cancel
Showing results for 
Search instead for 
Did you mean: 

Head's Up! These forums are read-only. All users and content have migrated. Please join us at community.neo4j.com.

Does anyone know where Jaccard similarity algorithm is?

Hello:
All the documentation states that Jaccard is an alpha algorithm and should be in:
gds.alpha.similarity.jaccard
but it's not there!

This is so in both a desktop version and a community edition. I wonder if it has now been restricted to just the enterprise edition.

The Jaccard algorithm compares two lists for similar entities, whereas node.similarity tests the members of a single list against one another.

My task is to take a list of deceased overdose, and another of still surviving, and find the closest matches. Presumably in a single list, the deceased will match one another more!

Thanks for any help in locating the Jaccard. I will try the sandbox, but because the data is in a secure environment, won't be able to use Sandbox.

Thanks
Ernie

5 REPLIES 5

Hello @ecsendmail

The Jaccard Similarity algorithm is in the Graph Data Science plugin.

The plugin must be installed on your database, there are a few tutorials here depending of where you want to install it.

Regards,
Cobra

Thanks for your help, Cobra.
I did find the answer to what was puzzling me. I did have GDS loaded, but it seems that the auto-completion when one types in gds. and then a whole list comes up, so if I go further and type in gds.alpha I get a further list.....and the completion of gds.alpha.Jaccard is empty, and so is gds.alpha.sim....which brings up a whole bunch of alpha similarity algorithms such as cosine and overlap BUT not Jaccard.

In fact, when one just types in "call gds.alpha.similarity.jaccard the error message that this procedure is not recorded for this database instance

However, the real gold mine is in the command call dbms.functions( ) which return ALL functions, and way down the list one can find gds.alpha.similarity.jaccard, with all the parameter descriptions.

The completions list is weird because I sometimes see jaccard in it, but most of the time NOT....anyway, I have been able to use it in my app, and not even having to use CALL but just as a function within a RETURN......

which is also nice to know....

Thanks, Cobra.....if you have any insights into the completion peek-a-boo, thanks for sharing!

Ernie

Which version of Neo4j database and GDS are you using?

Oh I updated to 4.3.5 for my dbms that I am working on, and to 1.7.1 for gds, and 4.3.0.3 for apoc, took a while to go the 10 steps from 4.2.5 to 4.3.5 and then had to reconstruct the whole schema! but now I am back to where I was, have used jaccard, but the odd completion business in the desktop browser for cypher entry is the only puzzle....and the workaround that I have adopted is to look in dbms.functions() if I really need to find an algorithm or a function....

Still, a puzzle....that trips the unwary (like me...now warier by the stumble!)

Thanks again, Cobra 😆
Ernie

ps I also updated the desktop.....so la lucha continua....