Context: A paper has been conditionally accepted (subject to my major revision). Main issues that have been raised by two reviewers are:
The author should compare the testing times of different algorithms.
Some methods can be added for comparison.
Negative example used in experiment is 50%. According to the size of the dataset, the algorithm may overfit.
The algorithm should be evaluated according to the protocol of a known dataset.
It is clear that the main issues are related to the experiment.
Question: Is it mandatory to conduct new experiments to address issues made by reviewers? If I do not conduct new experiments, what is the probability of getting rejected?
I mean if reviewers are not convinced with experiments, the editor is supposed to reject the paper, but he didn't do that.
What should be done under such circumstances?