I have seen a lot of questions on tagged with machine learning that are along the lines of "here is my ML training script and a link to a dataset. I am getting x% accuracy, and how do I improve it?". I am not sure how to answer these kind of questions.
The description for the machine learning tag says
NOTE: If you want to use this tag for a question not directly concerning implementation, then consider posting on Computer Science, Cross Validated, Data Science, or Artificial Intelligence instead. Otherwise you're probably off-topic.
However, I am not sure what exactly includes implementation. If someone is asking questions that are clearly conceptual in nature, voting to close as 'off-topic' it and leaving a comment asking to post it on some other Stack Exchange site is probably the way to go. However, I am not sure if questions asking ways to get performance improvement would be welcome there as well.
In my opinion, these should be flagged as "too-broad" as it's not possible to give a clear single answer without a lot of experimentation which just amounts to doing the work of the OP for them. On the other hand, if we try to answer in terms of "possible things to try", that will be inviting somewhat opinion based answers since it's not possible to know what will actually work and what won't for the given data.
Or should these be tagged as "off-topic".
What is the way to go in these type of situations?
Should these type of questions be allowed on Stack Overflow?