5

This question has been asked on SO, but I want to find a clearer solution.

Given X is 100x2 data, and labels is the label vector ( from 1 to 9) I plot the scatter plot as following:

pl.scatter(X[:,0], X[:,1], c = labels)
pl.show()

enter image description here

How to add legend to explain the colors in just one line of code? Other solutions plot each label separately:

a = pl.scatter(X1[:,0], X1[:,1], color = "red")
b = pl.scatter(X2[:,0], X2[:,1], color = "green")
c = pl.scatter(X3[:,0], X3[:,1], color = "blue")
pl.legend((a,b,c), ("line 1", "line 2", "line 3")
pl.show()
kdopen
  • 8,032
  • 7
  • 44
  • 52
Dzung Nguyen
  • 3,794
  • 9
  • 48
  • 86
  • 1
    I know you've probably seen this but in case you haven't: http://stackoverflow.com/questions/17411940/matplotlib-scatter-plot-legend – s16h Apr 15 '14 at 21:50
  • Just label each plot and invoke legend(): plt.scatter(x1,y1,label=str(pointset1)); ...; plt.legend(loc='upper right', numpoints=1, ncol=3, fontsize=8, bbox_to_anchor=(1,1)); plt.show() – MasterControlProgram Oct 20 '16 at 13:28

2 Answers2

3

You could create a dummy scatter plot with the desired legend as follows:

pl.scatter(X[:,0], X[:,1], c = labels)

for item in labels:
    #dummy plot just to create the legend
    pl.scatter([], [], c = item, label = item)
#loc = 0 is for the best position of the legend
#scatterpoints = 1 will only show one point in the legend instead of multiple points
plt.legend(loc = 0, scatterpoints = 1)
pl.show()
sherminator35
  • 189
  • 1
  • 10
2

Just label each plot and invoke legend() as you do :)

plt.scatter(x1,y1,label=str(pointset1))
plt.scatter(x2,y2,label=str(pointset2))
plt.scatter(x3,y3,label=str(pointset3))

plt.legend(loc='upper right', numpoints=1, ncol=3, fontsize=8, bbox_to_anchor=(1,1))
plt.show()
MasterControlProgram
  • 349
  • 2
  • 7