Matplotlib¶
Python offers sophisticated plotting capabilities. Plotting can be approached in two ways:
- Pythonic Approach: In this method, an empty object is created, and plots are constructed programmatically, then assigned to the empty object using code.
- Non-Pythonic Approach: This approach relies on external libraries like
matplotlib, which provides user-friendly tools for interactive plotting. A common shorthand for importing this module isimport matplotlib.pyplot as plt.
In [1]:
Copied!
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
CHD=pd.read_csv('../data/CHD_test.csv',index_col=False)
CHD.head()
import matplotlib.pyplot as plt import numpy as np import pandas as pd CHD=pd.read_csv('../data/CHD_test.csv',index_col=False) CHD.head()
Out[1]:
| latitude | housing_median_age | total_rooms | total_bedrooms | population | households | median_income | median_house_value | famlev | |
|---|---|---|---|---|---|---|---|---|---|
| 0 | 34.19 | 15 | 5612 | 1283 | 1015 | 472 | 1.4936 | 66900 | L |
| 1 | 34.40 | 19 | 7650 | 1901 | 1129 | 463 | 1.8200 | 80100 | L |
| 2 | 33.69 | 17 | 720 | 174 | 333 | 117 | 1.6509 | 85700 | L |
| 3 | 33.64 | 14 | 1501 | 337 | 515 | 226 | 3.1917 | 73400 | M |
| 4 | 33.57 | 20 | 1454 | 326 | 624 | 262 | 1.9250 | 65500 | L |
Scatter plot¶
The most commonly used plot is the scatter plot. Here are the following scripts that generate random numbers and create a scatter plot:
In [2]:
Copied!
x = CHD.median_house_value
y = CHD.median_income
plt.scatter(x, y)
plt.xlabel('median_house_value')
plt.ylabel('median_income')
plt.title('Scatter Plot')
plt.show(block=False)
x = CHD.median_house_value y = CHD.median_income plt.scatter(x, y) plt.xlabel('median_house_value') plt.ylabel('median_income') plt.title('Scatter Plot') plt.show(block=False)
Indeed, you can customize the appearance of a scatter plot using various arguments:
- Size of point (
s): you can adjust the size of the points, for example,s=10make the point smaller, - colour (
c): You can specify the color of pointy. For example,c=redwill make the points red. marker: you can choose different marker styles for points. For examplemarker=sw ill use squares for presenting points.
In [7]:
Copied!
plt.scatter(x, y,s=40, c='red', marker='s')
plt.xlabel('median_house_value')
plt.ylabel('median_income')
plt.title('Scatter Plot')
plt.show(block=False)
plt.scatter(x, y,s=40, c='red', marker='s') plt.xlabel('median_house_value') plt.ylabel('median_income') plt.title('Scatter Plot') plt.show(block=False)
In this example, we've customized the scatter plot to use red square for data points with a larger size (s=40) to label the points. You can further explore the links provided for more marker styles and line properties in matplotlib. The following figure displays a more sophisticated plot.
In [8]:
Copied!
select = (CHD.famlev == 'L')
plt.scatter(x, y, alpha=0.3)
plt.xlabel('median_house_value')
plt.ylabel('median_income')
plt.title('Scatter Plot')
plt.scatter(x[select], y[select], facecolor='none', edgecolors='r')
plt.show(block=False)
select = (CHD.famlev == 'L') plt.scatter(x, y, alpha=0.3) plt.xlabel('median_house_value') plt.ylabel('median_income') plt.title('Scatter Plot') plt.scatter(x[select], y[select], facecolor='none', edgecolors='r') plt.show(block=False)
Fit a linear model to a sample dataset.
In [9]:
Copied!
fig, ax = plt.subplots()
ax.scatter(x, y, alpha=0.5, color='orchid')
fig.suptitle('Scatter Plot')
fig.tight_layout(pad=2);
ax.grid(True)
fit = np.polyfit(x, y, deg=1)
ax.plot(x, fit[0]*x + fit[1], '-',color='red', linewidth=2)
fig, ax = plt.subplots() ax.scatter(x, y, alpha=0.5, color='orchid') fig.suptitle('Scatter Plot') fig.tight_layout(pad=2); ax.grid(True) fit = np.polyfit(x, y, deg=1) ax.plot(x, fit[0]*x + fit[1], '-',color='red', linewidth=2)
Out[9]:
[<matplotlib.lines.Line2D at 0x14052de10>]
subplot¶
You can create multiple subplots in a single figure using the .subplot(#row,#col,position) method.
In [10]:
Copied!
plt.subplot(2, 1, 1)
plt.scatter(x, y)
plt.title("Fig1")
plt.xlabel("median house value")
plt.ylabel("median income")
plt.subplot(2, 1, 2)
plt.scatter(x, CHD.population)
plt.title("Fig2")
plt.xlabel("median house value")
plt.ylabel("population")
plt.show(block=False)
plt.subplot(2, 1, 1) plt.scatter(x, y) plt.title("Fig1") plt.xlabel("median house value") plt.ylabel("median income") plt.subplot(2, 1, 2) plt.scatter(x, CHD.population) plt.title("Fig2") plt.xlabel("median house value") plt.ylabel("population") plt.show(block=False)
Seaborn¶
Seaborn provides advanced graphical capabilities for creating sophisticated statistical visualizations with ease. It simplifies the process of generating complex plots from pandas DataFrames using simple commands.
In [8]:
Copied!
import seaborn as sns
sns.set(color_codes=True)
CHD['median_income'] = (CHD['median_income'] -CHD['median_income'].mean()) / CHD['median_income'].std()
CHD['median_house_value'] = (CHD['median_house_value'] -CHD['median_house_value'].mean()) / CHD['median_house_value'].std()
import seaborn as sns sns.set(color_codes=True) CHD['median_income'] = (CHD['median_income'] -CHD['median_income'].mean()) / CHD['median_income'].std() CHD['median_house_value'] = (CHD['median_house_value'] -CHD['median_house_value'].mean()) / CHD['median_house_value'].std()
In [6]:
Copied!
for col in ['median_income','median_house_value']:
plt.hist(CHD[col], density=True)
for col in ['median_income','median_house_value']: plt.hist(CHD[col], density=True)
We can get a smooth estimate of the distribution using a kernel density estimation (KDE):
In [9]:
Copied!
import warnings
warnings.filterwarnings("ignore")
sns.kdeplot(data=CHD, x='median_income', y='median_house_value')
import warnings warnings.filterwarnings("ignore") sns.kdeplot(data=CHD, x='median_income', y='median_house_value')
Out[9]:
<Axes: xlabel='median_income', ylabel='median_house_value'>
You can create a hexagonally-based histogram using jointplot:
In [10]:
Copied!
sns.jointplot(data=CHD, x='median_income', y='median_house_value',kind="hex")
sns.jointplot(data=CHD, x='median_income', y='median_house_value',kind="hex")
Out[10]:
<seaborn.axisgrid.JointGrid at 0x168785bd0>
In [11]:
Copied!
sns.jointplot(data=CHD, x='median_income', y='median_house_value',kind="kde", hue='famlev')
sns.jointplot(data=CHD, x='median_income', y='median_house_value',kind="kde", hue='famlev')
Out[11]:
<seaborn.axisgrid.JointGrid at 0x168f77e10>
The following illustrates how to draw a box plot for different family levels.
In [14]:
Copied!
g=sns.catplot(data=CHD, x='median_income', y='famlev', kind="box")
g.set_axis_labels("Income", "Family level");
g=sns.catplot(data=CHD, x='median_income', y='famlev', kind="box") g.set_axis_labels("Income", "Family level");
Pairplots¶
We can generalize joint plots for multidimensional data, which is very useful for exploring correlations between multiple dimensions of data.
In [11]:
Copied!
sns.pairplot(CHD, hue='famlev');
sns.pairplot(CHD, hue='famlev');
--------------------------------------------------------------------------- NameError Traceback (most recent call last) /Users/samamiri/Library/CloudStorage/GoogleDrive-saeid.amiri1@gmail.com/My Drive/python/Python-for-Data-Analysis/notebooks/06-visualization.ipynb Cell 25 line 1 ----> <a href='vscode-notebook-cell:/Users/samamiri/Library/CloudStorage/GoogleDrive-saeid.amiri1%40gmail.com/My%20Drive/python/Python-for-Data-Analysis/notebooks/06-visualization.ipynb#X31sZmlsZQ%3D%3D?line=0'>1</a> sns.pairplot(CHD, hue='famlev'); NameError: name 'sns' is not defined
GGPLOT2¶
ggplot2 is a very useful package in R for creating advanced plots. In Python, the plotnine library is used to create ggplot2-like plots. You can import the module using import plotnine as p9. Generating plots in ggplot2 (plotnine) follows a structured series of steps, which can be accomplished via:
- initialize it
In [14]:
Copied!
import plotnine as p9
p9.ggplot(data=CHD)
import plotnine as p9 p9.ggplot(data=CHD)
Out[14]:
<Figure Size: (640 x 480)>
- Define aesthetics using
aesand specify your arguments. The most important aesthetics include:x,y,alpha,color,colour,fill,linetype,shape,size, andstroke. To create variations of the plot with different parameters, you can assign it to a variable.
In [15]:
Copied!
CHD_plot=p9.ggplot(data=CHD,mapping=p9.aes(x='median_income', y='median_house_value'))
CHD_plot=p9.ggplot(data=CHD,mapping=p9.aes(x='median_income', y='median_house_value'))
- Specify what you want to display and use the
+operator to add layers and customize your plot.
In [16]:
Copied!
CHD_plot+p9.geom_point()
CHD_plot+p9.geom_point()
Out[16]:
<Figure Size: (640 x 480)>
You can easily add scale and define label:
In [35]:
Copied!
CHD_plot+ p9.geom_point(alpha=0.15)+ p9.xlab("median_income") + p9.ylab("median_house_value")+ p9.scale_x_log10()+ p9.theme_bw()+ p9.theme(text=p9.element_text(size=10))
CHD_plot+ p9.geom_point(alpha=0.15)+ p9.xlab("median_income") + p9.ylab("median_house_value")+ p9.scale_x_log10()+ p9.theme_bw()+ p9.theme(text=p9.element_text(size=10))
Out[35]:
<Figure Size: (640 x 480)>
- After creating your plot, you can save it to a file in your favourite format
In [36]:
Copied!
CHD_plot2 = CHD_plot+p9.geom_point()
CHD_plot2.save("CHD_plot.png", dpi=300)
CHD_plot2 = CHD_plot+p9.geom_point() CHD_plot2.save("CHD_plot.png", dpi=300)
/home/sam/venv/lib/python3.10/site-packages/plotnine/ggplot.py:587: PlotnineWarning: Saving 6.4 x 4.8 in image. /home/sam/venv/lib/python3.10/site-packages/plotnine/ggplot.py:588: PlotnineWarning: Filename: CHD_plot.png
bar chart¶
To generate a bar chart, you can use geom_bar()
In [37]:
Copied!
(p9.ggplot(data=CHD,mapping=p9.aes(x='famlev'))+ p9.geom_bar())
(p9.ggplot(data=CHD,mapping=p9.aes(x='famlev'))+ p9.geom_bar())
Out[37]:
<Figure Size: (640 x 480)>
Plotting distributions¶
- A boxplot can be created using
geom_boxplot():
In [38]:
Copied!
(p9.ggplot(data=CHD,
mapping=p9.aes(x='famlev',
y='median_income'))
+ p9.geom_boxplot()
+ p9.scale_y_log10()
)
(p9.ggplot(data=CHD, mapping=p9.aes(x='famlev', y='median_income')) + p9.geom_boxplot() + p9.scale_y_log10() )
Out[38]:
<Figure Size: (640 x 480)>
- To add points behind the boxplot, you can use geom_jitter() to plot the points with some random noise to avoid overlapping points. This will create a visual representation of the data points behind the boxplot. Here's an example:
In [39]:
Copied!
(p9.ggplot(data=CHD,
mapping=p9.aes(x='famlev',
y='median_income'))
+ p9.geom_boxplot()
+ p9.geom_jitter(alpha=0.1, color="green")
+ p9.scale_y_log10()
)
(p9.ggplot(data=CHD, mapping=p9.aes(x='famlev', y='median_income')) + p9.geom_boxplot() + p9.geom_jitter(alpha=0.1, color="green") + p9.scale_y_log10() )
Out[39]:
<Figure Size: (640 x 480)>