Visualizing Time Series Data¶
Let's go through a few key points of creatng nice time visualizations!
Import¶
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
# Optional for interactive
# %matplotlib notebook (watch video for full details)
Get Data (CSV)¶
mcdon = pd.read_csv('mcdonalds.csv',index_col='Date',parse_dates=True)
mcdon.head()
Not a good practice to plot all columns in one chart! The Y-axis "Adj Close" and "Adj Volume" are on totally different scale
mcdon.plot()
Instead you should plot them individually
mcdon['Adj. Close'].plot()
mcdon['Adj. Volume'].plot()
mcdon['Adj. Close'].plot(figsize=(12,8))
mcdon['Adj. Close'].plot(figsize=(12,8))
plt.ylabel('Close Price')
plt.xlabel('Overwrite Date Index')
plt.title('Mcdonalds')
mcdon['Adj. Close'].plot(figsize=(12,8),title='Pandas Title')
Range to Diplay (xlim & ylim)¶
Use xlim to specify a period for a closer look. Note the ylim won't automatically adjust
mcdon['Adj. Close'].plot(xlim=['2007-01-01','2009-01-01'])
You need to specify ylim as well
mcdon['Adj. Close'].plot(xlim=['2007-01-01','2009-01-01'],ylim=[0,50])
Color and Style¶
Use ls and c to change the line style and color respectively
mcdon['Adj. Close'].plot(xlim=['2007-01-01','2007-05-01'],ylim=[0,40],ls='--',c='r')
Customize X Ticks (date), Grid, Minor Axis¶
This is where you will need the power of matplotlib to do heavy lifting if you want some serious customization!
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.dates as dates
mcdon['Adj. Close'].plot(xlim=['2007-01-01','2007-05-01'],ylim=[0,40])
To Grab the whole index
idx = mcdon.index
To Grab a portion of the index
Use loc[] to grab a portion of index. Note we use : for slicing
idx = mcdon.loc['2007-01-01':'2007-05-01'].index
Also grab a portion of stock data
stock = mcdon.loc['2007-01-01':'2007-05-01']['Adj. Close']
idx
stock
Basic matplotlib plot¶
Use plot_date() instead of plot(). plot_date() can handle date time index information
fig, ax = plt.subplots()
ax.plot_date(idx, stock,'-')
plt.tight_layout()
plt.show()
Note the overlapping label still shown after plt.tight_layout() is called
Fix the Overlap¶
fig, ax = plt.subplots()
ax.plot_date(idx, stock,'-')
fig.autofmt_xdate() # Auto fixes the overlap!
plt.tight_layout()
plt.show()
Customize Grid¶
Use ax.xaxis.grid(True) and yaxis.grid(True) to add grids to x-axis and y-axis respectively
fig, ax = plt.subplots()
ax.plot_date(idx, stock,'-')
ax.yaxis.grid(True)
ax.xaxis.grid(True)
fig.autofmt_xdate() # Auto fixes the overlap!
plt.tight_layout()
plt.show()
Format Dates on Major Axis¶
This is a two step process. The first step is SELECT display location followed by FORMATTING
Use xaxis.set_major_locator() to select, then use xaxis.set_major_formatter() to format
Example of format string: %b-%Y "Apr-2007", %B-%Y "April-2007", %b\n%Y
fig, ax = plt.subplots()
ax.plot_date(idx, stock,'-')
# Grids
ax.yaxis.grid(True)
ax.xaxis.grid(True)
# Major Axis
ax.xaxis.set_major_locator(dates.MonthLocator())
ax.xaxis.set_major_formatter(dates.DateFormatter('%b\n%Y'))
fig.autofmt_xdate() # Auto fixes the overlap!
plt.tight_layout()
plt.show()
Adding Minor Axis¶
In the following example, we pass in the format string \n\n\n\n%Y--%B' to dates.DateFormatter() in order to add some space for inserting a minor axis
Same as major axis, we locate and format the minor axis by calling xaxis.set_minor_locator() followed by xaxis.set_minor_formatter()
fig, ax = plt.subplots()
ax.plot_date(idx, stock,'-')
ax.xaxis.set_major_locator(dates.MonthLocator())
ax.xaxis.set_major_formatter(dates.DateFormatter('\n\n%Y--%B'))
Minor Axis¶
ax.xaxis.set_minor_locator(dates.WeekdayLocator())
ax.xaxis.set_minor_formatter(dates.DateFormatter('%d'))
Set Grids¶
ax.yaxis.grid(True)
ax.xaxis.grid(True)
fig.autofmt_xdate() # Auto fixes the overlap!
plt.tight_layout()
plt.show()
The parameter byweekday=1 pass to dates.WeekdayLocator() set the actual day of the week to display. '0' is Monday, '1' is Tuesday etc.
The format string %B-%d-%a pass to dates.DateFormatter(): %a: 'Tue', %d: day of month, %B: full month name
fig, ax = plt.subplots(figsize=(10,8))
ax.plot_date(idx, stock,'-')
ax.xaxis.set_major_locator(dates.WeekdayLocator(byweekday=1))
ax.xaxis.set_major_formatter(dates.DateFormatter('%B-%d-%a'))
Set Grids¶
ax.yaxis.grid(True)
ax.xaxis.grid(True)
Auto fixes the overlap!¶
fig.autofmt_xdate()
plt.tight_layout()
plt.show()
Trick: Make Jupyter Notebook Interactive¶
- Restart the kernel. From pull down menu:
Kernel > Restart - Change
%matplotlib inlineto%matplotlib notebookimport pandas as pd import matplotlib.pyplot as plt %matplotlib notebook
- Run the cells again
