predictionio-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kenneth Chan <kenn...@apache.org>
Subject Re: Recommendation template, Implicit Feedback, only "buy" events
Date Thu, 16 Feb 2017 07:56:37 GMT
1. yes.
try follow this to modify template to use trainImplicit
http://predictionio.incubator.apache.org/templates/recommendation/training-with-implicit-preference/

2. just import event to PIO as
 {
event: buy
entitiy: user
entityId: uid
targetEntity: article
targetEntityId: article_id
eventTime: time of the buy event
}


On Tue, Feb 14, 2017 at 2:42 AM, Kasper Bjerke <kasper.bjerke91@gmail.com>
wrote:

> Hi
>
> I'm currently trying to set up PredictionIO with the standard
> Recommendation template as a project. I have 4 years of data, but the data
> I have can only qualify as "buy" events (I realise that each of the "buy"
> events can be interpreted as: first "view" then "buy") from E-commerce, so
> I have changed the code to match:
>
> http://predictionio.incubator.apache.org/templates/
> recommendation/training-with-implicit-preference/
>
> But instead of the sample "view" event I made it "buy". My questions are
> as follows:
>
> 1) Is it possible to use this template based off of only "buy" events?
> 2) And how should the testdata be formatted? Because all the examples I
> can find are based off of rating events: uid::movie_id::rating
>
> I was thinking: uid::article_id::"buy"? I'm fairly new to this, I have
> looked into Spark MLLib, but all the examples use the MovieLens dataset.
>
> Thank you,
> - Kasper
>

Mime
View raw message