The imbalanced may happen because the unnecessary features where the selection reduce these features and remove the non useful ones, this can help the classifier to be balanced and trained better.

I think feature selection/feature reduction and balancing imbalance classes are two different things. Feature selection/feature reduction procedure eliminate unnecessary variables from the dataset (e.g. recursive feature elimination in sklearn library or stepwise regression) or transform the dataset into lower dimension while preserving maximum information (e.g. PCA). Feature elimination/reduction helps classifier to avoid learning from redundant information and overfitting. Contrarily, unbalance class in the target variable makes the classifier biased towards the particular classes which have maximum samples. To overcome biasness, you can apply sampling techniques such as oversampling or underdamping or combination of both. Here is an useful resource about sampling techniques of unbalance data -

https://machinelearningmastery.com/data-sampling-methods-for-imbalanced-classification/

I agree with Md Didarul Islam. Feature selection optimizes the input to your model and can improve the model robustness, speed and generalization. Class imbalance has to do with the sample class distribution in your training dataset. There are various methods to deal with class imbalance including sampling techniques and giving different weights to each class in the cost function.