A thesis requires a lot of work and thought. My advice is to find a way to make it fun and find a subject you are passionate about where you are able to use ML. Or if there is a problem you see in your life or in the general world that you feel ML is able to produce solution(s) for, do a little research on that to see if it is feasible to accomplish with ML.
Good luck and let me know if you have any specific questions,
Gerald (currently working on the final draft of his VR/AR Survey Research Paper)