How Popular Is Your Dog?

                                                                                                By: Justin W Brown

We all have that one photo of our little "Man's Best Friend" that we just can't resist showing everyone when the topic of dogs comes up in conversation. We just can't get enough of the cuteness contained inside the borders of the image and we can't wait to hear the chorus of "Awe's" that will come from those who cast their eyes upon it. But are they just being nice because you are standing in front of them? Or is your pooch really the emitome of everything that defines adorable?

Now you have a way to find out once and for all if your friends are just humoring you or if you have the next Lassie running around your home. Through the use of social media, WeRateDogs® has provided a mechanism for you to get your photo in front of the masses and see where the chips (aka Likes and Retweets) may fall.

Utilizing data from Twitter, analysis of over 5000 tweets was perfomed using image recognition and data analysis techniques. One thing became immediately clear. You retriever owners are either everywhere, or you really love to brag about your dogs (or both)! Based on the number of images in the data set in which the image prediction software had some level of confidence that it was looking at an actual dog, golden retrievers account for over 1200 total points awarded (out of 10), more than 1.5 times as many as any other breed. When the golden and labrador retriever gang up, together they outpace any other breed by 250% in terms of points awarded.

total_points_by_breed.jpg

Looking a level deeper, it becomes clear that, in fact, there are more retrievers represented in the data than any other breed, so, of course, the total number of points awarded is higher.

total_tweets_by_breed.jpg

Let's level the playing field, shall we? By normalizing this data by the number of tweets per breed, we will come up with the average points per tweet for the breeds. This will be a much better indicator of which breed may have claim to the throne.

average_points_by_breed.jpg

This paints a completely different picture! Now the Clumber Spaniel is the clear standout. Before we can award this breed the trophy for most popular breed, I wonder how many Clumber Spaniels are represented in this data.

average_points_table.jpg

So, one tweet on which the rating was 27 / 10 caused the Clumber to rise to the top, despite other breeds being much better represented in the data. This hardly seems fair.

Looking into the data a little more, there are other data points that may well represent a better measure of popularity. Once a person has Tweeted their photo, the world (yes, literally the world) can now react to this photo. There are two main ways that one can react on Twitter. You can "Favorite" a Tweet, or you can "Retweet" a Tweet. Confusing, I know. It seems that rather than an arbitrary, non-standard rating, the pure, unfettered emotions of the masses, clicking Favorite and Retweet to their hearts' content is probabaly a better indication of the popularity of a photo. And, when taken in total, there may yet be a breed that amasses the most popularity in this contest of cuteness. Lets first look at Retweets. And we will skip straight to averages to eliminate the built in bias by those pesky retriever owners.

average_retweets_by_breed.jpg

Now this looks quite interesting. The English Springer seems to be a clear crowd favorite! Averaging over 11,000 retweets, there is clearly something special about this particular breed... or at least its representation in the WeRateDogs world. (Don't get too excited yet, English Springer owners! While the world might be retweeting the photo, they don't know that your little pooch chewed up Grandma's 100 year old quilt right after you snapped the photo!)

To see if this may have been affected by a small sample size, let's look at the totals again, bringing this assortment into view.

average_retweets_table.jpg

Ok, so they have a better representation than the Clumbers did, but still not too impressive. However, the fact that Retweets ... in this case an average of over 11,000 ... represent many opinions, there must still be something about these photos that is catching the eye of Social Media.

Let's take a look at one more data point before we make our final decision about the most popular breed. This one may be even more helpful, because Favoriting a Tweet can be done "in the privacy of one's own cyberworld," whereas Retweeting requires you to put yourself out there for the world to judge. And what if the social media world disagrees with your assessment of a Retweetable Tweet? You cyber reputation can only take a few missteps like this! So Favorites being a more discrete way to cast your vote, we may see even more opinions, and thus another opportunity for a standout breed. I wonder how the English Springer will fare!

average_favorites_by_breed.jpg

It seems the race has tightened, but the English Springer was able to hold on to the lead averaging over 22,000 likes per Tweet! Sure it is a small number of photos, but for a breed to garner that much attention with so few photos seems to bolster its potential claim to the most popular breed!

So, just maybe, the question isn't most popular breed, but instead, who what the most popular Tweet. Taking a look at the tweets that obtained the most favorites, me seem to be narrowing in on the outliers that are skewing our data.

maximum_fav_by_breed.jpg

The picture changes again, with Chihuahua and French Bulldog taking a commanding lead, but the Springer is hanging in there. So, without further ado, let's take a look at the images that have risen to the top and give our breed owners something to be proud of!

Chihuahua

Wait a minute!!!

French Bulldog

Ummmm.......

English Springer

Ok, I am seeing a pattern here. Chihuahua = Corgie. French Bulldog = Boston Terrier. English Springer = Basset Hound.

OK... so this entire thing has been a ruse brought about by the machine learning that attempted to recognize the breed!! It seems before we can use analysis to award the most popular breed we need to work on image recognition.

Sorry for the confusion... I have to go now! My Pit Bull is chewing up Grandma's 100 year old Quilt!!

Justin is a Program Manager for Software Development at AT&T. He and Bo are studying to becme Data Scientists because they are confident that some day, on some unsuspecting Analyst's screen will be displayed the cure to cancer, even though they were only trying to analyze photos of dogs.

Image sources: WeRateDogs

If you want to see the original Tweets that the above images were taken from, you can find them at the following URLs:

https://twitter.com/dog_rates/status/807106840509214720/video/1

https://twitter.com/dog_rates/status/866450705531457537/photo/1

https://twitter.com/dog_rates/status/879415818425184262/video/1