How great would it be to live in a world where every pet has a loving, caring home?
Millions of unwanted dogs and cats end up in animal shelters across the nation every year. And while many of them are adopted into loving homes, adoption rates still need improvement.
But how do we do this?
The Austin Animal Shelter in Austin, Texas, houses over 18,000 animals a year, making it the largest No-Kill shelter in the country. Their shelter data is public and recently, in honor of National Adopt-a-Shelter Pet Day, Thinkful data science student Joanne Lin decided to determine how shelters like Austin Animal Shelter could improve their adoption rates.
She did this by using data to answer the question, “what makes an animal more likely to be adopted?”
Is there information on adoptable animals that impacts adoption rates? To better understand what data was already available, Joanne conducted an exploratory analysis.
Since not much is known about what factors affect adoption rates, Joanne started by using feature engineering to create data sets to better understand why animals may or may not be adopted at a certain rate. Rather than looking at broad attributes like species or breed, Joanne needed to delve deeper into features like coat color and coat pattern.
She also looked at whether or not animals had names, and if those names affected adoption rates.
To better understand factors impacting adoption rates, Joanne used a machine learning algorithm known as the “random forest classification.” This algorithm calculates the relationship between different variables so that decision trees can be created and evaluated. Decision trees visually depict the possible outcomes of a decision-making process. In this case, a decision tree illustrates the series of decisions that result in either adoption or no adoption. The ultimate outcome (adoption or no adoption) was determined using a more advanced processing model known as a “boundary.”
As you might imagine, Joanne found that fur color and breed affect adoptability, but those were by no means the only criteria. The analysis also uncovered some fascinating associations that should be utilized by shelters as part of adoption campaigns.
Joanne’s findings may surprise you, especially because the factors that impact adoptability differ between dogs and cats.
For dogs, breed was the most important feature, followed by whether they were spayed or neutered. For cats, breed was less important than whether the cat was spayed or neutered, but having a name was the second most important feature in the eyes of prospective adopters.
You may even be surprised to learn that the adoptable pets’ names factored into adoptability, at least for Austin-area animals. The most common dog names were Max, Bella, and Daisy, but Ginger was the dog name that resulted in the most adoptions. Among cat names, Cookie was the clear winner. Of 24 adoptable cats with this name, 21 were adopted -- an astonishing 87%.
The animal’s age also played an important role: two month old cats and dogs were the most popular. Of course, dogs and cats of other ages were adopted, as well, but animals two months old were the most likely to be adopted.
Even the day of the week was a factor, with the majority of adoptions occurring during the weekend for both cats and dogs. Other time-related data points factored into adoption rates as well. For example, the length of time that animals had been waiting to be adopted factored into adoption rates for both cats and dogs.
With this information, shelters all across the country can now determine the best course of action to maximize their animal adoptions. And while you can’t name every dog Ginger or every cat Cookie, the data does suggest ways of increasing a pet’s potential adoptability, such as giving each cat a name.
These findings are important, not just for animal adoption in Austin, but across the country, where millions of animals wait in anticipation of finding their own “forever home”.
As you can see, data science isn’t just relegated to the back office of the server farm, nor is it the express domain of the engineer or researcher. At companies like Thinkful, there are students leveraging the latest tools and discoveries to make data more meaningful and impactful across a wide variety of industries.
Thinkful’s programming enables everyday people to use the power of data science to make a real-world impact on the issues they care about most, including helping animals in need.