Plotting a Pandas DataSeries.GroupBy

Question

I am new to python and pandas, and have the following DataFrame.

How can I plot the DataFrame where each ModelID is a separate plot, saledate is the x-axis and MeanToDate is the y-axis?

Attempt

data[40:76].groupby('ModelID').plot()

DataFrame

Solution 1

You can make the plots by looping over the groups from groupby:

import matplotlib.pyplot as plt

for title, group in df.groupby('ModelID'):
    group.plot(x='saleDate', y='MeanToDate', title=title)

See for more information on plotting with pandas dataframes:
http://pandas.pydata.org/pandas-docs/stable/visualization.html
and for looping over a groupby-object:
http://pandas.pydata.org/pandas-docs/stable/groupby.html#iterating-through-groups

Solution 2

Example with aggregation:

I wanted to do something like the following, if pandas had a colour aesthetic like ggplot:

aggregated = df.groupby(['model', 'training_examples']).aggregate(np.mean)
aggregated.plot(x='training_examples', y='accuracy', label='model')

(columns: model is a string, training_examples is an integer, accuracy is a decimal)

But that just produces a mess.

Thanks to joris's answer, I ended up with:

for index, group in df.groupby(['model']):
    group_agg = group.groupby(['training_examples']).aggregate(np.mean)
    group_agg.plot(y='accuracy', label=index)

I found that title= was just replacing the single title of the plot on each loop iteration, but label= does what you'd expect -- after running plt.legend(), of course.

Comments

[gented]

The plots will be displayed on independent figures even without the plt.figure() first.

[glS]

another nice tutorial on using groupby is here: chrisalbon.com/python/pandas_apply_operations_to_groups.html