May i know the detail regarding the Train numeric prediction in data flow. like : * What is minimum node size means? * What is maximum Depth means? * What is Maximum confidence means? * What is Train Partition percent change going to impact to? * What is Balancing method under sample, over sample, none explanation?

Oracle Communities

Oracle Analytics and AI

Child Item

Detail about Machine Learning model in dataflow

Ferdy Hermawan

May i know the detail regarding the Train numeric prediction in data flow. like :

What is minimum node size means?
What is maximum Depth means?
What is Maximum confidence means?
What is Train Partition percent change going to impact to?
What is Balancing method under sample, over sample, none explanation?

Find more posts tagged with

Machine Learning

Accepted answers

Lalitha Venkataraman-Oracle

Please refer to the following blogs that describe many of these hyper parameters:

https://oracledataviz.blogspot.com/search?q=random+forest

https://oracledataviz.blogspot.com/search?q=hyper

Philippe Lions-Oracle

Hi @Ferdy Hermawan if you are interested in training ML models without going throug that tuning manually, you should leverage the Auto ML capability of OAC. Check this video as a brief example for this :

https://youtu.be/Cgw1LLH77Go

if not, if you need the manual process, in addition to the blogs above, this old video should still show most of the process :

https://youtu.be/0qmsJgxFea4

Philippe Lions-Oracle

Hi @Ferdy Hermawan if you are talking about the 'Explain' feature itself, yes, it's included in any OA instance by default (no plugin or extra install). It currently works with Datasets, but we are enhancing it to support Subject Areas next calendar year. The following short videos will give you a brief overview :

https://youtu.be/mcQ9MGdcKoA

https://youtu.be/jmIFsPzwtTs

https://youtu.be/kfKZg3eoSjo

let us know if not what you meant

thanks

All comments

Steve F-2

Hi @Ferdy Hermawan,

Welcome to the community!

You would need to have a basic understanding of the chosen predictive algorithm (i.e. CART, etc.) to be able to understand those values.

Other comments, welcomed.

@Lalitha Venkataraman-Oracle

@Philippe Lions-Oracle

Benjamin Arnulf-Oracle

I would recommend to start with the documentation:

https://docs.oracle.com/en/cloud/paas/analytics-cloud/acubi/create-and-use-oracle-analytics-predictive-models.html#GUID-F0FEED7B-FD45-4194-B738-7C116295A900

Lalitha Venkataraman-Oracle

Please refer to the following blogs that describe many of these hyper parameters:

https://oracledataviz.blogspot.com/search?q=random+forest

https://oracledataviz.blogspot.com/search?q=hyper

Philippe Lions-Oracle

https://youtu.be/Cgw1LLH77Go

if not, if you need the manual process, in addition to the blogs above, this old video should still show most of the process :

https://youtu.be/0qmsJgxFea4

Ferdy Hermawan

https://community.oracle.com/products/oracleanalytics/discussion/comment/61696#Comment_61696

Many Thanks. finally there is explanation regarding this parameter.

Ferdy Hermawan

https://community.oracle.com/products/oracleanalytics/discussion/comment/61713#Comment_61713

Hi, thanks for the information.

May i ask for more? in the first Video there is auto ML feature where it can show extra menu that start with "explain <attribute from subject area>", is this feature already embedded in the current OA or i need to upload any plugin to enable the "explain <attribute from subject area>" menu?

Philippe Lions-Oracle

https://youtu.be/mcQ9MGdcKoA

https://youtu.be/jmIFsPzwtTs

https://youtu.be/kfKZg3eoSjo

let us know if not what you meant

thanks