How to label data points in matplotlib scatter plot while looping through pandas dataframes?

Question

I have a pandas dataframe including the following columns:

label = ('A' , 'D' , 'K', 'L', 'P')
x = (1 , 4 , 9, 6, 4)
y = (2 , 6 , 5, 8, 9)
plot_id = (1 , 1 , 2, 2, 3)

I want to creat 3 seperate scatter plots - one for each individual plot_id. So the first scatter plot should consists all entries where plot_id == 1 and hence the points (1,2) and (4,6). Each data point should be labelled by label. Hence the first plot should have the labels Aand B.

I understand I can use annotate to label, and I am familiar with for loops. But I have no idea how to combine the two.

I wish I could post better code snippet of what I have done so far - but it's just terrible. Here it is:

for i in range(len(df.plot_id)):
    plt.scatter(df.x[i],df.y[i])
    plt.show()

That's all I got - unfortunately. Any ideas on how to procede?

Sorry, I edited the question while commented. I basically am trying to make 3 plots - for each individual `plot_id`. — Rachel, Dec 05 '16 at 16:44
No. I want to label/annotate the data entries (or glyphs if you will) with `label`. — Rachel, Dec 05 '16 at 16:48
You need to be very precise about the following: How many plots do you want to create? How many points do you want each plot to have? Where should the labels appear in the plot? Is it correct that you want to have exactly one point per plot? — ImportanceOfBeingErnest, Dec 05 '16 at 16:48
Sorry. I edited the question above. Do I make myself understandable? — Rachel, Dec 05 '16 at 16:55
how many different `plot_ids` do you have ? If it's a small number you can do some subplots. — MMF, Dec 05 '16 at 16:56

piRSquared · Accepted Answer · 2016-12-06T07:04:20.603

4

updated answer
save separate image files

def annotate(row, ax):
    ax.annotate(row.label, (row.x, row.y),
                xytext=(10, -5), textcoords='offset points')

for pid, grp in df.groupby('plot_id'):
    ax = grp.plot.scatter('x', 'y')
    grp.apply(annotate, ax=ax, axis=1)
    plt.savefig('{}.png'.format(pid))
    plt.close()

1.png

2.png

3.png

old answer
for those who want something like this

def annotate(row, ax):
    ax.annotate(row.label, (row.x, row.y),
                xytext=(10, -5), textcoords='offset points')

fig, axes = plt.subplots(df.plot_id.nunique(), 1)
for i, (pid, grp) in enumerate(df.groupby('plot_id')):
    ax = axes[i]
    grp.plot.scatter('x', 'y', ax=ax)
    grp.apply(annotate, ax=ax, axis=1)
fig.tight_layout()

setup

label = ('A' , 'D' , 'K', 'L', 'P')
x = (1 , 4 , 9, 6, 4)
y = (2 , 6 , 5, 8, 9)
plot_id = (1 , 1 , 2, 2, 3)

df = pd.DataFrame(dict(label=label, x=x, y=y, plot_id=plot_id))

edited Dec 06 '16 at 07:04

answered Dec 05 '16 at 17:13

piRSquared

285,575
57
475
624

1

since there are 54 `plot_ids`, I don't think subplots might be a good idea. am I wrong ? – MMF Dec 05 '16 at 17:15
1

I'm sorry, you weren't clear. You said you wanted separate plots. – piRSquared Dec 05 '16 at 18:31
Yes, indeed. I need 54 individual plots. I will try to be more clear next time! – Rachel Dec 05 '16 at 18:39
Great solution! Thank you! – Rachel Dec 06 '16 at 07:19

MMF · Answer 2 · 2016-12-05T17:19:27.053

1

Here is a simple way to deal with your problem :

zipped = zip(zip(zip(df.x, df.y), df.plot_id), df.label)
# Result : [(((1, 2), 1), 'A'),
#           (((4, 6), 1), 'D'),
#           (((9, 5), 2), 'K'),
#           (((6, 8), 2), 'L'),
#           (((4, 9), 3), 'P')]

To retrieve the positions, the plot index and the labels, you can loop as below :

for (pos, plot), label in zipped:
    ...
    print pos
    print plot
    print label

Now here is what you can do in your case :

import matplotlib.pyplot as plt

for (pos, plot), label in zipped:
    plt.figure(plot)
    x, y = pos
    plt.scatter(x, y)
    plt.annotate(label, xy=pos)

It will create as much figures as plot_ids and for each figure display the scatter plot of the points with the corresponding plot_ids value. What's more it overlays the label on each point.

edited Dec 05 '16 at 17:19

answered Dec 05 '16 at 17:14

MMF

5,750
3
16
20

Wow! This is great! Is there a way to save the plots on the loop too? I tried to adapt the code and save but unfortunately replace too... – Rachel Dec 05 '16 at 18:40
I get a figure for each `pos` . So given the example brought forward, I get 6 figures. How do I combine them into 3? – Rachel Dec 05 '16 at 20:19
@Rachel Are you sure that you get a figure for each `pos` ? It works perfectly for me ... – MMF Dec 05 '16 at 21:09
Yes. Your print command suggests you use Python 2 whilst I use python 3? Maybe that's why? – Rachel Dec 05 '16 at 21:14
can you edit your question with your new piece of code and the variables you use ? I'll check it out – MMF Dec 05 '16 at 21:20
I copied your code exactly (I always do, before applying it to my own code). I do indeed get a figure for each `pos`. Strange! Any ideas? – Rachel Dec 06 '16 at 06:45

score 0 · Answer 3 · answered Sep 07 '22 at 16:58

This is a function to create these plots (based on @piRSquared answer)

def plotter2(data,x,y,grp,lbl):

    def annotate(row, ax):
       ax.annotate(row[lbl], (row[x], row[y]),
            xytext=(3, 0), textcoords='offset points')

   for pid, grp in data.groupby(grp):
       ax = grp.plot.scatter(x, y)
       grp.apply(annotate, ax=ax, axis=1)
       plt.show()
       plt.savefig('{}.png'.format(pid))

How to label data points in matplotlib scatter plot while looping through pandas dataframes?

3 Answers3

Linked