Seattle has a reputation for being a pet friendly city. By some estimates, there are more dogs in the city than there are children, an impressive feat for a place as populous as Seattle. Seattle's open data portal contains, among other things, information on licensed pets.

Awesome!

Let's explore the kinds of insights that can be found by looking at this data.

In [1]:

import pandas as pd
import matplotlib as mpl
import matplotlib.pyplot as plt
from matplotlib_venn_wordcloud import venn2_wordcloud

%matplotlib inline

In [2]:

license_file = "pet_licenses.csv"
# Alternatively, the license file may be read from here
# license_file = "https://data.seattle.gov/api/views/jguv-t9rb/rows.csv?accessType=DOWNLOAD"

pets = pd.read_csv(license_file)

Cleaning the data¶

Let's inspect the dataset to ensure it's properly cleaned before any analysis is run on it.

First, what does our data look like?

In [3]:

pets.shape  # (rows, columns)

Out[3]:

(53663, 7)

In [4]:

pets.columns

Out[4]:

Index(['License Issue Date', 'License Number', 'Animal's Name', 'Species',
       'Primary Breed', 'Secondary Breed', 'ZIP Code'],
      dtype='object')

Let's rename the columns so they're easier to work with

In [5]:

pets.columns = [
    "license_issued", "license", "name",
    "species", "primary_breed",
    "secondary_breed", "zipcode",
]

In [6]:

# how many NaNs does each colum have?
# https://stackoverflow.com/a/26266451
pets.isna().sum()

Out[6]:

license_issued         0
license                0
name                 709
species                0
primary_breed          0
secondary_breed    32083
zipcode              361
dtype: int64

Given the amount of NaN values in the secondary_breed column, most owners either don't know what breed their pet is mixed with, or left that section blank when applying for a license.

A fair number of licenses also don't have the name listed. It's possible these animals were either babies or recently adopted at the time of their licensing.

cute beagle puppy Cute!

In either case, let's replace the missing values from secondary_breed and drop the rest of the NaNs

In [7]:

pets = (pets
    .fillna({"secondary_breed": ""})
    .dropna()
)

The zipcode field is useful if we want to analyze pet information based on location. To make it easy, let's only use rows that contain a 5-digit ZIP code.

In [8]:

pets = pets[pets.zipcode.str.match("^\d{5}$")]

In [9]:

pets.describe()

Out[9]:

	license_issued	license	name	species	primary_breed	secondary_breed	zipcode
count	52444	52444	52444	52444	52444	52444	52444
unique	1060	52419	13977	4	335	256	93
top	January 16 2018	133239	Lucy	Dog	Domestic Shorthair		98115
freq	448	3	453	35188	9936	31467	4851

license_issued looks like it's a string - it would be more useful as a datetime object

In [10]:

pets["license_issued"] = pd.to_datetime(pets["license_issued"])

There are also duplicate licenses that we should get rid of

In [11]:

pets.drop_duplicates(subset=["license"], inplace=True)

Our data should be clean enough now, let's move on to the analysis.

Cats & Dogs¶

Unsurprisingly, cats and dogs are the most commonly licensed pet.

cat and dog together

although there are a few farm animals as well.

In [12]:

pets.species.value_counts()

Out[12]:

Dog     35170
Cat     17207
Goat       38
Pig         4
Name: species, dtype: int64

In [13]:

cats = pets[pets.species == "Cat"]
dogs = pets[pets.species == "Dog"]

When do owners register their pets?¶

We've all heard of the cliché of getting your loved ones a pet for Valentine's day, but apparently February is the least popular month to license a pet.

In [14]:

ax = (pets
 .license_issued
 .apply(lambda date: date.strftime("%m - %b"))
 .value_counts(normalize=True).apply(lambda percent: percent * 100)
 .sort_index()
 .plot.bar()
)

ax.set_ylabel("Percent of licenses issued", rotation=0, labelpad=60)
ax.set_xlabel("Month")

ax

Out[14]:

<matplotlib.axes._subplots.AxesSubplot at 0x7f7eb21a7748>

Most common names¶

If you're ever on the street and bump into a dog (or cat), your best bet is to call it Lucy.

In [15]:

cat_names = cats.name.value_counts()
dog_names = dogs.name.value_counts()

In [16]:

fig, ax = plt.subplots(1, 2, figsize=(15, 5))

cat_names.head(10).plot.bar(
    title="Most popular cat names",
    ax=ax[0],
)

dog_names.head(10).plot.bar(
    title="Most popular dog names",
    ax=ax[1],
)

plt.show()

In [17]:

fig, ax = plt.subplots(1, 1)

ax.set_title("Top cat and dog names")
venn2_wordcloud(
    [
        set(cat_names.head(10).index),
        set(dog_names.head(10).index),
    ],
    set_labels=("Cats", "Dogs"),
    ax=ax,
)

plt.show()

Breed¶

Retrievers (Labrador and Golden) are the most popular type of dog.

Domestic shorthair cats are the most popular cat by a wide margin.

In [18]:

fig, ax = plt.subplots(1, 2, figsize=(15, 5))

cats.primary_breed.value_counts().head(10).plot.bar(title="Top 10 cat breeds", ax=ax[0])
dogs.primary_breed.value_counts().head(10).plot.bar(title="Top 10 dog breeds", ax=ax[1])

Out[18]:

<matplotlib.axes._subplots.AxesSubplot at 0x7f7eb358e7b8>

Where are they?¶

Are there specific places within Seattle that have more pets than others?

In [19]:

pets.zipcode.value_counts().head(10)

Out[19]:

98115    4850
98103    4539
98117    3929
98125    3019
98107    2518
98118    2496
98122    2484
98199    2337
98116    2186
98105    2138
Name: zipcode, dtype: int64

In [20]:

import folium
import requests

SEATTLE_COORDINATES = (47.63, -122.27)
DEFAULT_ZOOM = 10
# Thank you, SeattleIO, for providing this!
ZIPCODE_GEO_URL = (
    "https://raw.githubusercontent.com/seattleio/seattle-boundaries-data/master/data/zip-codes.geojson"
)
ZIPCODE_GEO_TEXT = requests.get(ZIPCODE_GEO_URL).text

def plot_data_by_zipcode(data,
                         location=SEATTLE_COORDINATES,
                         zoom_start=DEFAULT_ZOOM,
                         geo_data=ZIPCODE_GEO_TEXT, 
                         key_on="feature.properties.ZCTA5CE10",
                         fill_color="PuBuGn",
                         **kwargs):
    """Plot data on a choropleth map
    
    With the exception of map_coordinates and default_zoom, all arguments
    in this function are passed to folium.Map.choropleth. map_coordinates
    and default_zoom arguments are passed in as parameters to the folium.Map
    constructor.
    
    Args:
        data - data to pass into the choropleth map. Corresponds to the
               data argument passed into folium.Map.choropleth and is usually
               a pandas DataFrame or Series.
        **kwargs - Additional arguments to pass to folium.Map.choropleth
        
    Returns:
        A folium.Map instance containing a choropleth of mapped data
    """    
    map_ = folium.Map(location=location, zoom_start=zoom_start)

    map_.choropleth(
        geo_data=geo_data,
        data=data,
        key_on=key_on,
        fill_color=fill_color,
        **kwargs,
    )
    
    return map_

In [21]:

plot_data_by_zipcode(
    data=pets.zipcode.value_counts(),
    legend_name="Pet Population by Zipcode",
)

Out[21]:

North Seattle, especially around the Green Lake area, has the highest population of pets. If you ever want to go dog watching, you can use this data as a guide to find the best place to do so!