I have multiple pandas dataframes and I would like to write a function that will take out the values in each column of the dataframes and put them into their own numpy array.
Example dataframe
In [1]: df = pd.DataFrame([[1, 2], [1, 3], [4, 6]], columns=['A', 'B'])
In [2]: df
Out[2]:
A B
0 1 2
1 1 3
2 4 6
How would I generate two different numpy arrays out of the values in column A and B
df['A'].values will return numpy arrays. see below:
for col in df.columns:
print(col, type(df[col].values))
A <class 'numpy.ndarray'>
B <class 'numpy.ndarray'>
You can use this method to obtain a list of numpy arrays for each column. You can put the arrays in a dictionary too but I prefer a list over the former because it is ordered and we can always use a zip function with df.columns to get tupled data!
np_arr_list = []
for i in df.columns:
new_np_arr = np.array(df.loc[:,i])
np_arr_list.append(new_np_arr)
Output:
[array([ 1, 21, 213, 32], dtype=int64),
array([ 4, 5, 32, 3], dtype=int64),
array([213, 23, 23, 1], dtype=int64)]
for (col_name,np_arr) in zip(df.columns, np_arr_list): ` print("Column Name :", col_name,"\nArray",np_arr,"\n")` It binds col_name to iterate through df.columns and np_arr through np_arr_list. I use it for association of column names and values in some use cases where I want to keep data and names separate. Other way is to use a dictionary with keys as column name and values as numpy arrays.You should use to_numpy() methods because others will be deprecated. If your pandas library is older than 0.24 you should update it first because this method has come after from that version.
>>> df = pd.DataFrame([[1, 2], [1, 3], [4, 6]], columns=['A', 'B'])
>>> arr = []
>>> for column in df.columns:
... arr.append(df[column].to_numpy())
...
>>> arr
[array([1, 1, 4], dtype=int64), array([2, 3, 6], dtype=int64)]
>>>
df.values;