As we've established, this question is hard to answer because it's hard (impossible?) to define an objectively "correct" fingering. But I do think, for a given player, you can say that some fingerings are better, or more optimal, than others.
An aside: Musicians get touchy about correct usage of sharps and flats. For good reason, too, although it's not immediately obvious to most beginners. It's kind of like replacing the word "the" in every sentence with "coconut". I can eventually understand you, but it makes everything unnecessarily harder. This is not the place for a full explanation, but you can find one elsewhere on this site.
If I read between the lines, you're basically looking for truth data to train your model. I think you might be able to find a better source than static chord fingerings. Why?
- Because it's so context dependent. For example, if I'm playing root position C Major, I might use 1-3-5. Or I might use 1-2-4, if the next chord is a second inversion F Major.
- Because a lot of chords aren't major, minor or seventh triads, or inversions thereof.
- Because it's very dependent on the physical layout of each player. Big hands, different fingerings become optimal. You get the idea.
So, what then? I think your best source of truth data would be to get a series of pianists to play a real piece, and record their fingering. Note that they might actually change what they do on different runs of the same piece. I think this would be a much more "real-world" data source. If you can't get real live keyboard players, YouTube is your friend. Of course, the quality of player may vary.
Without specific knowledge of how you're training this model, this may not be practical advice. But I think you could train it on several different players, and see what comes out. It would actually be really interesting to compare the difference between two training sets. Lots of scope for future work if interested.
Of course, all of this might not work at all. But then you've learnt something. So what's to lose?