mattprout
commented
6 years ago Per your suggestion, changed the focus of the model to provide recommendations on how to improve the outcome of all animals. Added a sentence to discuss a future enhancement to match people with animals to adopt.
The idea is to help animal shelters utilize their limited resources effectively to help the most animals possible by creating a model that enables shelters to predict whether an animal will be adopted, returned, or euthanized based on the characteristics of the animal and the circumstances under which it was brought to the shelter.
I like the beginning. It very clearly identifies the problem and tells the reader why it matters (6.5 million animals, 1.5 million euthanized, etc) as well as identifying the client as animal shelters and their motivation (use limited resources to maximize good for animals).
In the middle, you describe the example of giving a sick animal veterinary help to avoid euthanasia and promote adoption. I'm unfamiliar with animal shelters, but is that not something the shelters would do already? It might be interesting to see if the model can help identify traits that are most likely to cause an animal to swing from "adopted" to "euthanized" status and have the shelters focus their time/energies on promoting the "adopted" traits and minimizing the "euthanized" traits. Maybe a limp in a dog's walk doesn't affect adoptability but having worms does so veterinary resources should be prioritized towards dogs with the latter issue over the former, etc.
Something that stuck out to me is that while the problem is defined as helping all the animals, the focus afterwards became about the euthanized animals. This is clearly the worst case outcome for animals in a shelter so it makes sense to highlight it but I think your model also can benefit animals that fall into the adopted category too so I think it would be good to mention that possibility too. The model could help increase the adoption rate of animals if the proper traits can be determined and prioritized which benefits the larger group. There's maybe even a possibility of expanding the model to include characteristics about adopters so animals can be shown to potential adoptees who are most likely to be compatible. My guess would be that there are too many animals in a shelter for an adoptee to see all of them and make a good decision so it would benefit the shelter to show the top few that are most likely. This might be tricky since adoptee data is probably harder to come by but food for thought.