Filtering Ideas In Machine Learning problems

If the question has other problems (such as being a homework dump), guide them to improve their question before reposting it. Among several ideas that come to mind throughout the research process, how do I filter them and evaluate my assumptions before experiments and coding? Is there any general hands-on justification?