June 8, 2022
Pandas is a data analysis tool that also offers great options for data visualization. Here’s how to get started plotting in Pandas.
Data visualization is an essential step in making data science projects successful — an effective plot tells a thousand words.
Data visualization is a powerful way to capture trends and share the insights gained from data. There are plenty of data visualization tools on the shelf with a lot of outstanding features, but in this tutorial, we’re going to learn plotting with the Pandas package.
As you may already know, Pandas is a data analysis tool, but it provides some great options for data visualization. At the end of this tutorial, you’ll see how easy and straightforward plotting with Pandas can be.
We assume that you know the fundamentals of Pandas DataFrames. If you’re not familiar with the Pandas library, you might like to try our Pandas and NumPy Fundamentals course.
Let’s dive in.
Pandas comes with a couple of plotting functionalities applicable on DataFrame- or series objects that use the Matplotlib library under the hood, which means any plot created by the Pandas library is a Matplotlib object.
Technically, the Pandas
plot() method provides a set of plot styles through the
kind keyword argument to create decent-looking plots. The default value of the
kind argument is the
line string value. However, there are eleven different string values that can be assigned to the
kind argument, which determines what kind of plot we’ll create.
.plot is also an attribute of Pandas DataFrame and series objects, providing a small subset of plots available with Matplotlib. In fact, Pandas makes plotting as simple as just writing a single line of code by automating much of the data visualization procedure for us.
Importing the Libraries and the Dataset
In this tutorial, we’re going to work on the weekly closing price of the Facebook, Microsoft, and Apple stocks over the last previous months. The following code imports the necessary libraries and the dataset required for visualization and then displays the content of the DataFrame on the output. The
%matplotlib inline magic command is also added to the code to ensure the plotted figures appear in the notebook cells correctly:
import pandas as pdimport numpy as npimport matplotlib.pyplot as pltdataset_url = ('https://raw.githubusercontent.com/m-mehdi/pandas_tutorials/main/weekly_stocks.csv')df = pd.read_csv(dataset_url, parse_dates=['Date'], index_col='Date')pd.set_option('display.max.columns', None)print(df.head())
MSFT FB AAPLDate 2021-05-24 249.679993 328.730011 124.6100012021-05-31 250.789993 330.350006 125.8899992021-06-07 257.890015 331.260010 127.3499982021-06-14 259.429993 329.660004 130.4600072021-06-21 265.019989 341.369995 133.110001
We’re now ready to explore and visualize the data with Pandas.
The default plot is the line plot that plots the index on the x-axis and the other numeric columns in the DataFrame on the y-axis.
Let’s plot a line plot and see how Microsoft performed over the previous 12 months:
figsize argument takes two arguments, width and height in inches, and allows us to change the size of the output figure. The default values of the width and height are 6.4 and 4.8, respectively.
We can plot multiple lines from the data by providing a list of column names and assigning it to the y-axis. For example, let’s see how the three companies performed over the previous year:
df.plot.line(y=['FB', 'AAPL', 'MSFT'], figsize=(10,6))
We can use the other parameters provided by the
plot() method to add more details to a plot, like this:
df.plot(y='FB', figsize=(10,6), title='Facebook Stock', ylabel='USD')
As we see in the figure, the
title argument adds a title to the plot, and the
ylabel sets a label for the y-axis of the plot. The plot’s legend display by default, however, we may set the
legend argument to
false to hide the legend.
A bar chart is a basic visualization for comparing values between data groups and representing categorical data with rectangular bars. This plot may include the count of a specific category or any defined value, and the lengths of the bars correspond to the values they represent.
In the following example, we’ll create a bar chart based on the average monthly stock price to compare the average stock price of each company to others in a particular month. To do so, first, we need to resample data by month-end and then use the
mean() method to calculate the average stock price in each month. We also select the last three months of data, like this:
df_3Months = df.resample(rule='M').mean()[-3:]print(df_3Months)
MSFT FB AAPLDate 2022-03-31 298.400002 212.692505 166.9349982022-04-30 282.087494 204.272499 163.7049942022-05-31 262.803335 198.643331 147.326665
Now, we’re ready to create a bar chart based on the aggregated data by assigning the
bar string value to the
df_3Months.plot(kind='bar', figsize=(10,6), ylabel='Price')
We can create horizontal bar charts by assigning the
barh string value to the
kind argument. Let’s do it:
We can also plot the data on the stacked vertical or horizontal bar charts, which represent different groups on top of each other. The height of the resulting bar shows the combined result of the groups. To create a stacked bar chart we need to assign
True to the
stacked argument, like this:
df_3Months.plot(kind='bar', stacked=True, figsize=(9,6))
A histogram is a type of bar chart that represents the distribution of numerical data where the x-axis represents the bin ranges while the y-axis represents the data frequency within a certain interval.
df[['MSFT', 'FB']].plot(kind='hist', bins=25, alpha=0.6, figsize=(9,6))
In the example above, the
bins argument specifies the number of bin intervals, and the
alpha argument specifies the degree of transparency.
A histogram can also be stacked. Let’s try it out:
df[['MSFT', 'FB']].plot(kind='hist', bins=25, alpha=0.6, stacked=True, figsize=(9,6))
A box plot consists of three quartiles and two whiskers that summarize the data in a set of indicators: minimum, first quartile, median, third quartile, and maximum values. A box plot conveys useful information, such as the interquartile range (IQR), the median, and the outliers of each data group. Let’s see how it works:
We can create horizontal box plots, like horizontal bar charts, by assigning
False to the
vert argument. Like this:
df.plot(kind='box', vert=False, figsize=(9,6))
An area plot is an extension of a line chart that fills the region between the line chart and the x-axis with a color. If more than one area chart displays in the same plot, different colors distinguish different area charts. Let’s try it out:
plot() method creates a stacked area plot by default. It’s a common task to unstack the area chart by assigning
False to the
df.plot(kind='area', stacked=False, figsize=(9,6))
If we’re interested in ratios, a pie plot is a great proportional representation of numerical data in a column. The following example shows the average Apple stock price distribution over the previous three months:
df_3Months.index=['March', 'April', 'May']df_3Months.plot(kind='pie', y='AAPL', legend=False, autopct='%.f')
A legend will display on pie plots by default, so we assigned
False to the
legend keyword to hide the legend.
The new keyword argument in the code above is
autopct, which shows the percent value on the pie chart slices.
If we want to represent the data of all the columns in multiple pie charts as subplots, we can assign
True to the
subplots argument, like this:
df_3Months.plot(kind='pie', legend=False, autopct='%.f', subplots=True, figsize=(14,8))
, , ], dtype=object)
Scatter plots plot data points on the x and y axes to show the correlation between two variables. Like this:
df.plot(kind='scatter', x='MSFT', y='AAPL', figsize=(9,6), color='Green')
As we see in the plot above, the scatter plot shows the relationship between Microsoft and Apple stock prices.
When the data is very dense, a hexagon bin plot, also known as a hexbin plot, can be an alternative to a scatter plot. In other words, when the number of data points is enormous, and each data point can’t be plotted separately, it’s better to use this kind of plot that represents data in the form of a honeycomb. Also, the color of each hexbin defines the density of data points in that range.
df.plot(kind='hexbin', x='MSFT', y='AAPL', gridsize=10, figsize=(10,6))
gridsize argument specifies the number of hexagons in the x-direction. A larger grid size means more and smaller bins. The default value of the
gridsize argument is 100.
The last plot we want to discuss in this tutorial is the Kernel Density Estimate, also known as KDE, which visualizes the probability density of a continuous and non-parametric data variable. This plot uses Gaussian kernels to estimate the probability density function (PDF) internally. Let’s try it out:
We’re can also specify the bandwidth that affects the plot smoothness in the KDE plot, like this:
As we can see, selecting a small bandwidth leads to under-smoothing, which means the density plot appears as a combination of individual peaks. On the contrary, a huge bandwidth leads to over-smoothing, which means the density plot appears as unimodal distribution.
In this tutorial, we discussed the capabilities of the Pandas library as an easy-to-learn and straightforward data visualization tool. Then, we covered all the plots provided in Pandas by implementing some examples with very few lines of code.
Feel free to connect with me on LinkedIn and Twitter.