[
https://issues.apache.org/jira/browse/MAHOUT-1423?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14056810#comment-14056810
]
Matteo commented on MAHOUT-1423:
--------------------------------
We tried to refactor the proposed interface to make it more general. The first
assumption we made, is that the focus is narrowed to time series anomaly
detection as described by the issue. The second assumption is that a time
series is represented as a Matrix, where each row corresponds to a time series
point, and each point is represented as a Vector of features. For this reason
the interface uses now Matrix to exchange the data.
Answering to the points made in the last feedback:
a) We improved the README and added a simple test, even though the tests should
still be improved.
b)
- the interface is used to provide a skeleton for time series anomaly detection
to the users, which uses the t-digest to spot the anomalies
- we changed the data format to be more general (from Vector to Matrix)
- the proposed interface is indeed not suitable for an on-line anomaly
detection; it is possible to discuss how to adapt it to consider this use-case
Any further comments will be appreciated!
> Add time series anomaly detection example
> -----------------------------------------
>
> Key: MAHOUT-1423
> URL: https://issues.apache.org/jira/browse/MAHOUT-1423
> Project: Mahout
> Issue Type: New Feature
> Components: Examples
> Affects Versions: 0.9
> Reporter: Andrew Musselman
> Assignee: Andrew Musselman
> Priority: Minor
> Fix For: 1.0
>
>
> Add an example of anomaly detection in a time series along the lines of Ted's
> EKG example starting about slide 20 here:
> http://www.slideshare.net/tdunning/strata-2014-anomaly-detection
--
This message was sent by Atlassian JIRA
(v6.2#6252)
