Age | Commit message (Collapse) | Author | Files | Lines |
|
Change-Id: Iafdd12a573b58ab3ba9c69699cd97c5a4e598601
Signed-off-by: Tibor Frank <tifrank@cisco.com>
|
|
+ Do not declare BitCountingClassifier.classify() as class method.
+ Make BitCountingGroupList subclass of list.
+ Inherit from abstract classes whenever possible.
+ Drop unneeded imports.
+ Add module docstrings and class docstrings anywhere.
+ Add TODOs hinting at possible improvements.
Change-Id: Iccfff5c0e7be0607d6cfa74314083fcfe5a4d7d9
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Dashboard tables should now report anomalies from last week only.
+ Changed handling of Nan to report regression.
Change-Id: I624b0bc84a93702a31fc79fd670bd645b963f1f7
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|
|
+ Keep the original detection,
+ add the new one as subdirectory
(both in source and in rendered tree).
- The new detection is not rebased over "Add dpdk mrr tests to trending".
New detection features:
+ Do not remove (nor detect) outliers.
+ Trend line shows the constant average within a group.
+ Anomaly circles are placed at the changed average.
+ Small bias against too similar averages.
+ Should be ready for moving the detection library out to pip.
Change-Id: I7ab1a92b79eeeed53ba65a071b1305e927816a89
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
|