Redundant Legends: Matplotlib

Question

My scatter plot has redundant legends. Here is the image of my plot.

On this issue, I already checked the following existing question at StackOverflow: too many legend with array column data in matplotlib

Nevertheless, it did not help. I presume that I am encountering a totally different issue. Please advise me how to resolve this.

Here is my code:

import matplotlib.cm as cm
colors = cm.rainbow(np.linspace(0, 1, N_Clus))
cluster_labels_2 = list(range(1, N_Clus+1))
print("cluster_labels: ", cluster_labels_2)
# Create a figure
plt.figure(figsize=(15,8))
s=0
for color, label in zip(colors, np.asarray(cluster_labels_2).flatten()):
    subset = WorkingDF2[WorkingDF2.Cluster == label]    
    for i in subset.index:
        x=np.asarray(subset["Standardized COVID-19 Index"][i]).flatten()
        y=np.asarray(subset["Standardized CSS Index"][i]).flatten() 
        plt.text(x, y, str(subset['Neighbourhood'][i]), rotation=25) 
        s += 1
        plt.scatter(x, y, c=np.array([color]), label='cluster'+str(label),alpha=0.5)
plt.legend(loc='lower right', fontsize=15)
plt.xlabel('Standardized COVID-19 Index', fontsize=18)
plt.ylabel('Standardized CSS Index', fontsize=18)
plt.title("[Hierarchical Clustering: {} Cluster] \n 
 Mapping of Non-Outlier Neighbourhoods \n 
 onto Standardized CSS-COVID19 Indices Space \n
 ".format(N_Clus), fontsize=18)
print('# of Neighbours: ', s)

This is not reproducible as posted. Include sufficient data and imports for a [Minimal, Reproducible Example (e.g. code, data, errors, current output, expected output), as text](https://stackoverflow.com/help/minimal-reproducible-example). `NameError: name 'N_Clus' is not defined`. — Trenton McKinney, Aug 14 '20 at 18:18
Do the `plot` command outside of the loop that labels the `Neighbourhood`. — Jody Klymak, Aug 14 '20 at 18:19
Hi Trenton McKinney , Thanks for your note. I will delete this for now and think how I can create reproducible code, since the original one is too long. Sorry that I am new here and still learning about the rule. Thanks — Pulpito, Aug 14 '20 at 18:26
See [How to provide a reproducible copy of your DataFrame using `df.head(30).to_clipboard()`](https://stackoverflow.com/questions/52413246), then **[edit] your question**, and paste the clipboard into a code block. _A proper question **MUST** provide **ALL** of the information necessary in order for a proper answer to be given._ — Trenton McKinney, Aug 14 '20 at 18:27
@TrentonMcKinney, , as per your kind advice, I tried to_clipboard() several times, but unfortunately not successful. Nevertheless, thank you very much for your note on the policy. I am still new here. So, I will comply with the policy goingforward. Thanks again. — Pulpito, Aug 20 '20 at 10:06
I know it doesn’t work for google cola if you’re using that. There are alternate instructions at the bottom of the linked answer if using colas. There are other links in the answer for creating a synthetic test set. — Trenton McKinney, Aug 20 '20 at 14:35

Guimoute · Accepted Answer · 2020-08-21T09:27:50.660

The issue comes from the line

plt.scatter(x, y, c=np.array([color]), label='cluster'+str(label),alpha=0.5)

Here, you give the colored dot a label 'cluster' + str(label) even if such a label already exists, so plt.legend() will create many identical legend elements. I would keep track of the previous labels and set the current plot's one to None if it's not new so that the plt.legend() ignores it.

Notice that your naming choice can be a little confusing as matplotlib uses "label" as the name of a curve that appears in the legend, while you use it as a cluster number. Can we just call it cluster_number?

Here is the implementation:

import matplotlib.cm as cm

colors = cm.rainbow(np.linspace(0, 1, N_Clus))
cluster_labels_2 = list(range(1, N_Clus+1))
print("cluster_labels: ", cluster_labels_2)

# Create a figure.
plt.figure(figsize=(15, 8))
s=0
clusters_already_in_the_legend = []
for color, cluster_number in zip(colors, np.asarray(cluster_labels_2).flatten()):
    subset = WorkingDF2[WorkingDF2.Cluster == cluster_number]    
    for i in subset.index:
        x = np.asarray(subset["Standardized COVID-19 Index"][i]).flatten()
        y = np.asarray(subset["Standardized CSS Index"][i]).flatten() 
        plt.text(x, y, str(subset['Neighbourhood'][i]), rotation=25) 
        s += 1

        # Keeping track of the labels so that we don't legend them multiple times.
        if cluster_number not in clusters_already_in_the_legend:
            clusters_already_in_the_legend.append(cluster_number)
            label = f"Cluster {cluster_number}"
        else:
            label = None
        plt.scatter(x, y, c=np.array([color]), label=label, alpha=0.5)

plt.legend(loc='lower right', fontsize=15)
plt.xlabel('Standardized COVID-19 Index', fontsize=18)
plt.ylabel('Standardized CSS Index', fontsize=18)
plt.title("[Hierarchical Clustering: {} Cluster] \n 
 Mapping of Non-Outlier Neighbourhoods \n 
 onto Standardized CSS-COVID19 Indices Space \n
 ".format(N_Clus), fontsize=18)
print('# of Neighbours: ', s)

Thanks for your advice. After some minor modification, it worked! Brilliant! — Pulpito, Aug 20 '20 at 09:57
Two modifications: 1) subset = WorkingDF2[WorkingDF2.Cluster == cluster_number]; 2) label==> label_ — Pulpito, Aug 20 '20 at 10:03

Redundant Legends: Matplotlib

1 Answers1