Not that easy. I have sibo and depending on what I eat and how I eat it, it changes the effects it has on me. Also the amount and frequency. For example, if I have raisins at the start of my meal then I don't see issues, but if I have them at the end of my meal then I do. If I have it here and there then no issues, but if I have it a few meals in a row then it'll give me issues.

And raisins affect me within 20min, but if I were to face something like potatoes or tomatoes then it doesn't give me issues until hours later (after my next meal or even the next day).

> If people would log all what they eat and how they feel.

Dietary data is one of the hardest things to collect. Either you need to hospitalize people and control their whole diet, and your sample size is small, because of costs, or you need to rely on self reporting, and your data is unreliable.

Basically, much easier said than done.

Biological systems are highly nonlinear and time-dependent, and also very individual! People have been trying learning techniques in glucose control since the early 2000s, and have only seen limited success in very controlled scenarios. I think a simple neural net would fall short of catching all that complexity, at least using the current architectures

People won't log all the food they eat, but supermarkets actually have data on what food people buy.

So correlating that data with people's health could be very interesting!

I don't think that has much chance getting past an IRB though.