[jira] [Commented] (METRON-389) Create Java API to Read Profile Data During Model Scoring

["Nick Allen (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/METRON-389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15434997#comment-15434997 ] 

Nick Allen commented on METRON-389:
-----------------------------------

When performing model scoring, a user will know the specific entity that needs scored along with the name of the profile.  

The user will also need to specify a time horizon over which the values should be read. A user may want to specify this as (1) a date range; Aug22 - Aug 24 (2) a time period; last 7 days (3) number of samples; the last 52 samples.  Not all need supported in this first draft.

Based on METRON-367, one profile may store its values as Integers while the other may be Doubles.  A user will need to also specify the type of values that they expect to receive from a profile.  

To earn a gold start, some level of conversion could also be performed.  For example, if a profile stores its values as Doubles, but a user doesn't need that level of precision, they could specify the read to return Integers.  Not necessary for first draft.

To summarize the following will be needed as input for performing a read operation.
1. Profile Name 
2. Entity Name - Ex: 10.0.0.1, Wires-App, webserver1.company.com
3. Time Horizon 
4. Data Type - Double.class, Integer.class

The output will be a collection of values of the specified data type.  Something like Double[], Collection<Double>, List<Double>.

> Create Java API to Read Profile Data During Model Scoring
> ---------------------------------------------------------
>
>                 Key: METRON-389
>                 URL: https://issues.apache.org/jira/browse/METRON-389
>             Project: Metron
>          Issue Type: Improvement
>            Reporter: Nick Allen
>            Assignee: Nick Allen
>
> Functionality already exists to create Profiles in HBase.  A Java API is needed to read existing Profile data from HBase.  The API should be optimized for use when scoring models.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)