You have only "categoria":["INVESTIMENTO”] type properties, so the name is 
“categoria” and value is ["INVESTIMENTO”]


{
  "user": "10000179",
    "fields": [
    {
      "name": "categoria",
      "values": ["INVESTIMENTO"],
      "bias": 1.02
    }
  ]
}

If you $set a property of “fasciaEta” with a value of something like ["45-60”] 
then your query will work.

You are using a bias of 1.02 so this will be multiplied times the score for 
every recommendation of "user": “10000179” that matches the property specified 
so the recommendations that match will be scored higher by the query and ranked 
higher than they would be without the fields query.



On Feb 18, 2017, at 2:52 AM, Andrea Santurbano <[email protected]> wrote:

Thank you Pat!
Just one last question:
Is the model that i have designed fine to be queried in this way?
https://gist.github.com/conker84/c0e7f25eadcaef03d908ef6d5ea5c3ed 
<https://gist.github.com/conker84/c0e7f25eadcaef03d908ef6d5ea5c3ed>

Il giorno sab 18 feb 2017 alle ore 03:15 Pat Ferrel <[email protected] 
<mailto:[email protected]>> ha scritto:
What is the problem?

Scores of more than one are certainly quite normal. The only thing you care 
about is the ranking/ordering of the results, not the absolute score. Scores of 
0.0 being the exception, which indicates that the item is returned from the 
popularity model.


On Feb 17, 2017, at 12:36 PM, Andrea Santurbano <[email protected] 
<mailto:[email protected]>> wrote:

Hi Pat,
here the last iteration of the model:
https://gist.github.com/conker84/39734a8a6a5d9564449439b838ab8ab7 
<https://gist.github.com/conker84/39734a8a6a5d9564449439b838ab8ab7>
This is how i passed the events:
https://gist.github.com/conker84/1fe6788fc527f591488fbb730b599542 
<https://gist.github.com/conker84/1fe6788fc527f591488fbb730b599542>
I think i made some other mistake becouse the score values are greater that 1:
https://gist.github.com/conker84/12e58c5af470677d56f9a8bfa99608e8 
<https://gist.github.com/conker84/12e58c5af470677d56f9a8bfa99608e8>

Thanks
Andrea


Il giorno gio 16 feb 2017 alle ore 00:19 Pat Ferrel <[email protected] 
<mailto:[email protected]>> ha scritto:
The properties are input incorrectly, they should all be arrays of strings, 
even if there is only one string in the array.

Docs: http://actionml.com/docs/ur_input <http://actionml.com/docs/ur_input>



On Feb 15, 2017, at 1:31 PM, Andrea Santurbano <[email protected] 
<mailto:[email protected]>> wrote:

Thanks now the system responds,
but as you can see in this gist (is the response for a simple query {"user": 
"<USER_ID>"}):
https://gist.github.com/conker84/12e58c5af470677d56f9a8bfa99608e8 
<https://gist.github.com/conker84/12e58c5af470677d56f9a8bfa99608e8>
The "score" field has a strange format.
I updated the model stored in elastic if you want to check it out:
https://gist.github.com/conker84/39734a8a6a5d9564449439b838ab8ab7 
<https://gist.github.com/conker84/39734a8a6a5d9564449439b838ab8ab7>
Il giorno lun 13 feb 2017 alle ore 19:32 Pat Ferrel <[email protected] 
<mailto:[email protected]>> ha scritto:
The date ranges are meant for examples and are used in the integration test but 
will block all your recommendations if you don’t use them. Start from 
engine.json.minimum instead. I’ll change this to be more clear in the next 
release. Specifically:

    "availableDateName": "available",
    "expireDateName": "expires",
    "dateName": "date”,

will cause trouble if you don’t set values to correspond in the query or item 
properties. Don’t use them until you need to.


On Feb 13, 2017, at 10:04 AM, Andrea Santurbano <[email protected] 
<mailto:[email protected]>> wrote:

Hi Pat,
following my engine.json:
https://gist.github.com/conker84/11a76a1644aaffeaa286c70c6038c3c8 
<https://gist.github.com/conker84/11a76a1644aaffeaa286c70c6038c3c8>
Thanks
Andrea

Il giorno lun 13 feb 2017 alle ore 17:34 Pat Ferrel <[email protected] 
<mailto:[email protected]>> ha scritto:
Then the config is wrong so send engine.json. Did you leave the date ranges and 
other example config in the engine?


On Feb 13, 2017, at 12:34 AM, Andrea Santurbano <[email protected] 
<mailto:[email protected]>> wrote:

Hi Pat!
Thanks for the quick response, the problem is that i get an empty response even 
if i made a simple empty query like "{}"... :(



Il giorno dom 12 feb 2017 alle ore 21:00 Pat Ferrel <[email protected] 
<mailto:[email protected]>> ha scritto:
The model looks ok. It looks like you have “purchase” as the conversion event 
and “categories” properties for items. 

Since the popularity is being calculated you can only get a blank result if you 
have supplied a filter in the query (that disallows everything) or there is 
some other config problem. I bet you are specifying a category filter that is 
ruling out everything in some way.

Can you supply a sample query JSON?

 
On Feb 12, 2017, at 10:56 AM, Andrea Santurbano <[email protected] 
<mailto:[email protected]>> wrote:

Hi community,
i'm testing the UR but i'm facing some problem; after the deploy of the 
algorithm if i try some simple queries, like the most popular items for 
instance, the result is always empty.
I tried to debug the model like explained in this page:
http://actionml.com/docs/ur_elasticsearch_debugging 
<http://actionml.com/docs/ur_elasticsearch_debugging>
but i can't figure out where is the problem.
At this url a sample of the model:
https://gist.github.com/conker84/39734a8a6a5d9564449439b838ab8ab7 
<https://gist.github.com/conker84/39734a8a6a5d9564449439b838ab8ab7>
Thanks for the help!
Andrea






Reply via email to