Pandas: Visualization#

For more cf. Pandas Chart Visualization.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

Plotting Features#

plt.figure(figsize=(10,5))

x = np.arange(start =-10, stop = 10, step=0.5)
y = np.sin(.6*x)

plt.plot(x, y, '--b', linewidth=2, label='curve')

plt.title('my nice plot')
plt.xlabel('x [a.u.]')
plt.ylabel('y [a.u.]')
plt.legend(loc='best', numpoints=3)
plt.grid()

plt.savefig('abbildungen/my_plot.pdf')
_images/DM_Pandas_3_3_0.png

Random Walk with Series#

dates = pd.date_range('2000-01-01', '2001-01-01', freq='B')
ts = pd.Series(np.random.randn(len(dates)), index=dates)
ts.head()

2000-01-03    0.986342
2000-01-04    1.148306
2000-01-05   -1.725368
2000-01-06   -0.633592
2000-01-07   -0.086018
Freq: B, dtype: float64

plt.plot(ts);
_images/DM_Pandas_3_6_0.png

This can be plotted nicer using pandas plot method:

plt.figure(figsize=(8,6))
plt.grid()

plt.subplot(2,1,1)
ts.plot(label='raw')
plt.legend()

plt.subplot(2,1,2)
ts.cumsum().plot(style='s-r', label='cumsum')
plt.legend();
_images/DM_Pandas_3_8_0.png 
df = pd.DataFrame({'raw':ts, 'cumsum':ts.cumsum()}, columns=['raw','cumsum'])
df.tail()
raw cumsum
2000-12-26 1.507907 -13.362592
2000-12-27 -1.551516 -14.914108
2000-12-28 0.690827 -14.223281
2000-12-29 2.444438 -11.778843
2001-01-01 1.119174 -10.659669 
df.plot(figsize=(8,6), subplots=True, style=['.-r','o-g'], sharex=False);
_images/DM_Pandas_3_10_0.png 
df.plot(style=['b','r']);
_images/DM_Pandas_3_11_0.png

Bar Plots#

Bar Plots of Series:

s = pd.Series(np.random.rand(4), index=['A','B','C','D'])
s

A    0.716410
B    0.723100
C    0.286061
D    0.967280
dtype: float64

s.plot(kind='bar');
_images/DM_Pandas_3_15_0.png 
s.plot(kind='barh', color=['r','b','g','y']);
_images/DM_Pandas_3_16_0.png

*Bar Plots of Data Frames:

df = pd.DataFrame(np.random.rand(6,4), 
                  index=['one','two','three','four','five','six'],
                  columns=['A','B','C','D'])
df
A B C D
one 0.547516 0.919091 0.749656 0.489452
two 0.126288 0.566826 0.665307 0.647030
three 0.935753 0.633396 0.916266 0.805464
four 0.038519 0.676239 0.241718 0.808550
five 0.578486 0.355255 0.991703 0.161532
six 0.424310 0.396545 0.403273 0.621927 
df.plot(kind='bar');
_images/DM_Pandas_3_19_0.png 
df.T.plot(kind='bar');
_images/DM_Pandas_3_20_0.png 
df.plot(kind='bar', stacked=True, alpha=0.5);
_images/DM_Pandas_3_21_0.png 
df.T.plot(kind='bar', stacked=True, alpha=0.5);
_images/DM_Pandas_3_22_0.png 
df2 = df/df.sum()
df2.T.plot(kind='bar', stacked=True, xlim= (0,6));
_images/DM_Pandas_3_23_0.png

Pie Plots#

plt.figure(figsize=(6,6))
plt.pie(s, labels=s.index)
plt.legend();
_images/DM_Pandas_3_25_0.png

Histograms and Kernel Density Estimate#

s = pd.Series(np.random.randn(1000))

plt.figure(figsize=(8,6))
s.hist(bins=30, density=True, alpha=0.3, label='histogram')
s.plot(kind='kde', label='kde', linewidth=2)
plt.grid(True)
plt.legend();
_images/DM_Pandas_3_27_0.png

Box and Scatter Plots#

http://de.wikipedia.org/wiki/Boxplot

df = pd.read_excel('daten/dshistory2013.xls',sheet_name='Price (EUR)',skiprows=1,index_col=0)
df.head()
hEXA01 hEXA02 hEXA03 hEXA04 hEXA05 hEXA06 hEXA07 hEXA08 hEXA09 hEXA10 ... bEXAoff1\n(01-08) bEXAoff2\n(21-24) bEXAdream (01-06) bEXAlunch (11-14) bEXAteatime (17-20) bEXAmoon\n(01-04) bEXAsun\n(05-08) bEXAearlyt\n(09-10) bEXAlatet\n(15-16) bEXAwakeup\n(07-08)
Delivery Date
2013-12-31 20.00 17.26 15.66 12.00 12.00 17.61 22.96 27.04 33.05 36.15 ... 18.07 27.64 15.76 32.06 37.40 16.23 19.90 34.60 32.00 25.00
2013-12-30 19.67 13.5 11.44 11.00 12.63 18.75 30.00 33.90 35.76 37.80 ... 18.86 34.53 14.50 36.28 43.43 13.90 23.82 36.78 34.33 31.95
2013-12-29 16.99 10.56 7.48 7.13 7.30 9.65 14.35 17.49 19.59 25.55 ... 11.37 36.30 9.85 31.53 40.95 10.54 12.20 22.57 29.43 15.92
2013-12-28 9.00 5.6 2.39 3.20 3.77 7.54 11.58 18.93 26.81 30.51 ... 7.75 32.26 5.25 34.51 44.78 5.05 10.46 28.66 35.26 15.26
2013-12-27 20.90 20.03 15.87 14.93 17.58 21.13 23.43 28.44 32.87 33.93 ... 20.29 17.66 18.41 32.62 33.50 17.93 22.65 33.40 29.92 25.94

5 rows × 38 columns

Find non-numeric values: cf. stackoverflow

The applymap method applies a function to each element in the data frame.

df.applymap(np.isreal).head()
hEXA01 hEXA02 hEXA03 hEXA04 hEXA05 hEXA06 hEXA07 hEXA08 hEXA09 hEXA10 ... bEXAoff1\n(01-08) bEXAoff2\n(21-24) bEXAdream (01-06) bEXAlunch (11-14) bEXAteatime (17-20) bEXAmoon\n(01-04) bEXAsun\n(05-08) bEXAearlyt\n(09-10) bEXAlatet\n(15-16) bEXAwakeup\n(07-08)
Delivery Date
2013-12-31 True True True True True True True True True True ... True True True True True True True True True True
2013-12-30 True True True True True True True True True True ... True True True True True True True True True True
2013-12-29 True True True True True True True True True True ... True True True True True True True True True True
2013-12-28 True True True True True True True True True True ... True True True True True True True True True True
2013-12-27 True True True True True True True True True True ... True True True True True True True True True True

5 rows × 38 columns

If all in the row are True then they are all numeric:

df.applymap(np.isreal).all(1).head()

Delivery Date
2013-12-31    True
2013-12-30    True
2013-12-29    True
2013-12-28    True
2013-12-27    True
dtype: bool

Use the negation to index the data frame for those rows which have at least one non-real value:

df[~df.applymap(np.isreal).all(1)]
hEXA01 hEXA02 hEXA03 hEXA04 hEXA05 hEXA06 hEXA07 hEXA08 hEXA09 hEXA10 ... bEXAoff1\n(01-08) bEXAoff2\n(21-24) bEXAdream (01-06) bEXAlunch (11-14) bEXAteatime (17-20) bEXAmoon\n(01-04) bEXAsun\n(05-08) bEXAearlyt\n(09-10) bEXAlatet\n(15-16) bEXAwakeup\n(07-08)
Delivery Date
2013-10-27 12.3 6,74 | 4,25 4.0 0.1 -0.2 -0.57 0.98 6.68 3.98 9.45 ... 3.81 13.68 3.8 4.48 21.25 5.48 1.72 6.72 -5.75 3.83

1 rows × 38 columns

df.loc['2013-10-27','hEXA02'] = 5.1
df.loc['2013-10-27','hEXA02'] 

5.1

Use only the first 24 columns of month data and map all values to floating point numbers.

df = df.iloc[:,:24].applymap(float)
df.columns

Index(['hEXA01', 'hEXA02', 'hEXA03', 'hEXA04', 'hEXA05', 'hEXA06', 'hEXA07',
       'hEXA08', 'hEXA09', 'hEXA10', 'hEXA11', 'hEXA12', 'hEXA13', 'hEXA14',
       'hEXA15', 'hEXA16', 'hEXA17', 'hEXA18', 'hEXA19', 'hEXA20', 'hEXA21',
       'hEXA22', 'hEXA23', 'hEXA24'],
      dtype='object')

df.boxplot(figsize=(18,9))
plt.ylabel('price [EUR]');
_images/DM_Pandas_3_40_0.png 
df.loc[:'2013-06-01'].plot(kind='scatter', x='hEXA10', y='hEXA13', figsize=(8,8));
plt.grid()
_images/DM_Pandas_3_41_0.png 
pd.plotting.scatter_matrix(df.iloc[:,:6]);
_images/DM_Pandas_3_42_0.png 
df['hEXA12'].plot();
plt.grid()
_images/DM_Pandas_3_43_0.png 
plt.figure(figsize=(10,8))
plt.plot(df['hEXA12'].sort_values(ascending=False).values)

plt.xlabel('days')
plt.ylabel('price')
plt.grid()
_images/DM_Pandas_3_44_0.png