JIYIK CN >

We will introduce to_numpy()the method to pandas.Dataframeconvert a to NumPyan array, which is introduced in pandas v0.24.0, replacing the old .valuesmethod. We can define it on Index, Series, and DataFrameobjects to_numpy.

The old DataFrame.valueshas inconsistent behavior and we do not recommend using it according to the Pandas API documentation. However, if you are using an older version, we will explore examples of this method.

Another old method that is deprecated is DataFrame.as_matrix(), please don't use it!

We will also introduce another DataFrame.to_records()way to convert a given DataFrameinto NumPyan array of records using the method.

to_numpyMethod DataFrameconverts to NumPyan array

pandas.Dataframeis a two-dimensional tabular data structure with rows and columns. to_numpyThis data structure can be converted to NumPyan array using the method:

# python 3.x
import pandas as pd
import numpy as np

df = pd.DataFrame(data=np.random.randint(0, 10, (6, 4)), columns=["a", "b", "c", "d"])
nmp = df.to_numpy()
print(nmp)
print(type(nmp))

Output:

[[5 5 1 3]
 [1 6 6 0]
 [9 1 2 0]
 [9 3 5 3]
 [7 9 4 9]
 [8 1 8 9]]
<class 'numpy.ndarray'>

Dataframe.valuesThis can be achieved using the method in the following way :

# python 3.x
import pandas as pd
import numpy as np

df = pd.DataFrame(data=np.random.randint(0, 10, (6, 4)), columns=["a", "b", "c", "d"])
nmp = df.values
print(nmp)
print(type(nmp))

Output:

[[8 8 5 0]
 [1 7 7 5]
 [0 2 4 2]
 [6 8 0 7]
 [6 4 5 1]
 [1 8 4 7]]
<class 'numpy.ndarray'>

If we want to NumPyinclude in the array indexes, we need to Dataframe.valuesapply reset_index():

# python 3.x
import pandas as pd
import numpy as np

df = pd.DataFrame(data=np.random.randint(0, 10, (6, 4)), columns=["a", "b", "c", "d"])
nmp = df.reset_index().values
print(nmp)
print(type(nmp))

Output:

[[0 1 0 3 7]
 [1 8 2 5 1]
 [2 2 2 7 3]
 [3 3 4 3 7]
 [4 5 4 4 3]
 [5 2 9 7 6]]
<class 'numpy.ndarray'>

to_records()Method DataFrameconverts to NumPyan array of records

If you need this dtypes, then to_records()is the best choice. In terms of performance, to_numpy()it to_records()is almost the same as:

# python 3.x
import pandas as pd
import numpy as np

df = pd.DataFrame(data=np.random.randint(0, 10, (6, 4)), columns=["a", "b", "c", "d"])
nmp = df.to_records()
print(nmp)
print(type(nmp))

Output:

[(0, 0, 4, 6, 1) 
 (1, 3, 1, 7, 1) 
 (2, 9, 1, 6, 4) 
 (3, 1, 4, 6, 9)
 (4, 9, 1, 3, 9)
 (5, 2, 5, 7, 9)]
<class 'numpy.recarray'>