Dear researchers
When the data set for a particular prediction is unbalanced( negative class is more than positive class OR negative class is less than positive class), Which evaluation parameter is appropriate to evaluate the performance of the predictive model? Why?
These parameters are including Specificity(SPE), Sensitivity(SEN), Receiver operating characteristics (ROC), Area Under Curve (AUC or AUROC), Accuracy(ACC), and f-measure(F-M)
Thanks for your guidance
I am waiting for your answer
Dear Shima Shafiee , you might want to cosider calcuating Cohen's Kappa. See for instance this paper:
https://www.knime.com/blog/cohens-kappa-an-overview
Shima Shafiee
I suggest Precision-Recall Metrics. The F-Measure is a well-known statistic for unbalanced classification.
Traditional model assessment approaches, on the other hand, do not effectively quantify model performance when confronted with unbalanced datasets. Standard classifier techniques, such as Decision Tree and Logistic Regression, favor classes with a large number of occurrences. They are only good at predicting data from the majority of classes.
DEAR Annika Hamachers
Thanks for sharing the link that is definitely effective and related to mentioned parameters.
GOOD LUCK
Qamar Ul Islam
I read your answers eagerly because they are very useful and complete
If you know the reference, please introduce it?
Thanks
GOOD LUCK
Hi Shima Shafiee
For unbalance data set most preferred evaluation matrix is ROC/AUC score.
For Unbalanced data use ROC/AUC, precision recall or use confusion matrix.
Hi
Ashish Mishra
Thanks for sharing your answer
and why? please explain if you have time
GOOD LUCK
Intersectoral distance, information distance, chi-squared
Hi
Eugene Veniaminovich Lutsenko
Thanks for sharing your answer and wonderful
- Badges
- Science topic