Geek asks and answers

The value of properly classified weka instances

I recently started using weka, and I'm trying to classify tweets as positive or negative using Naive Bayes. So, I have a training set with tweets for which I gave a shortcut, and a test set with tweets that have a “positive” shortcut. When I ran Naive Bayes, I get the following results:

Correctly classified specimens: 69 92% Incorrectly classified specimens: 6 8%

Then, if I changed the tweet labels in the test set to “negative” and ran to Naive Bayes again, the results will be inverted:

Correctly classified specimens: 6 8% Incorrectly classified specimens: 69 92%

I thought that properly classified instances show Naive Bayes accuracy and that it should be the same regardless of the tweet labels in the test set. Is there something wrong with my data or am I misunderstanding the meaning of correctly classified instances?

Thanks so much for your time,

Nantia

+5
source share
2 answers

Labels on the test case must be correct. Performance is calculated by asking the classifier to give a better idea of ​​the classification for each instance in the test case. The predicted classifications are then compared with actual classifications to determine accuracy. Therefore, if you flip the “correct” values ​​that you give them, the results will also be flipped.

+5
source

, 69,92% . , , , , 69,92%. (, , ) , , , , .

, . . , , , . , , , .

+1

More articles:


All Articles