Dear Suhail,

I would like to provide you some useful links that you may be interested in : Article The Relationship Between Variable Selection and Data Agument...

https://stackoverflow.com/questions/39265746/data-augmentation-techniques-for-general-datasets/39272735

https://arxiv.org/pdf/1609.08764.pdf

https://bair.berkeley.edu/blog/2019/06/07/data_aug/

I recomend

Article Bootstrap Methods

Article Bootstrap percolation

Article Analysis of cost data in randomized trials: an application o...

Article Bootstrap-DEA analysis of BRICS’ energy efficiency based on ...

Article Chapter 52 The Bootstrap

Article Bootstrapping clustered data

https://link.springer.com/content/pdf/10.1007/JHEP03(2014)100.pdf

https://amstat.tandfonline.com/doi/abs/10.1080/01621459.1994.10476768

https://projecteuclid.org/download/pdf_1/euclid.bj/1174324983

https://arxiv.org/ftp/arxiv/papers/1809/1809.04016.pdf

Augmenting the images is easier and simple as relationship between the pixels and label assignment can be maintained. Whereas, perturbing a dataset with categorical and numeric features can perturb the data sample into entirely different class. However, unsupervised machine learning algorithms can be used for randomly perturb features in each subset using the distribution’s mean and standard deviation as perturbation bounds.

Thanks for sharing the links Muhammad Ali and Cristian Ramos-Vera

Thank you Ritika Lohiya for the answer

Do you mean that the numeric data can be interpolated or extrapolated while keeping the mean and standard deviation within acceptable bounds?

Can you share a research publication or a link for the same?

Please refer the following link, this might help:

https://towardsdatascience.com/augmenting-categorical-datasets-with-synthetic-data-for-machine-learning-a25095d6d7c8

Ritika Lohiya the link is indeed helpful.

Yes,

I already applied it on numerical dataset using varitional auto-encoder..

Here you can find the details and the python code...

Conference Paper Industrial Internet of Things Based Ransomware Detection usi...

Code Python code Deep Stacked Varitional Neural Network

Dear Muna Al-Hawawreh

Thanks for sharing your publication and the python code.

Yes, data-augmentation techniques are useful in the unbalanced-data area. Generative Adversarial Networks (GAN) can generate realistic data, which is beneficial to train the model.

Thanks for the answer Xingjie Li