mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Owen <sro...@gmail.com>
Subject Re: Questions regarding recommendations
Date Tue, 18 Jan 2011 12:25:22 GMT
That's right. Though long-term I think you will want to use a numeric ID for
simplicity and performance. It's pretty common to have, for example, a
unique ID for each entity even if you also have another unique identifier.
For example, a user probably has a (unique) email address but any usual DB
schema is going to assign it a numeric primary key too. I'd suggest you try
to add a numeric ID to your users/items in the long term.

On Tue, Jan 18, 2011 at 11:18 AM, Sebastian Schelter <ssc@apache.org> wrote:

> Hi Komal,
>
> you have to use an IDMigrator (
> https://hudson.apache.org/hudson/job/Mahout-Quality/javadoc/org/apache/mahout/cf/taste/model/IDMigrator.html)
to map your alphanumeric ids to long ids and vice versa.
>
> --sebastian
>
>
> On 18.01.2011 12:09, Komal Goyal wrote:
>
>> Hi,
>>
>> I was looking into the recommendation documentation on the apache mahout
>> site which says:
>>   *Users and items are identified solely by an ID value in the framework.
>> Further, this ID value must be numeric; it is a Java long type through the
>> APIs.*
>>
>> but IDs in my database are alpha numeric.
>>
>> I am having the following:
>> Users:
>>    key
>>    name
>>    email
>>    other_details
>>    url_resource
>>    time_stamp
>>
>> Behavior:
>>    key
>>    value
>>     time_stamp
>>     type
>>
>>
>> UserBehavior:
>>    key
>>    userkey
>>    behaviorkey
>>
>> Currently behavior type has only one value.
>>
>> I want to search for behaviors that can be recommended to a particular
>> user
>> from the other closest users.
>>
>> Please suggest me which algorithms and  data model I can use to get the
>> recommendations?
>>
>> Thanks,
>>
>> *Komal Goyal*
>>
>>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message