Include majority class and random baselines in reports

[reckart]

Would it be possible to always include the majority baseline and random baseline in experiment reports? It may not be obvious to the user that even high fscores are essentially worthless until she looks at the actual prediction files and discovers that the classifier essentially just predicts the majority class.

[Horsmann]

I think this should be doable. I will have a look next week. I like the idea.

• [x] CrfSuite
• [x] Weka
• [x] Lib*
• [x] DeepLearning-DyNet
• [x] DeepLearning-Keras
• [x] DeepLearning-Dl4J
• [x] Excel-Output BatchTrainTestReport
• [x] Excel-Output BatchCrossValidationReport

[Horsmann]

What would be a reasonable majority value for regression, i.e. mean value of all encountered regression values? Or would one rather have no majority value in case of regression? @zesch

[zesch]

doesn't make sense for regression in my view

2018-03-08 11:14 GMT+01:00 Tobias Horsmann notifications@github.com:

What would be a reasonable majority value for regression, i.e. mean value of all encountered regression values? Or would one rather have no majority value in case of regression? @zesch https://github.com/zesch

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/dkpro/dkpro-tc/issues/454#issuecomment-371443332, or mute the thread https://github.com/notifications/unsubscribe-auth/ACkQ4EgtjlcfDDYS-WfFgmuxiYuWQQ5hks5tcQSGgaJpZM4Sa2f5 .

[Horsmann]

@reckart should be available now