Data Visualization is about taking data and representing it visually to make large data interpretable to humans. In this tutorial, we will be using matplotlib and seaborn.
When you have a big amount of air quality data or pollen data it's hard to visualize it or know what to do with it. Data visualization allows us to look at trends and patterns in the data to facilitate decision making.
Python has a lot of data visualization libraries for common type of visualizations. Some of the major libraries are:
In this tutorial, we will be using matplotlib and seaborn and looking at common types of plots. We will also be looking at easy ways to make graphs prettier.
Note: Definitions taken straight from Wikipedia.
Matplotlib has 3 different layers, each layer has different level of customization.
Different layers of matplotlib are:
We will be looking at the scripting layer since its the most easy to use. Scripting layer can be used using matplotlib.pyplot.
Lets start by importing the libraries. We'll import pandas for reading the data and matplotlib for plotting.
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib import style
We are going to use two datasets, just for demonstrating the things we can do with matplotlib and seaborn.
df1=pd.read_csv("India GDP from 1961 to 2017.csv")
If you want to plot something quickly, you can use pyplot.plot function. If you pass a single column it'll plot it against the index. If you pass two columns, you'll get a line graph by default and a scatter plot if you pass the string 'o'. We can also change the color of the data points being plotted by passing two character strings like 'ro'. We can also use different markers such as '+'.
You might have noticed that our plots do not look visually pleasing. One easy trick to make it instantly look better is to use matplotlib.style. It comes with a large amount of styles, so feel free to experiment. In this tutorial, we are going to use the style 'ggplot' which is based on the famous R library ggplot2.
We can use style.available to get a list of all the styles.
Now, let us set style by using style.use(). We'll use 'ggplot' as argument.
We can notice that our plot looks instantly better.
Scatter plot can be generated by using plt.scatter() and passing two arguments for x and y.
Now, lets look at coloring the data points by their species. To do this, we need more than a single line of code. Lets see how we can do it. First, we need to create groups using pandas and we can plot the scatter plot using a for loop. We shall also go ahead and add labels for x and y axis and add a title. We can do this by using xlabel, ylabel and title methods in pyplot.
for name, group in groups:
plt.title('Sepal Length VS Sepal Width')
We can plot a bar plot by using plt.bar() and passing x and y values. x-axis generally contains categorical value and y contains a numerical value.
Plotting a pie chart in matplotlib is relatively easy. First, let us extract the counts for the species column and then plot it using matplotlib. We shall add the legend as well and set its position using bbox_to_anchor parameter.
Line plots are pretty easy in matplotlib. We can use plt.plot to generate it. We shall visualize India's GDP percentage from 1960-2017
We can generate histogram using plt.hist() method and pass a numerical column as parameter, we can also set the number of bins and play with a few style options.
We can generate boxplot by using plt.boxplot(). However, if we want to group the boxplot by species like we did with scatter, we can use panda's implementation of boxplot using df.boxplot() instead since its pretty straight forward.
Seaborn is a plotting library build on top of matplotlib. It provides an easy way to produce good looking plots. Since it is built upon matplotlib, both can be used together to enhance plots generated by the other. We will be looking at few examples for using seaborn to enhance matplotlib plots.
Lets start by using seaborn's aesthetics methods to style matplotlib plots. We generally use the alias 'sns' for seaborn.
sns.set() will set seaborn styling to matplotlib plots.
import seaborn as sns
sns.set_style() can be used to set the grid style. The following example shows how to use it.
sns.set_palette() can be used select color palettes. Seaborn has some built in palettes but built in or custom palettes can also be used. We will be looking into it later in the tutorial.
Now, lets look at how to plot different types of charts in seaborn. We will be looking at the same types of charts that we looked at when we were using matplotlib.
Scatter Plot can be generated using sns.scatterplot(). We can also group data points to categories just by passing the hue parameter with the column containing categorical values.
Let us create a custom palette and use it. It is as simple as creating a list of colors in hexadecimal values.
Now, let us alter the scatter plot and add a fourth variable which we will represent using 'size' parameter. The resulting plot is called a bubble chart.
Bar plot can be generated using sns.barplot and passing x and y values.
seaborn also provides countplot method from which you can see the counts of a categorical column.
Box plot can be generated using sns.boxplot() method. Passing a single column will generate a single box plot. Passing the categorical variable with the column will group the values by the categorical variable values. We can pass it in any order and it'll only result in horizontal and vertical alignment differences.
Histogram can be generated using sns.distplot(). Seaborn provides both bar histogram and line histograms by default.
Line plot can be generated using sns.lineplot and passing x and y values. As simple as that.
Unfortunately, seaborn does not support pie charts. But, we can still use seaborn's styling to generate pie charts using matplotlib.
A good selection of colors can really enhance data visualization experience. Although both matplotlib and seaborn come with built in styling and palettes and we can always use custom palettes, it is really handy to have a wide selection of existing palettes available. Enter palettable. It is a library which contains a wide variety of palettes from different data visualization tools and libraries that can be used alongside with matplotlib and seaborn. Let us see how we can use it.
We will be using one of my favorite palettes called Prism. Here are a few examples:
from palettable.cartocolors.qualitative import Prism_10
There are multiple palettes available in palettable so feel free to mess around.
We can save the generated graphs in high resolution using plt.savefig() method. We can set the figure size and dpi using plt.figure(). If we want a transparent background, we can use transparent=True in savefig() method.
That's it for this tutorial. In the future, I'll be covering interactive plots using plotly. Thanks for reading and have fun 'plotting'.