mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sebastian Schelter <...@apache.org>
Subject Re: ItemSimilarityJob as UserSimilarityJob
Date Fri, 08 Apr 2011 10:15:47 GMT
Please don't use the similarity jobs from Mahout 0.4 as they have a 
serious bug, use the trunk.

--sebastian

On 08.04.2011 12:12, Sean Owen wrote:
> Yep you can definitely do that, no problem.
>
> On Fri, Apr 8, 2011 at 11:04 AM, Thomas Rewig<trewig@mufin.com>  wrote:
>
>>   Hello,
>>
>> I am testing at the moment a bit with mahout-jobs in hadoop.
>>
>> As I understand it:
>> * the Recommender Job computes item-recommendations for all users
>> * and the ItemSimilarityJob computes all item-item-similaritys
>>
>> I wonder if there is a job for the calculation of user-user-similaritys ..
>> a UserSimilarityJob (Or have I write my own recommenderJob?). In my opinion
>> the RowSimilartityJob is not for this purpose, or?
>>
>> So if i have this data:
>>
>> |userID1,itemID1,preferencevalue|
>> |userID1,itemID2,preferencevalue
>> ||userID2,itemID1,preferencevalue|
>> |userID2,itemID3,preferencevalue
>> ...
>>
>> and transform it to
>> |
>> |itemID1||,||userID1||,preferencevalue|
>> |itemID1||||,||userID2||,preferencevalue
>> ||itemID||2,||userID1||,preferencevalue|
>> |itemID3||||,||userID2||,preferencevalue
>> ...
>>
>> ||i can use the |ItemSimilarityJob to get all user-user-similaritys and
>> this should be the same result I would expect in a UserSimilarityJob. But is
>> there an easier way without a transformation of the data i use to get all
>> user-user-similartitys?
>>
>> Or have I missed something?
>>
>> I use mahout 0.4.
>>
>> Thanks in advance!
>> Thomas
>>
>>
>>
>>
>

Mime
View raw message