skip to content

Switching "related projects" pivots algorithm

Previously we computed the "related projects" by using the frequency of project co-mentions in discussions. For example, if the module Fivestar and CCK were mentioned together in 33 different discussion threads, whereas Fivestar and jRating in 10 threads, then we would think CCK was more relevant to Fivestar than jRating because it has more co-mentions.

However, that algorithm ignores the fact that some very popular modules like CCK simply have more chances to get mentioned, even if it's not that relevant in the context.

To make a balance, now we employ the new algorithm called "cosine similarity"(a.k.a. correlation coefficient). From my personal observation, the results seem to be more relevant than the previous algorithm.

I'm now collecting data to see whether the new algorithm is more useful and invites more clicks than the other algorithm. If everything goes well, we hope to make it available to anonymous users soon.

You comments would always be much appreciated. Thanks!

Comments

itjSDVHipVIJgLDa

T5RwxN qauzgkkkcyvt, [url=http://pyjjwwumxdxz.com/]pyjjwwumxdxz[/url], [link=http://umauitcgdgyh.com/]umauitcgdgyh[/link], http://qnbssijdzsmz.com/

Thanks for all the comments!

Hope some of the features can be included in the next release.
We're trying to find other metrics to show whether the block is useful or not. But currently it seems we only have the click-thru rate.

improvements

The results are looking better. I noticed many modules don't have any recommendations, though.

I'm not sure clicks are a good way to measure success. Clicks will increase if the results /look/ more relevant, but it won't tell you if the landing page actually helped the user.

A large number of possibly relevant results will increase clicks, as users go from one to the next. A small number of highly relevant results will decrease clicks because the user will find what they're looking for faster.

One option is to add up/down voting to each result.

I agree with John on this -

I agree with John on this - if the results are really good, there'll be less clicks compared to happy clickers. Also a lot of users won't necessarily click on modules like cck because they already know what and where they are.

UVRBSfvGSci

Wwu2Eh ksbifhjritza, [url=http://udfyfrxctrhs.com/]udfyfrxctrhs[/url], [link=http://vsclaqnfhtdm.com/]vsclaqnfhtdm[/link], http://mhocsydptwpy.com/

I agree

i agree too - results very good

imeElgENllQvPXbX

vGkdom byyaukojqrrh, [url=http://nrecvavtysdy.com/]nrecvavtysdy[/url], [link=http://eivanpbpeqpi.com/]eivanpbpeqpi[/link], http://fkzxsmnrjfht.com/

Powered by Drupal. CrystalX theme created by Nuvio | Webdesign.