5

I would like to annotate a heatmap with the values that I pass from a dataframe into the function below. I have looked at matplotlib.text but have not been able to get the values from my dataframe in a desired way in my heatmap. I have pasted in my function for generating a heatmap below, after that my dataframe and the output from the heatmap call. I would like to plot each value from my dataframe in the center of each cell in the heatmap.

Function for generating a heatmap:

import matplotlib as mpl
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors

def heatmap_binary(df,
            edgecolors='w',
            #cmap=mpl.cm.RdYlGn,
            log=False):    
    width = len(df.columns)/7*10
    height = len(df.index)/7*10

    fig, ax = plt.subplots(figsize=(20,10))#(figsize=(width,height))

    cmap, norm = mcolors.from_levels_and_colors([0, 0.05, 1],['Teal', 'MidnightBlue'] ) # ['MidnightBlue', Teal]['Darkgreen', 'Darkred']

    heatmap = ax.pcolor(df ,
                        edgecolors=edgecolors,  # put white lines between squares in heatmap
                        cmap=cmap,
                        norm=norm)


    ax.autoscale(tight=True)  # get rid of whitespace in margins of heatmap
    ax.set_aspect('equal')  # ensure heatmap cells are square
    ax.xaxis.set_ticks_position('top')  # put column labels at the top
    ax.tick_params(bottom='off', top='off', left='off', right='off')  # turn off ticks

    plt.yticks(np.arange(len(df.index)) + 0.5, df.index, size=20)
    plt.xticks(np.arange(len(df.columns)) + 0.5, df.columns, rotation=90, size= 15)

    # ugliness from http://matplotlib.org/users/tight_layout_guide.html
    from mpl_toolkits.axes_grid1 import make_axes_locatable
    divider = make_axes_locatable(ax)
    cax = divider.append_axes("right", "3%", pad="1%")
    plt.colorbar(heatmap, cax=cax)


plt.show()

Herre is an example of My dataframe :

dataframe :

             0-5 km / h  5-40 km / h  40-80 km / h  80-120 km / h  \
NORDIC         0.113955     0.191888      0.017485      -0.277528   
MIDDLE  EU     0.117903     0.197084     -0.001447      -0.332677   
KOREA          0.314008     0.236503     -0.067174      -0.396518   
CHINA          0.314008     0.236503     -0.067174      -0.396518   

             120-160 km / h  160-190 km / h  190 km / h  
NORDIC            -0.054365        0.006107    0.002458  
MIDDLE  EU         0.002441        0.012097    0.004599  
KOREA             -0.087191        0.000331    0.000040  
CHINA             -0.087191        0.000331    0.000040

Generating the heatmap:

heatmap_binary(dataframe)

enter image description here

Any ideas?

Update to clarify my problem

I tried the proposed solution from question which has the result I'm looking for: how to annotate heatmap with text in matplotlib? However, I still have a problem using the matplotlib.text function for positioning the values in the heatmap: Here is my cod for trying this solution:

import matplotlib.pyplot as plt
import numpy as np


data = dataframe.values
heatmap_binary(dataframe)

for y in range(data.shape[0]):
    for x in range(data.shape[1]):
        plt.text(data[y,x] +0.05 , data[y,x] + 0.05, '%.4f' % data[y, x], #data[y,x] +0.05 , data[y,x] + 0.05
                 horizontalalignment='center',
                 verticalalignment='center',
                 color='w')

#plt.colorbar(heatmap)

plt.show()

added plot: (different coloring but same problem) enter image description here

Improve this question
4
  • What goes wrong when you try this? Can you post an image? Commented Jan 15, 2014 at 16:57
  • Added a picture, different coloring but same problem though. The text call places the values in the right hand corner.... Commented Jan 15, 2014 at 17:16
  • Can you post a copy of your code all at once, not in two distinct chunks? I think the problem is that you're running your for loop after you've added the colorbar. Try running the text for loop immediately after the call to pcolor. Commented Jan 15, 2014 at 21:14
  • I tried to put the for loop iside the heatmap_binary function after the pcolor...did not work though...I got the same result Commented Jan 16, 2014 at 15:04

4 Answers 4

Reset to default
9

This functionality is provided by the seaborn package. It can produce maps like

Example annotated heatmap

An example usage of seaborn is

import seaborn as sns
sns.set()

# Load the example flights dataset and conver to long-form
flights_long = sns.load_dataset("flights")
flights = flights_long.pivot("month", "year", "passengers")

# Draw a heatmap with the numeric values in each cell
sns.heatmap(flights, annot=True, fmt="d", linewidths=.5)
Improve this answer
Sign up to request clarification or add additional context in comments.

Comments

7

The values you were using for your coordinates in your for loop were screwed up. Also you were using plt.colorbar instead of something cleaner like fig.colorbar. Try this (it gets the job done, with no effort made to otherwise cleanup the code):

def heatmap_binary(df,
            edgecolors='w',
            #cmap=mpl.cm.RdYlGn,
            log=False):    
    width = len(df.columns)/7*10
    height = len(df.index)/7*10

    fig, ax = plt.subplots(figsize=(20,10))#(figsize=(width,height))

    cmap, norm = mcolors.from_levels_and_colors([0, 0.05, 1],['Teal', 'MidnightBlue'] ) # ['MidnightBlue', Teal]['Darkgreen', 'Darkred']

    heatmap = ax.pcolor(df ,
                        edgecolors=edgecolors,  # put white lines between squares in heatmap
                        cmap=cmap,
                        norm=norm)
    data = df.values
    for y in range(data.shape[0]):
        for x in range(data.shape[1]):
            plt.text(x + 0.5 , y + 0.5, '%.4f' % data[y, x], #data[y,x] +0.05 , data[y,x] + 0.05
                 horizontalalignment='center',
                 verticalalignment='center',
                 color='w')


    ax.autoscale(tight=True)  # get rid of whitespace in margins of heatmap
    ax.set_aspect('equal')  # ensure heatmap cells are square
    ax.xaxis.set_ticks_position('top')  # put column labels at the top
    ax.tick_params(bottom='off', top='off', left='off', right='off')  # turn off ticks

    ax.set_yticks(np.arange(len(df.index)) + 0.5)
    ax.set_yticklabels(df.index, size=20)
    ax.set_xticks(np.arange(len(df.columns)) + 0.5)
    ax.set_xticklabels(df.columns, rotation=90, size= 15)

    # ugliness from http://matplotlib.org/users/tight_layout_guide.html
    from mpl_toolkits.axes_grid1 import make_axes_locatable
    divider = make_axes_locatable(ax)
    cax = divider.append_axes("right", "3%", pad="1%")
    fig.colorbar(heatmap, cax=cax)

Then

df1 = pd.DataFrame(np.random.choice([0, 0.75], size=(4,5)), columns=list('ABCDE'), index=list('WXYZ'))
heatmap_binary(df1)

gives:

The Answer

Improve this answer

2 Comments

Happy to help. I wonder if this is worth proposing as an enhancement to Pandas. There are all sorts of things I wish I could do with Pandas DataFrames in this spirit. Color only certain columns. Normalize before coloring. Etc. The open question for me is "what's the right general approach"? I.e. Is it something based on pcolor or is it something that should involve the DataFrame HTML representation? What do you guys think?
Yes, such enhancement in diagraming would be helpful. In the spirit of simplifying data processing, they are very much welcomed. Especially, the labeling of values on heatmap is very much needed. So common, it would save some effort with some straightforward API.
1

This is because you're using plt.text after you've added another axes.

The state machine will plot on the current axes, and after you've added a new one with divider.append_axes, the colorbar's axes is the current one. (Just calling plt.colorbar will not cause this, as it sets the current axes back to the original one afterwards if it creates the axes itself. If a specific axes object is passed in using the cax kwarg, it doesn't reset the "current" axes, as that's not what you'd normally want.)

Things like this are the main reason that you'll see so many people advising that you use the OO interface to matplotlib instead of the state machine interface. That way you know which axes object that you're plotting on.

For example, in your case, you could have heatmap_binary return the ax object that it creates, and the plot using ax.text instead of plt.text (and similar for the other plotting methods).

Improve this answer

1 Comment

Thank you. So, do you have any suggestions of how to return an axis object from my function. I'm afraid I'm not so skilled in the python language..
1

You also can use plotly.figure_factory to create heatmap from DataFrame, but you have convert it into list.

    import plotly.figure_factory as ff

    z = [your_dataframe].values.tolist()
    x = [your_dataframe].columns.tolist()
    y = [your_dataframe].index.tolist()

    fig = ff.create_annotated_heatmap(z, x=x, y=y, annotation_text=z, colorscale='viridis')

    # for add annotation into Heatmap
    for i in range(len(fig.layout.annotations)):
        fig.layout.annotations[i].font.size = 12

    # show your Heatmap
    fig.show()
Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.