Prevent overlapping labels in grouped bar chart using matplotlib python-CodePudding

I am using the following function to plot a grouped bar chart that compares between the performance of models built using imbalanced and balanced data.

# plot results to compare between balanced and imbalanced data

def barChartBalancing(imbalancedResults, rusResults, smoteResults, score, title, string):
    import matplotlib
    import matplotlib.pyplot as plt
    import numpy as np

    #make font bigger
    font = {'size'   : 15}

    matplotlib.rc('font', **font)
    
    labels = names
    
    x = np.arange(len(labels))  # the label locations
    width = 0.2  # the width of the bars

    fig, ax = plt.subplots(figsize=(10,9))
    rects1 = ax.bar(x - width, imbalancedResults, width, label='Imbalanced Dataset')
    rects2 = ax.bar(x , rusResults, width, label='RandomUnderSampler')
    rects3 = ax.bar(x   width, smoteResults, width, label='SMOTE')
    

    # Add some text for labels, title and custom x-axis tick labels, etc.
    ax.set_ylabel(score)
    ax.set_title(title)
    ax.set_xticks(x)
    ax.set_xticklabels(labels)
    ax.legend(loc='upper center')
    

    ax.bar_label(rects1, padding=5, fmt='%.2f', label_type='edge')
    ax.bar_label(rects2, padding=5, fmt='%.2f', label_type='edge')
    ax.bar_label(rects3, padding=5, fmt='%.2f', label_type='edge')

    fig.tight_layout()
    
    fileName = string  '.png'
    print(fileName)
    plt.savefig('figures/resampling/'   fileName)
    
    plt.show()

However, when I run this, the labels above the grouped bars are overlapping, like so:

I tried changing the 'padding' value but then realised that is the distance between the and the end of the bar. I also tried changing the font size using fontsize argument in bar_label based on Rabinzel's comment, this kind of helped, but I have to make the font extremely small to prevent the overlapping completely, to the point where it is very hard to read.

Any ideas how I can fix the overlapping? I wonder if I can position the labels vertically on the bar? I think that would solve the issue without having to use extremely small font sizes.

CodePudding user response：

I dont know if you have consider plotting your figure horizontally, but with this approach you will be able to:

have longer names as labels
you will not have the overlapping problem.
an unlimited number of models (entries) in the same figure.

import matplotlib.pyplot as plt
import numpy as np

fig, ax = plt.subplots()

# Example data
models = ('Linear regresion', 'Singular vector machine', 'linear vector clasification',
          'Naive Bayes Classifier', 'Gradient boosting')
y_pos = np.arange(len(models))
distance = 3   10 * np.random.rand(len(models))
speed = 3   10 * np.random.rand(len(models))
age = 3   10 * np.random.rand(len(models))
width = 0.2 

rects1=ax.barh(y_pos-width, distance, width, align='center', label="Unvalance dataset")
rects2=ax.barh(y_pos, speed, width, align='center', label="Random under sapler")
rects3=ax.barh(y_pos   width, age, width, align='center', label="SMOTE")
ax.set_yticks(y_pos, labels=people)
ax.invert_yaxis()  # labels read top-to-bottom
ax.set_xlabel('Performance')
ax.bar_label(rects1, padding=5, fmt='%.2f', label_type='edge')
ax.bar_label(rects2, padding=5, fmt='%.2f', label_type='edge')
ax.bar_label(rects3, padding=5, fmt='%.2f', label_type='edge')
plt.legend()
plt.xlim(0,15)
plt.show()

CodePudding user response：

Here is how you change fontsize and rotate the bar_label:

example on one of them:

ax.bar_label(rects1, padding=5, fmt='%.2f', label_type='edge', fontsize=9, rotation='vertical')

You can also pass an integer to rotation if you like to have a different rotation: rotation=45