Seaborn bar graph after group by to create top N x label in pandas

Question

I want to create a seaborn bar graph, problem is before creating the bar graph I need to create database based on column value(descending). For further explanation below is the data type

City	Complaint Type	Value
ARVERNE	Blocked Driveway	50
ARVERNE	Derelict Vehicle	32
ARVERNE	Disorderly Youth	2
ARVERNE	Drinking	1
ASTORIA	Animal Abuse	170
ASTORIA	Bike/Roller/Skate Chronic	16
ASTORIA	Blocked Driveway	3436
ASTORIA	Derelict Vehicle	426
BAYSIDE	Animal Abuse	53
BAYSIDE	Blocked Driveway	514
BAYSIDE	Derelict Vehicle	231
BAYSIDE	Disorderly Youth	2
BELLEROSE	Animal Abuse	15
BELLEROSE	Bike/Roller/Skate Chronic	1
BELLEROSE	Blocked Driveway	138
BELLEROSE	Derelict Vehicle	120
BREEZY POINT	Animal Abuse	2
BREEZY POINT	Blocked Driveway	3
BREEZY POINT	Derelict Vehicle	3
BREEZY POINT	Illegal Parking	16

Now I want to create graph with top 3 city with complaint and there major complain type, so if I use top 2 complain type then my data should look like the below

City	Complaint Type	Value
ASTORIA	Blocked Driveway	3436
ASTORIA	Derelict Vehicle	426
BAYSIDE	Blocked Driveway	514
BAYSIDE	Derelict Vehicle	231
BELLEROSE	Blocked Driveway	138
BELLEROSE	Derelict Vehicle	120
ARVERNE	Blocked Driveway	50
ARVERNE	Derelict Vehicle	32
BREEZY POINT	Illegal Parking	16
BREEZY POINT	Derelict Vehicle	3

Here you can clearly see that data is sorted/group by City but values are in descending order, plus only 2 major complain are selected. Can you please help on how to build this data/or plot a graph in pandas

I have tried few code where I can select the top 2 complain by city but unable to sort the city based on the values. Even when I sort the data based on values then I lose the group by. Below is the code I am currently using

df1 = df.groupby(['City','Complaint Type']).size().reset_index(name = 'size')   
df2 = df1.sort_values(by = ['City', 'size'], ascending = [True, False]).groupby('City').head(3)

Asked By: Vipin

||

Source

Answer 1

To get the dataframe the way you want, you need to first get the top 3 cities. To do this, you can sort and then call unique() on city column. Then sort the dataframe by these 3 cities (using sort_value) and get the top 2 complaints (using groupby and head()). This will give you the top 6 rows. If you want them grouped by city, you can use pd.categorical() on the data. Finally plot the graph using seaborn catplot. Code is below.

#Get top 3 cities in a list
top3=df.sort_values(by ='Value', ascending = False)['City'].unique()[0:3]
print(top3)

#Filter by these cities and pick top2 entries
df1=df[df['City'].isin(top3)].sort_values('Value', ascending = False).groupby('City').head(2)

#If you want the data grouped by city, then use categorical ordering
df1['City'] = pd.Categorical(df1['City'], top3)
df1.sort_values('City', inplace=True)
print(df1)

#Finally plot your graph
sns.catplot(data=df1, kind='bar', x='Complaint Type', y='Value', hue='City')

Note that you can exchange x and hue if you want to plot by each city in X axis

Outputs

top3

[‘ASTORIA’ ‘BAYSIDE’ ‘BELLEROSE’]

df1

         City    Complaint Type  Value
6     ASTORIA  Blocked Driveway   3436
7     ASTORIA  Derelict Vehicle    426
9     BAYSIDE  Blocked Driveway    514
10    BAYSIDE  Derelict Vehicle    231
14  BELLEROSE  Blocked Driveway    138
15  BELLEROSE  Derelict Vehicle    120

Plot

Answered By: Redox

Seaborn bar graph after group by to create top N x label in pandas

Question:

Answers: