Plotting multiple line graph using pandas and matplotlib

Question

I have the following data in a pandas dataframe

       date  template     score
0  20140605         0  0.138786
1  20140605         1  0.846441
2  20140605         2  0.766636
3  20140605         3  0.259632
4  20140605         4  0.497366
5  20140606         0  0.138139
6  20140606         1  0.845320
7  20140606         2  0.762876
8  20140606         3  0.261035
9  20140606         4  0.498010

For every day there will be 5 templates and each template will have a score.

I want to plot the date in the x axis and score in the y axis and a separate line graph for each template in the same figure.

Is it possible to do this using matplotlib?

just fast-fast : try to start from the samples : matplotlib.org/examples/pylab_examples/date_demo_rrule.html — Louis
– Louis, Commented Jun 6, 2014 at 12:21
@Louis Just to be clear. I want to know how to plot a grouped dataframe, not about processing dates — Sudar
– Sudar, Commented Jun 6, 2014 at 13:06

Thomas Cokelaer · Accepted Answer · 2014-06-07 15:47:06Z

56

You can use the groupby method:

data.groupby("template").plot(x="date", y="score")

answered Jun 7, 2014 at 15:47

Thomas Cokelaer

1,0591 gold badge8 silver badges12 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

Sudar Over a year ago

wow!! this is very crisp and gets the work done. It never occurred to me that I can group by template as well. If I had not accepted the previous answer, I would have accepted this one though.

Conrad.Dean Over a year ago

is there a way to have the groups populate the legend?

dshepherd Over a year ago

To get them on a single figure I had to do fig, ax = plt.subplots(1,1); data.groupby("template").plot(x="date", y="score", ax=ax). It seems like there should be a nicer way to do this and get the legends right automatically as well.

dshepherd Over a year ago

I populated the legend using plt.legend([v[0] for v in pr.groupby('template')['template']]), which is messy but it works (groupby returns an iterator of data frames, within each data frame all values of 'template' are the same so we just take the first).

curtisp Over a year ago

This is great, it is the start of "small multiples" type charting with each groupby thing (in this case "template") getting its own chart. Just need to resize them so they are smaller. Nice.

|

Jon Hoffman · Accepted Answer · 2015-12-05 20:34:54Z

18

I think the easiest way to plot this data with all the lines on the same graph is to pivot it such that each "template" value is a column:

pivoted = pandas.pivot_table(data, values='score', columns='template', index='date')
# Now there will be an index column for date and value columns for 0,1,2,3,4
pivoted.plot()

answered Dec 5, 2015 at 20:34

Jon Hoffman

8626 silver badges11 bronze badges

2 Comments

curtisp Over a year ago

This is great also. Searched all over and this is best explanation of how pivot can be used with plot. I had two things I wanted to include on x axis. In my case I wanted month and day which were separate columns in my dataframe. So I used "index=['month','day']" which put both on x axis. Thanks!

etov Over a year ago

I prefer this one over the accepted solution because: 1. no need to define ax separately, 2. unified way to play with different axes and groupings

papafe · Accepted Answer · 2014-06-06 13:14:59Z

You can use an approach like the following one. You can simply slice the dataframe according to the values of each template, and subsequently use the dates and scores for the plot.

from pandas import *
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import datetime as dt

#The following part is just for generating something similar to your dataframe
date1 = "20140605"
date2 = "20140606"

d = {'date': Series([date1]*5 + [date2]*5), 'template': Series(range(5)*2),
'score': Series([random() for i in range(10)]) } 

data = DataFrame(d)
#end of dataset generation

fig, ax = plt.subplots()

for temp in range(5):
    dat = data[data['template']==temp]
    dates =  dat['date']
    dates_f = [dt.datetime.strptime(date,'%Y%m%d') for date in dates]
    ax.plot(dates_f, dat['score'], label = "Template: {0}".format(temp))

plt.xlabel("Date")
plt.ylabel("Score")
ax.legend()
plt.show()

animalito · Accepted Answer · 2014-10-10 18:45:42Z

1

You can add the legend according to the groups with:

plt.legend(pr['template'], loc='best')

answered Oct 10, 2014 at 18:45

animalito

3822 silver badges7 bronze badges

1 Comment

dshepherd Over a year ago

Actually you can't do this in general. pr['template'] = [0,1,2,3,4,0,1,2,3,4] in the case in the question, so it appears to work. However in general the first 4 elements of pr['template'] will not contain each of the labels in the correct order. For example if you sort the data before plotting then pr['template'][:4] = [0,0,1,1].

Collectives™ on Stack Overflow

Plotting multiple line graph using pandas and matplotlib

4 Answers 4

7 Comments

2 Comments

Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

7 Comments

2 Comments

Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related