predictionio-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vaghawan Ojha <vaghawan...@gmail.com>
Subject Re: Cart/Complementary purchase recommendation with UR
Date Wed, 19 Apr 2017 11:02:11 GMT
Hi Pat,

I was trying to accomplish the cart recommendation also with properties in
the events, so my events looks something like this:

{"eventTime":"2016-01-01T07:19:17+05:41","entityType":"user","targetEntityType":"item","event":"purchase","entityId":"firstcartid","targetEntityId":"product12","properties":{"store_id":"2"}}
{"eventTime":"2016-01-01T07:06:51+05:41","entityType":"user","targetEntityType":"item","event":"purchase","entityId":"secondcartid","targetEntityId":"product23","properties":{"store_id":"3"}}
{"eventTime":"2016-01-01T07:06:51+05:41","entityType":"user","targetEntityType":"item","event":"purchase","entityId":"firstcartid","targetEntityId":"product101","properties":{"store_id":"3"}}

And I would query it like this:

curl -H "Content-Type: application/json" -d '
> {
>     "user": "CM/00074305/16",
>     "num":4,
>     "fields": [
>     {
>       "name": "store_id",
>       "values": ["2"]
>     }
>   ]
> }' http://10.10.10.2:8123/queries.json

It works without bias. But I don't think it's a valid way. I meant, if I'm
not wrong, I should have $set in the events, but if I use $set in the
events, then I won't have option to put both cart id as user and product id
as item.

Can you please add your suggestion?


On Sat, Apr 8, 2017 at 10:13 AM, Vaghawan Ojha <vaghawan781@gmail.com>
wrote:

> Hi Pat,
>
> Thanks I will try this in my existing app.
>
> Thanks
> Vaghawan
>
> On Thu, Apr 6, 2017 at 9:49 PM, Pat Ferrel <pat@occamsmachete.com> wrote:
>
>> Yes it can be done but support is not ideal so UR v0.6.0 coming just
>> after the release of PIO-0.11.0 (in RC now) has some enhancements that give
>> you better options and easier queries.
>>
>> To get truly complimentary purchases you need to train on shopping carts,
>> not users purchases. So the primary input would of the form:
>>
>> (shopping-cart-id, cart-purchase, product-id)
>>
>> Then in < 0.6.0 you query with the currently active shopping cart id as
>> if it is the user-id
>>
>> {
>>     “user”: “some-shopping-cart-id”
>> }
>>
>> In v0.6.0 you still need to train on shopping carts to get true
>> “complimentary purchase” but the query may be easier:
>>
>> {
>>     “Item-set”: [“product1”, “product2”]
>> }
>>
>> The query is the current contents of the shopping cart, no fake “user-id”
>> required.
>>
>> UR 0.6.0 will also allow you to query against a model trained on user
>> purchases with the same query but, though the results will be "things
>> purchased by people with similar purchasing behavior as you query
>> item-set", it won’t be strictly “complimentary purchases”.
>>
>> Since 2 methods are supported with UR 0.6.0 you could A/B test for which
>> has the best results. If using user purchase (instead of shopping carts) is
>> close to as good it would be much easier since you can have user-based,
>> item-based and shopping carts recs from the same model. Otherwise one for
>> user-based and item-based and another for “complimentary purchases” would
>> make 2 trains and 2 models.
>>
>>
>>
>> On Apr 5, 2017, at 8:08 PM, Vaghawan Ojha <vaghawan781@gmail.com> wrote:
>>
>> Hi Pat,
>>
>> As far as the actionml docs tells me that I could be able to use the UR
>> to recommend the cart/complementary purchase, but there seems to be no
>> documentation regarding this.
>>
>> It would be helpful if there was a documentation for this feature. I know
>> people have worked out and implemented the cart recommendation, but for
>> begineer like me, it would still be like "what to do now".
>>
>> So could you help me how can I do that? If documentation is not ready yet
>> for this, I could write one for this feature once I got what to do.
>>
>> Thanks
>> Vaghawan
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "actionml-user" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to actionml-user+unsubscribe@googlegroups.com.
>> To post to this group, send email to actionml-user@googlegroups.com.
>> To view this discussion on the web visit https://groups.google.com/d/ms
>> gid/actionml-user/CA%2B69RXYDeMRXOG25v%3Dadr%3DSXLk6BO811DSC
>> YX0t6vp8EVPP1Xw%40mail.gmail.com
>> <https://groups.google.com/d/msgid/actionml-user/CA%2B69RXYDeMRXOG25v%3Dadr%3DSXLk6BO811DSCYX0t6vp8EVPP1Xw%40mail.gmail.com?utm_medium=email&utm_source=footer>
>> .
>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>

Mime
View raw message