|classifire Accuracy in openVibE
|Page 1 of 1|
|Author:||HA3 [ Fri Jul 03, 2015 7:32 am ]|
|Post subject:||classifire Accuracy in openVibE|
I build a system, gives a 88.7% test data accuracy(when building the classification model), But when I run the data on the classifier model and measure the accuracy with "classifier Accuracy measure" Box, the accuracy decreases to 40% ! Is that normal?
|Author:||fabien.lotte [ Tue Jul 21, 2015 1:32 am ]|
|Post subject:||Re: classifire Accuracy in openVibE|
you used the classifier accuracy measure on the same data in which you trained your classifier is that right?
If so, the fact you obtained decreased accuracy may come from the fact that you gave all your data (including the non-control state - i.e., the data between trial that do not correspond to one of the class the classifier has been trained on) to the classifier accuracy measure, and not only the data corresponding to one of the classes the classifier has been optimized to classify. Could that be the case? If so you need to use something like a stimulation based epoching to only use as input to the classifier accuracy measure box data that actually can be (and need to be) classified.
I hope this helps,
|Page 1 of 1||All times are UTC|
|Powered by phpBB® Forum Software © phpBB Group