Criteria to get machine learning benchmarking papers published in ACM

Question

I have read lots of data analytics and machine learning related papers coming up with new algorithms, implementing models that outperform the existing one on a particular dataset. In case of publishing to a top journals on probably ACM, what kinds of difference should be created for getting accepted?

For example, on Liver cancer dataset, 5 researchers have published different models and the latest one outperformed the earlier ones. If I come up with a novel approach/algorithm and use some models to outperform the latest one, will it be considered publishable in any ACM journals?

Answer 1

What is publishable is up to an editor with advice from reviewers who are more or less experts in the particular subject. You need to write something that extends the state of the art enough that the reviewers suggest acceptance to the editor and the editor decides it is ripe for publication.

The only way to absolutely know that is to write the paper and submit it. But reading the most current literature will give you a pretty good idea of the state of the art. Collaborative relationships with other researchers in the field is even better.

Some things that make only minor improvements might not make the cut, especially if the process to achieve it offers little insight going forward. On the other hand, a tiny improvement might be significant if it offers new insight.

And, of course, every accepted paper moves the goal posts.

Do you best work. Write it up. Submit it. Then you will know.