[
https://issues.apache.org/jira/browse/PIO-113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16149337#comment-16149337
]
Ananda Verma commented on PIO-113:
----------------------------------
One approach which I can suggest you is logistic regression. Assuming you have
training data as
||User||Question||Answer||
|User1|What is Apache PredictionIO|Open Source Machine Learning Server
You can convert question into sentence vector and train your logistic model
considering users as output class.
Other approach can be taking sentence vector of posted question and finding
cosine distance between all answered question, the nearest question will be
similar and hence user who had given answer is highly likely answer that new
question.
> pridection of user who can able to give answer of posted question based on
> whole mysql data
> -------------------------------------------------------------------------------------------
>
> Key: PIO-113
> URL: https://issues.apache.org/jira/browse/PIO-113
> Project: PredictionIO
> Issue Type: Improvement
> Components: Templates
> Affects Versions: 0.11.0-incubating
> Environment: Linux centos 7 64 biy
> Reporter: Dipen Patel
> Labels: features
> Fix For: 0.12.0-incubating
>
>
> hello..
> i am looking below feature
> prediction of answer based on past user data from whole database.
> So can i used any existing template or required to develop new template ?
> i have find some template but i dont think it is helpful to me.
> Thanks
> Dipen
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
