Timeline for How to determine the quality of a multiclass classifier
Current License: CC BY-SA 3.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 1, 2016 at 23:57 | answer | added | Ashish Markanday | timeline score: 2 | |
| Jun 9, 2014 at 4:12 | comment | added | drsealks | Someone can find this work helpful. The work is dedicated exactly to the problem of qualifying multivariate cases in classification tasks. arxiv.org/abs/1202.6221 | |
| Dec 6, 2012 at 20:09 | vote | accept | Gere | ||
| Dec 6, 2012 at 20:09 | vote | accept | Gere | ||
| Dec 6, 2012 at 20:09 | |||||
| Nov 27, 2012 at 16:01 | history | tweeted | twitter.com/#!/StackStats/status/273456620858638337 | ||
| Nov 27, 2012 at 15:41 | answer | added | Kyle. | timeline score: 21 | |
| Nov 27, 2012 at 15:39 | answer | added | Emile | timeline score: 23 | |
| Nov 27, 2012 at 15:20 | comment | added | Gere | I suppose the meaning of positive depends on the particular problem and the importance of the classes. I was playing around with different self-made measures. Maybe there is a more standardized approach. | |
| Nov 27, 2012 at 14:47 | comment | added | steffen | @MichaelMcGowan I have asked the OP for clarification and afterwards performed an edit to explicitly reflect the multiclass issue, which was not obvious before the edit (IMHO). | |
| Nov 27, 2012 at 14:36 | comment | added | Michael McGowan | @steffen Those derived measures are primarily applicable to binary classification, whereas this question explicitly is dealing with more than two classes. This then requires a modified understanding of terms like "true positive." | |
| Nov 27, 2012 at 13:41 | history | edited | steffen | CC BY-SA 3.0 |
added more formal definitions; replaced "real" with true; added multi-class tag
|
| Nov 27, 2012 at 13:24 | comment | added | Gere | @steffen: I've seen the confusion matrix. In my particular case I have 4 classes. So I'm not sure which derived measures can be used and would make sense. Each $x_i$ belongs to only one class. However there are more than two possible classes overall $i\in [1,\cdots,N]$. | |
| Nov 27, 2012 at 9:30 | comment | added | steffen | I think this is the source of my confusion: In the first paragraph you state ..where yi is the real classes and...: Do you mean that an instance $x_i$ can belong to / has more than one class ? Or does every $x_i$ belongs to / has exactly one class ? Can you please clarify ? | |
| Nov 27, 2012 at 9:27 | comment | added | steffen | I am not sure if I understand the question correctly. Do you know the Confusion Matrix and derived measures ? Is this the answer to your question ? Or do you refer to something more complicated ? | |
| Nov 23, 2012 at 12:46 | history | asked | Gere | CC BY-SA 3.0 |