[
https://issues.apache.org/jira/browse/SPARK-8540?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15088746#comment-15088746
]
Rakesh Chalasani commented on SPARK-8540:
-----------------------------------------
I see that this hasn't moved forward, so trying to revive it. I will pick this
up.
After taking a fleeting glance at the KMeans API, we have two options:
1. Add this to KMeans/KMeansModel itself (which I don't like after what
[~josephkb] said above)
(or)
2. We need KMeansOutlier and KMeansOutlierModel as separate classes;
KMeansOutlier can extend KMeans itself with additional parameters for
supporting the above mentioned (a) and (b). KMeansOutlierModel might have to
duplicate some parts of KMeansModel
For (a) setThreshold/getThreshold param need to be added and can be implemented
using simple 'where'; (b) setNumOutliers/getNumOutliers param need to be added
and requires orderBy followed by take (or something better?). (a) takes
precedence over (b).
Please let me know your thoughts.
> KMeans-based outlier detection
> ------------------------------
>
> Key: SPARK-8540
> URL: https://issues.apache.org/jira/browse/SPARK-8540
> Project: Spark
> Issue Type: Sub-task
> Components: ML
> Reporter: Joseph K. Bradley
> Original Estimate: 336h
> Remaining Estimate: 336h
>
> Proposal for K-Means-based outlier detection:
> * Cluster data using K-Means
> * Provide prediction/filtering functionality which returns outliers/anomalies
> ** This can take some threshold parameter which specifies either (a) how far
> off a point needs to be to be considered an outlier or (b) how many outliers
> should be returned.
> Note this will require a bit of API design, which should probably be posted
> and discussed on this JIRA before implementation.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
