Here are some tips to make a choice :

First, there should be a problem to solve with respect to some applications, second, methodologies for solving the problem should demonstrate some kind of innovation that is not seen in earlier works, third, there should be some kind of improvement vis-a-vis the state-of-the-art (preferably there should be some metric to measure this objectively). Finally, the improved results should have universal implications i.e. the improvements should not be dependent on some specific dataset.

Firstly, it depends on your research area and what it is based on. Secondly, are you targeting MSc or PhD? As PhD research is more rigorous compare to MSc. In either case, most especially in PhD research, you need to narrow down your area of research to a microscopic level. Because in the end, you are an expert in your field. That means you need to go through relevant literature to see what others have done, what is missing, gaps, etc. Then…If you need more info please let me know.