Python for data analysis
Python is more of a general purpose programming language than R or Matlab. It
has gradually become more popular for data analysis and scientific computing, but
additional modules are needed. Some of the more popular modules are:
NumPy N-dimensional array
SciPy Scientific computing (linear algebra, numerical integration,
optimization, etc)
Matplotlib 2D Plotting (similar to Matlab)
IPython Enhanced Interactive Console
Sympy Symbolic mathematics
Pandas Data analysis (provides a data frame structure similar to R)
NumPy, SciPy and Matplotlib are used in this presentation.
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 2 / 9
Creating N-dimensional arrays using NumPy
There are many ways to create N-dimensional arrays
import numpy as np
# Create 2X3 double precision array initialized to all zeroes
a = np.zeros((2,3), dtype=np.float64)
# Create array initialized by list of lists
a = np.array([[0,1,2],[3,4,5]], dtype=np.float64)
# Create array by reading CSV file
a = np.genfromtxt(’data.csv’, dtype=np.float64, delimiter=’,’)
# Create array using "arange" function
a = np.arange(6, dtype=np.float64).reshape(2,3)
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 3 / 9
Get values from N-dimensional array
NumPy provides many ways to extract data from arrays
# Print single element of 2D array
print a[0,0] # a scalar, not an array
# Print first row of 2D array
print a[0,:] # 1D array
# Print last column of array
print a[:,-1] # 1D array
# Print sub-matrix of 2D array
print a[0:2,1:3] # 2D array
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 4 / 9
Modifying N-dimensional arrays
NumPy uses the same basic syntax for modifying arrays
# Assign single value to single element of 2D array
a[0,0] = 25.0
# Assign 1D array to first row of 2D array
a[0,:] = np.array([10,11,12], dtype=np.float64)
# Assign 1D array to last column of 2D array
a[:,-1] = np.array([20,21], dtype=np.float64)
# Assign 2D array to sub-matrix of 2D array
a[0:2,1:3] = np.array([[10,11],[20,21]], dtype=np.float64)
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 5 / 9
Modifying arrays using broadcasting
# Assign scalar to first row of 2D array
a[0,:] = 10.0
# Assign 1D array to all rows of 2D array
a[:,:] = np.array([30,31,32], dtype=np.float64)
# Assign 1D array to all columns of 2D array
a[:,:] = np.array([40,41], dtype=np.float64).reshape(2,1)
# Assign scalar to sub-matrix of 2D array
a[0:2,1:3] = 100.0
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 6 / 9
Arithmetic on arrays
Operate on arrays using binary operators and NumPy functions
# Create 1D array
a = np.arange(4, dtype=np.float64)
# Add 1D arrays elementwise
a + a
# Multiply 1D arrays elementwise
a * a
# Sum elements of 1D array
a.sum()
# Compute dot product
np.dot(a, a) # same as: (a * a).sum()
# Compute cross product
np.dot(a.reshape(4,1), a.reshape(1,4))
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 7 / 9
SciPy Linear Algebra functions
import numpy as np
from scipy import linalg
a = np.array([[1, 2], [3, 4]], dtype=np.float64)
# Compute the inverse matrix
linalg.inv(a)
# Compute singular value decomposition
linalg.svd(a)
# Compute eigenvalues
linalg.eigvals(a)
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 8 / 9
2D plotting using Matplotlib
import numpy as np
import matplotlib.pyplot as plt
x = np.linspace(0.0, 2.0, 20)
plt.plot(x, np.sqrt(x), ’ro’) # red circles
plt.show()
plt.plot(x, np.sqrt(x), ’b-’) # blue lines
plt.show()
# Three plots in one figure
plt.plot(x, x, ’g--’, x, np.sqrt(x), ’ro’, x, np.sqrt(x), ’b-’)
plt.show()
Stephen Weston, Robert Bjornson (Yale) Introduction to Python Data Analysis July 2016 9 / 9
