With a dataframe and basic plot such as this:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
np.random.seed(123456)
rows = 75
df = pd.DataFrame(np.random.randint(-4,5,size=(rows, 3)), columns=['A', 'B', 'C'])
datelist = pd.date_range(pd.datetime(2017, 1, 1).strftime('%Y-%m-%d'), periods=rows).tolist()
df['dates'] = datelist
df = df.set_index(['dates'])
df.index = pd.to_datetime(df.index)
df = df.cumsum()
df.plot()
What is the best way of annotating the last points on the lines so that you get the result below?
In order to annotate a point use ax.annotate(). In this case it makes sense to specify the coordinates to annotate separately. I.e. the y coordinate is the data coordinate of the last point of the line (which you can get from line.get_ydata()[-1]) while the x coordinate is independent of the data and should be the right hand side of the axes (i.e. 1 in axes coordinates). You may then also want to offset the text a bit such that it does not overlap with the axes.
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
rows = 75
df = pd.DataFrame(np.random.randint(-4,5,size=(rows, 3)), columns=['A', 'B', 'C'])
datelist = pd.date_range(pd.datetime(2017, 1, 1).strftime('%Y-%m-%d'), periods=rows).tolist()
df['dates'] = datelist
df = df.set_index(['dates'])
df.index = pd.to_datetime(df.index)
df = df.cumsum()
ax = df.plot()
for line, name in zip(ax.lines, df.columns):
y = line.get_ydata()[-1]
ax.annotate(name, xy=(1,y), xytext=(6,0), color=line.get_color(),
xycoords = ax.get_yaxis_transform(), textcoords="offset points",
size=14, va="center")
plt.show()
Method 1
Here is one way, or at least a method, which you can adapt to aesthetically fit in whatever way you want, using the plt.annotate method:
[EDIT]: If you're going to use a method like this first one, the method outlined in ImportanceOfBeingErnest's answer is better than what I've proposed.
df.plot()
for col in df.columns:
plt.annotate(col,xy=(plt.xticks()[0][-1]+0.7, df[col].iloc[-1]))
plt.show()
For the xy argument, which is the x and y coordinates of the text, I chose the last x coordinate in plt.xticks(), and added 0.7 so that it is outside of your x axis, but you can coose to make it closer or further as you see fit.
METHOD 2:
You could also just use the right y axis, and label it with your 3 lines. For example:
fig, ax = plt.subplots()
df.plot(ax=ax)
ax2 = ax.twinx()
ax2.set_ylim(ax.get_ylim())
ax2.set_yticks([df[col].iloc[-1] for col in df.columns])
ax2.set_yticklabels(df.columns)
plt.show()
This gives you the following plot:
I've got some tips from the other answers and believe this is the easiest solution.
Here is a generic function to improve the labels of a line chart. Its advantages are:
you don't need to mess with the original DataFrame since it works over a line chart,
it will use the already set legend label,
removes the frame,
just copy'n paste it to improve your chart :-)
You can just call it after creating any line char:
def improve_legend(ax=None):
if ax is None:
ax = plt.gca()
for spine in ax.spines:
ax.spines[spine].set_visible(False)
for line in ax.lines:
data_x, data_y = line.get_data()
right_most_x = data_x[-1]
right_most_y = data_y[-1]
ax.annotate(
line.get_label(),
xy=(right_most_x, right_most_y),
xytext=(5, 0),
textcoords="offset points",
va="center",
color=line.get_color(),
)
ax.legend().set_visible(False)
This is the original chart:
Now you just need to call the function to improve your plot:
ax = df.plot()
improve_legend(ax)
The new chart:
Beware, it will probably not work well if a line has null values at the end.
Related
Question
I have used the secondary_y argument in pd.DataFrame.plot().
While trying to change the fontsize of legends by .legend(fontsize=20), I ended up having only 1 column name in the legend when I actually have 2 columns to be printed on the legend.
This problem (having only 1 column name in the legend) does not take place when I did not use secondary_y argument.
I want all the column names in my dataframe to be printed in the legend, and change the fontsize of the legend even when I use secondary_y while plotting dataframe.
Example
The following example with secondary_y shows only 1 column name A, when I have actually 2 columns, which are A and B.
The fontsize of the legend is changed, but only for 1 column name.
import pandas as pd
import numpy as np
np.random.seed(42)
df = pd.DataFrame(np.random.randn(24*3, 2),
index=pd.date_range('1/1/2019', periods=24*3, freq='h'))
df.columns = ['A', 'B']
df.plot(secondary_y = ["B"], figsize=(12,5)).legend(fontsize=20, loc="upper right")
When I do not use secondary_y, then legend shows both of the 2 columns A and B.
import pandas as pd
import numpy as np
np.random.seed(42)
df = pd.DataFrame(np.random.randn(24*3, 2),
index=pd.date_range('1/1/2019', periods=24*3, freq='h'))
df.columns = ['A', 'B']
df.plot(figsize=(12,5)).legend(fontsize=20, loc="upper right")
To manage to customize it you have to create your graph with subplots function of Matplotlib:
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
np.random.seed(42)
df = pd.DataFrame(np.random.randn(24*3, 2),
index=pd.date_range('1/1/2019', periods=24*3, freq='h'))
df.columns = ['A', 'B']
#define colors to use
col1 = 'steelblue'
col2 = 'red'
#define subplots
fig,ax = plt.subplots()
#add first line to plot
lns1=ax.plot(df.index,df['A'], color=col1)
#add x-axis label
ax.set_xlabel('dates', fontsize=14)
#add y-axis label
ax.set_ylabel('A', color=col1, fontsize=16)
#define second y-axis that shares x-axis with current plot
ax2 = ax.twinx()
#add second line to plot
lns2=ax2.plot(df.index,df['B'], color=col2)
#add second y-axis label
ax2.set_ylabel('B', color=col2, fontsize=16)
#legend
ax.legend(lns1+lns2,['A','B'],loc="upper right",fontsize=20)
#another solution is to create legend for fig,:
#fig.legend(['A','B'],loc="upper right")
plt.show()
result:
this is a somewhat late response, but something that worked for me was simply setting plt.legend(fontsize = wanted_fontsize) after the plot function.
I can't get the legends to show on the subplots which show up just fine and take the other formatting I've applied. What am I missing?
If I do a plot for the dataframe alone, it shows the legend. If I add a label to the plot for the subplots, it assigns that label to all three lines.
Here is image. plot vs subplot
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from functools import reduce
%matplotlib notebook
#Source for files
# Per Capita Personal Income
# Ann Arbor https://fred.stlouisfed.org/series/ANNA426PCPI
# MI https://fred.stlouisfed.org/series/MIPCPI
# USA https://fred.stlouisfed.org/series/A792RC0A052NBEA
dfAnnArbor_PCPI = pd.read_csv('PerCapitaPersonalIncomeAnnArborMI.csv', skiprows=1, names=['Date', 'PCPI'])
dfMI_PCPI = pd.read_csv('PerCapitaPersonalIncomeMI.csv', skiprows=1, names=['Date', 'PCPI'])
dfUSA_PCPI = pd.read_csv('PerCapitaPersonalIncomeUSA.csv', skiprows=1, names=['Date', 'PCPI'])
# consolidate three df into one using Date
dfAll = [dfAnnArbor_PCPI, dfMI_PCPI, dfUSA_PCPI]
dfPCPI = reduce(lambda left, right: pd.merge(left, right, on='Date', how='outer'), dfAll)
dfPCPI = dfPCPI.dropna() # drop rows with NaN
dfPCPI.columns = ['Date', 'AnnArbor', 'MI', 'USA'] # rename columns
dfPCPI['Date'] = dfPCPI['Date'].str[:4] # select only year
dfPCPI = dfPCPI.set_index('Date')
dfPCPI_Rel = dfPCPI.apply(lambda x: x / x[0])
dfPCPI_Small = dfPCPI.iloc[8:].copy()
dfPCPI_SmRel = dfPCPI_Small.apply(lambda x: x / x[0])
dfPCPI_SmRel.plot()
fig, ax = plt.subplots(1, 2)
ax0 = ax[0].plot(dfPCPI_Rel, '-', label='a')
ax1 = ax[1].plot(dfPCPI_SmRel, '-', label='test1')
ax[0].legend()
for x in fig.axes:
for label in x.get_xticklabels():
label.set_rotation(45)
ax[1].xaxis.set_major_locator(ticker.MultipleLocator(2))
plt.show()
The legend in pyplot refers to an axis instance. Therefore, if you want multiple plots to have their own legend, you need to call legend() for each axis. In your case
ax[0].legend()
ax[1].legend()
Additionally, as you are calling plot(), you may want to use the keyword label in each plot() call so as to have a label for each legend entry.
You should try fig.legend() instead of plt.legend()
I was trying to do a comparison of runtime between Naive matrix multiplication and Strassen's. For this, I was recording the runtime for a different dimension of the matrices. Then I was trying to plot the result in the same graph for the comparison.
But the problem is the plotting is not showing the proper result.
Here is the data...
2 3142
3 3531
4 4756
5 5781
6 8107
The leftmost column is denoting n, the dimension and rightmost column is denoting execution time.
The above data is for Naive method and the data for Strassen is in this pattern too.
I'm inserting this data to a pandas dataframe. And after plotting the data the image looks like this:
Here blue is for Naive and green is for Strassen's
This is certainly not true as Naive cannot be constant. But my code was correct. SO I decided to plot them separately and these are the result:
Naive
Strassen
As you can see it might happen because the scaling in Y axis is not the same?
Is this the reason?
The code I'm implementing for plotting is:
fig = plt.figure()
data_naive = pd.read_csv('naive.txt', sep="\t", header=None)
data_naive.columns = ["n", "time"]
plt.plot(data_naive['n'], data_naive['time'], 'g')
data_strassen = pd.read_csv('strassen.txt', sep="\t", header=None)
data_strassen.columns = ["n", "time"]
plt.plot(data_strassen['n'], data_strassen['time'], 'b')
plt.show()
fig.savefig('figure.png')
What I tried to work out?
fig = plt.figure()
data_naive = pd.read_csv('naive.txt', sep="\t", header=None)
data_naive.columns = ["n", "time"]
data_strassen = pd.read_csv('strassen.txt', sep="\t", header=None)
data_strassen.columns = ["n", "time"]
ax = data_naive.plot(x='n', y='time', c='blue', figsize=(20,10))
data_strassen.plot(x='n', y='time', c='green', figsize=(20,10), ax=ax)
plt.savefig('comparison.png')
plt.show()
But no luck!!!
How to plot them in the same figure without altering their actual orientation?
IIUC: Here is a solution using twinx
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
df = pd.DataFrame(np.random.randint(10, 100, (12,2)))
df[1] = np.random.dirichlet(np.ones(12)*1000., size=1)[0]
fig, ax1 = plt.subplots()
ax1.plot(df[0], color='r')
#Plot the secondary axis in the right side
ax2 = ax1.twinx()
ax2.plot(df[1], color='k')
fig.tight_layout()
plt.show()
Result produced:
I am trying to plot two different variables (linked by a relation of causality), delai_jour and date_sondage on a single FacetGrid. I can do it with this code:
g = sns.FacetGrid(df_verif_sum, col="prefecture", col_wrap=2, aspect=2, sharex=True,)
g = g.map(plt.plot, "date_sondage", "delai_jour", color="m", linewidth=2)
g = g.map(plt.bar, "date_sondage", "impossible")
which gives me this:
FacetGrid
(There are 33 of them in total).
I'm interested in comparing the patterns across the various prefecture, but due to the difference in magnitude I cannot see the changes in the line chart.
For this specific work, the best way to do it is to create a secondary y axis, but I can't seem to make anything work: it doesn't look like it's possible with FacetGrid, and I didn't understand the code not was able to replicate the examples i've seen with pure matplotlib.
How should I go about it?
I got this to work by iterating through the axes and plotting a secondary axis as in a typical Seaborn graph.
Using the OP example:
g = sns.FacetGrid(df_verif_sum, col="prefecture", col_wrap=2, aspect=2, sharex=True)
g = g.map(plt.plot, "date_sondage", "delai_jour", color="m", linewidth=2)
for ax, (_, subdata) in zip(g.axes, df_verif_sum.groupby('prefecture')):
ax2=ax.twinx()
subdata.plot(x='data_sondage',y='impossible', ax=ax2,legend=False,color='r')
If you do any formatting to the x-axis, you may have to do it to both ax and ax2.
Here's an example where you apply a custom mapping function to the dataframe of interest. Within the function, you can call plt.gca() to get the current axis at the facet being currently plotted in FacetGrid. Once you have the axis, twinx() can be called just like you would in plain old matplotlib plotting.
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import seaborn as sns
def facetgrid_two_axes(*args, **kwargs):
data = kwargs.pop('data')
dual_axis = kwargs.pop('dual_axis')
alpha = kwargs.pop('alpha', 0.2)
kwargs.pop('color')
ax = plt.gca()
if dual_axis:
ax2 = ax.twinx()
ax2.set_ylabel('Second Axis!')
ax.plot(data['x'],data['y1'], **kwargs, color='red',alpha=alpha)
if dual_axis:
ax2.plot(df['x'],df['y2'], **kwargs, color='blue',alpha=alpha)
df = pd.DataFrame()
df['x'] = np.arange(1,5,1)
df['y1'] = 1 / df['x']
df['y2'] = df['x'] * 100
df['facet'] = 'foo'
df2 = df.copy()
df2['facet'] = 'bar'
df3 = pd.concat([df,df2])
win_plot = sns.FacetGrid(df3, col='facet', size=6)
(win_plot.map_dataframe(facetgrid_two_axes, dual_axis=True)
.set_axis_labels("X", "First Y-axis"))
plt.show()
This isn't the prettiest plot as you might want to adjust the presence of the second y-axis' label, the spacing between plots, etc. but the code suffices to show how to plot two series of differing magnitudes within FacetGrids.
I have many data frames that I am plotting for a presentation. These all have different columns, but all contain the same additional column foobar. At the moment, I am plotting these different data frames using
df.plot(secondary_y='foobar')
Unfortunately, since these data frames all have different additional columns with different ordering, the color of foobar is always different. This makes the presentation slides unnecessary complicated. I would like, throughout the different plots, assign that foobar is plotted bold and black.
Looking at the docs, the only thing coming close appears to be the parameter colormap - I would need to ensure that the xth color in the color map is always black, where x is the order of foobar in the data frame. Seems to be more complicated than it should be, also this wouldn't make it bold.
Is there a (better) approach?
I would suggest using matplotlib directly rather than the dataframe plotting methods. If df.plot returned the artists it added instead of an Axes object it wouldn't be too bad to change the color of the line after it was plotted.
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
def pandas_plot(ax, df, callout_key):
"""
Parameters
----------
ax : mpl.Axes
The axes to draw to
df : DataFrame
Data to plot
callout_key : str
key to highlight
"""
artists = {}
x = df.index.values
for k, v in df.iteritems():
style_kwargs = {}
if k == callout_key:
style_kwargs['c'] = 'k'
style_kwargs['lw'] = 2
ln, = ax.plot(x, v.values, **style_kwargs)
artists[k] = ln
ax.legend()
ax.set_xlim(np.min(x), np.max(x))
return artists
Usage:
fig, ax = plt.subplots()
ax2 = ax.twinx()
th = np.linspace(0, 2*np.pi, 1024)
df = pd.DataFrame({'cos': np.cos(th), 'sin': np.sin(th),
'foo': np.sin(th + 1), 'bar': np.cos(th +1)}, index=th)
df2 = pd.DataFrame({'cos': -np.cos(th), 'sin': -np.sin(th)}, index=th)
pandas_plot(ax, df, 'sin')
pandas_plot(ax2, df2, 'sin')
Perhaps you could define a function which handles the special column in a separate plot call:
def emphasize_plot(ax, df, col, **emphargs):
columns = [c for c in df.columns if c != col]
df[columns].plot(ax=ax)
df[col].plot(ax=ax, **emphargs)
Using code from tcaswell's example,
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
def emphasize_plot(ax, df, col, **emphargs):
columns = [c for c in df.columns if c != col]
df[columns].plot(ax=ax)
df[col].plot(ax=ax, **emphargs)
fig, ax = plt.subplots()
th = np.linspace(0, 2*np.pi, 1024)
df = pd.DataFrame({'cos': np.cos(th), 'foobar': np.sin(th),
'foo': np.sin(th + 1), 'bar': np.cos(th +1)}, index=th)
df2 = pd.DataFrame({'cos': -np.cos(th), 'foobar': -np.sin(th)}, index=th)
emphasize_plot(ax, df, 'foobar', lw=2, c='k')
emphasize_plot(ax, df2, 'foobar', lw=2, c='k')
plt.show()
yields
I used #unutbut's answer and extended it to allow for a secondary y axis and correct legends:
def emphasize_plot(ax, df, col, **emphargs):
columns = [c for c in df.columns if c != col]
ax2 = ax.twinx()
df[columns].plot(ax=ax)
df[col].plot(ax=ax2, **emphargs)
lines, labels = ax.get_legend_handles_labels()
lines2, labels2 = ax2.get_legend_handles_labels()
ax2.legend(lines + lines2, labels + labels2, loc=0)