Pandas 数据框到 Seaborn 分组条形图

Question

I have the following dataframe which I have obtained from a larger dataframe which lists the worst 10 "Benchmark Returns" and their corresponding portfolio returns and dates:

I've managed to create a Seaborn bar plot which lists Benchmark Returns against their corresponding dates with this script:

import pandas as pd
import seaborn as sns

df = pd.read_csv('L:\\My Documents\\Desktop\\Data NEW.csv', parse_dates = True)

df = df.nsmallest(10, columns = 'Benchmark Returns')
df = df[['Date', 'Benchmark Returns', 'Portfolio Returns']]
p6 = sns.barplot(x = 'Date', y = 'Benchmark Returns', data = df)
p6.set(ylabel = 'Return (%)')
for x_ticks in p6.get_xticklabels():
    x_ticks.set_rotation(90)

And it produces this plot:

However, what I'd like is a grouped bar plot that contains both Benchmark Returns and Portfolio Returns, where two different colours are used to distinguish between these two categories.

I've tried several different methods but nothing seems to work.

Thanks in advance for all your help!

Answer 1

The trick is to transform the pandas df from wide to long format

Step 1: Prepare data

import seaborn as sns

np.random.seed(123)
index = np.random.randint(1,100,10)

x1 = pd.date_range('2000-01-01','2015-01-01').map(lambda t: t.strftime('%Y-%m-%d'))
dts = np.random.choice(x1,10)

benchmark = np.random.randn(10)
portfolio = np.random.randn(10)

df = pd.DataFrame({'Index': index,
                   'Dates': dts,
                   'Benchmark': benchmark,
                   'Portfolio': portfolio},
                    columns = ['Index','Dates','Benchmark','Portfolio'])

Step 2: From "wide" to "long" format

df1 = pd.melt(df, id_vars=['Index','Dates']).sort_values(['variable','value'])
df1

    Index   Dates   variable    value
9   48  2012-06-13  Benchmark   -1.410301
1   93  2002-07-31  Benchmark   -1.301489
8   97  2005-01-21  Benchmark   -1.100985
0   67  2011-06-01  Benchmark   0.126526
4   84  2003-09-25  Benchmark   0.465645
3   18  2009-07-13  Benchmark   0.522742
5   58  2007-12-04  Benchmark   0.724915
7   98  2002-12-28  Benchmark   0.746581
6   87  2009-02-07  Benchmark   1.495827
2   99  2000-04-21  Benchmark   2.207427
16  87  2009-02-07  Portfolio   -2.750224
14  84  2003-09-25  Portfolio   -1.855637
15  58  2007-12-04  Portfolio   -1.779455
19  48  2012-06-13  Portfolio   -1.774134
11  93  2002-07-31  Portfolio   -0.984868
12  99  2000-04-21  Portfolio   -0.748569
10  67  2011-06-01  Portfolio   -0.747651
18  97  2005-01-21  Portfolio   -0.695981
17  98  2002-12-28  Portfolio   -0.234158
13  18  2009-07-13  Portfolio   0.240367

Step 3: Plot

sns.barplot(x='Dates', y='value', hue='variable', data=df1)
plt.xticks(rotation=90)
plt.ylabel('Returns')
plt.title('Portfolio vs Benchmark Returns');