I did some interesting work in this space. With several thousand seasons of data I used an MLP network to visualize phenotypic plasticity (how plants react to a range of environmental conditions).
Accurately cleaning the input data proved to be extremely important, because there's a tremendous amount of "noise" at the individual level when dealing with living organisms, so lots of high-quality data is necessary to tease out relationships. Establishing causality was also important, considering the potential for confounding variables.
It also gave me a chance to brush up on my React/front end skills, but that was more ancillary.
Accurately cleaning the input data proved to be extremely important, because there's a tremendous amount of "noise" at the individual level when dealing with living organisms, so lots of high-quality data is necessary to tease out relationships. Establishing causality was also important, considering the potential for confounding variables.
It also gave me a chance to brush up on my React/front end skills, but that was more ancillary.
https://sproutling.ai/blog/harvest-simulations?jm2
https://sproutling.ai/blog/growth-simulations?jm2