During Covid-19, we saw a huge increase in the adoption rate of shelter animals. However, prior to Covid-19, it was difficult to convince people to rescue shelter animals. A lot of employees at animal shelters have described how the breed of different types of animals, age, and coloring of the animals had a significant impact on which animals were successfully adopted and which ones were not. As the world moves back in-person and the adoption rate hike is simmering down as people return to work, it could be worthwhile studying which animals are less desirable and how to help.
Using a small sample of data from the Austin Animal Center from October 1st, 2013 to March, 2016 studying the status of animals as they leave the no-kill Animal Center, certain relationships between types of outcomes when animals leave the shelter and their relevant attributes can be studied. These subtypes of outcomes include adoption, fostering, SCRP (now known as SNR or shelter, neuter, return), The goal of this project is to identify the relationship of an animal's species, breed, coloring, and age and their outcome to determine which animals should receive more attention when promoting adoption.
If successful, this can help animal shelters predict which animals may have a more difficult time being adopted. Therefore, they can spend more of their efforts on helping those particular animals find loving homes as opposed to overpromoting the animals that are more likely to be adopted. This will reduce the number of animals who are left without a home and can be implemented in shelters that do not have the no-kill policy that the Austin Animal Center has, saving lives.
We will use a Kaggle Dataset of Animal Adoptions from the Austin Animal Shelter to observe the following features for each song:
age_upon_outcome
animal_id
| age_upon_outcome | animal_id | animal_type | breed | color | date_of_birth | datetime | monthyear | name | outcome_subtype |
|---|---|---|---|---|---|---|---|---|---|
| 2 weeks | A684346 | Cat | domestic shorthair | orange | 2014-07-07 00:00:00 | 2014-07-22 16:04:00 | 2014-07-22T16:04:00 | Partner | |
| 1 month | A685067 | Cat | domestic shorthair | blue /white | 2014-06-16 00:00:00 | 2014-08-14 18:45:00 | 2014-08-14T18:45:00 | Lucy | |
| 3 months | A678580 | Cat | domestic shorthair | white/black | 2014-03-26 00:00:00 | 2014-06-29 17:45:00 | 2014-06-29T17:45:00 | *Frida | Offsite |
We are currently unclear as to how some of these outcome subtypes are classified, nor why some cells lack data and how to interpret them. We believe we can find more context clues in the description of the data set and from the Austin Animal Shelter website.
Of course, using this small subset of data from a single shelter during one period of time is not necessarily representative of animal shelters across the nation; however, this small step can lead to further data analysis and greater findings. Also, shelters tend to not be able to always accurately provide the correct data concerning animals since they often come in as strays without any background information. Therefore, breed, age, and date_of_birth could all be incorrect since they were most likely educated guesses.
We pose our problem as a regression (line of best fit) problem: given the listed features above (age, breed, color) we seek to estimate the popularity of which types of animals get adopted more frequently. One advantage of this approach is that it offers an intuitive output as each feature will explicitly be associated with some increase or decrease in adoption rates.
Our method is subject to change as we may find other methods may better suit our purposes in addition to a series of regressions.