 
  userID, 
  SUM(IF(itemId = "001", rating, 0)) AS item001,
  SUM(IF(itemId = "002", rating, 0)) AS item002 
FROM YourTable 
GROUP BY userId

Result as expected

userID  item001 item002  
001     5.0     4.0  
002     4.5     3.0

来源：https://stackoverflow.com/questions/37825142/build-adjacency-matrix-from-list-of-weighted-edges-in-bigquery

标签

matrix

google-bigquery

transpose