Dear Mehdi Neshat ,

An ensemble-based method can be used to deal with imbalanced datasets. The belief is that multiple learning methods are more effective than a single one. It's an approach that combines the performance or results of many classifiers to better the performance of a single classifier.

https://www.section.io

Regards,

Shafagat

Many thanks Dear Professor Mahmudova for your great suggestion. May you probably please share a reference in this way?

Thanks in advance,

Good morning, to handle with imbalanced datasets i used K-fold cross validation. I split dataset in training and testing set (respectively 70%-30% or 80%-20% of prepared dataset). I divided the training set in 10 parts and i run models ten times using each part as a validation dataset. Thus, during the training, your model seen all data. Some people used also resampling method (I never used it so i cannot give you more information about it). Indeed, it is more interesting to use ensembles learning machine methods especially boosting technics that learn of weak classifiers.

Dear Dr Neshat,

For imbalanced datasets, proper selection of performance indicators is important. Instead of accuracy, we had to include parameter like F1 measure.

However, we can include the resampling techniques to deal with imbalanced data such as Synthetic Minority Oversampling Technique Tomek Link (SMOTE Tomek) and Synthetic Minority Oversampling Technique ENN (SMOTE ENN).

You can see the paper we published recently:

Article Churn prediction in telecommunication industry using kernel ...

Hope it helps.

I am a fan of your work anyway.

Regards,

Son

You can follow this paper (Article Vietnamese hate and offensive detection using PhoBERT-CNN an...

Mehdi Neshat you can use library https://imbalanced-learn.org/ which contain a lot of methods