Change format of MulticlassConfusionMatrix result to be more self-explanatory#48174
Conversation
66e1005 to
071cbd7
Compare
071cbd7 to
fa6ab5b
Compare
|
Pinging @elastic/ml-core (:ml) |
|
run elasticsearch-ci/packaging-sample-matrix |
4ce3aa3 to
97628e2
Compare
|
@elasticmachine update branch |
1 similar comment
|
@elasticmachine update branch |
2fd6087 to
dfecdf6
Compare
benwtrent
left a comment
There was a problem hiding this comment.
I really like the new format. It also gives us flexibility to add new fields in the future if we deem them necessary.
There was a problem hiding this comment.
According to the ctor, both of these are nullable.
...ticsearch/client/ml/dataframe/evaluation/classification/MulticlassConfusionMatrixMetric.java
Outdated
Show resolved
Hide resolved
There was a problem hiding this comment.
Should this be nullable? I think it would throw if predictedClasses == null
There was a problem hiding this comment.
might be good to do a otherActualClassCount >= 0 check.
There was a problem hiding this comment.
| this.otherActualClassCount = in.readLong(); | |
| this.otherActualClassCount = in.readVLong(); |
There was a problem hiding this comment.
| out.writeLong(actualClassDocCount); | |
| out.writeVLong(actualClassDocCount); |
There was a problem hiding this comment.
| this.count = in.readLong(); | |
| this.count = in.readVLong(); |
There was a problem hiding this comment.
| out.writeLong(count); | |
| out.writeVLong(count); |
cc20654 to
9b1c359
Compare
|
run elasticsearch-ci/packaging-sample-matrix |
9b1c359 to
92a6dd4
Compare
|
run elasticsearch-ci/packaging-sample-matrix |
Current MulticlassConfusionMatrix result can be confusing to the API user.
They have no means to tell if the outer map is actual or predicted. Similarly for inner map.
This PR fixes that by making the format strict (naming each field).
Relates #46735