You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@samoa.apache.org by Yu Gong <gy...@163.com> on 2016/01/11 18:01:18 UTC

Some question about the prediction of VHT in SAMOA

Hi,


Recently, I am trying to write some codes so that it can be easier to serialize SAMOA machine learning model such as Hoeffding Tree, Clustream and so on. And make it easier to use the model in some external environment. I think it's quite useful.


However, I meet some problem about Hoeffding Tree in SAMOA. I print the **prediction** of each class (which is a double array) for some instances, and the true number of class for each instance is 2.
But some **predictions** only contains 1 score, and others contains two. I am wondering what is the meaning about the **predictions** containing 1 score. You can see in the following figure.
               


I am looking forward hearing from you. And thanks all of you for building this awesome platform for streaming data mining.


Thanks,
Yu

Re: Some question about the prediction of VHT in SAMOA

Posted by Nicolas Kourtellis <nk...@gmail.com>.
Hi Yu,

Thank you for your interest in the SAMOA platform.

Unfortunately, the pictures you attached did not pass through and it is not
easy to understand the problem.

Can you paste the actual output in text format, instead?

Thank you!

Nicolas


On Mon, Jan 11, 2016 at 6:01 PM, Yu Gong <gy...@163.com> wrote:

> Hi,
>
> Recently, I am trying to write some codes so that it can be easier to
> serialize SAMOA machine learning model such as Hoeffding Tree, Clustream
> and so on. And make it easier to use the model in some external
> environment. I think it's quite useful.
>
> However, I meet some problem about Hoeffding Tree in SAMOA. I print the
> **prediction** of each class (which is a double array) for some instances,
> and the true number of class for each instance is 2.
> But some **predictions** only contains 1 score, and others contains two. I
> am wondering what is the meaning about the **predictions** containing 1
> score. You can see in the following figure.
>
>
> I am looking forward hearing from you. And thanks all of you for building
> this awesome platform for streaming data mining.
>
> Thanks,
> Yu
>
>
>
>



-- 
Nicolas Kourtellis