Getting NaN converting pandas Dataframe to Series-CodePudding

I am trying to convert pandas Dataframe to Series based on accepted answer to Convert dataframe to series for multiple column

However I am getting NaN in my integer column 'y'.

Here is my code:

data = [['2021-10-14 18:12:00.000', '22811316'],['2021-10-14 18:42:00.000', '22700704']]
df = pd.DataFrame(data, columns = ['ds', 'y'])
series = pd.Series(df.y, index=df.ds)

printing series gives me:

ds
2021-10-14 18:12:00.000    NaN
2021-10-14 18:42:00.000    NaN
Name: y, dtype: object

What am I missing?

CodePudding user response：

I could find the answer in pandas.Series() Creation using DataFrame Columns returns NaN Data entries

The trick was to use:

series = pd.Series(df.y.values, index=df.ds)

CodePudding user response：

If you just take the series df.y, you will obtain a series with new indices starting from 0, 1, ...

print(df.y)

0    22811316
1    22700704
Name: y, dtype: object

These indices do not match with the values of the column ds that you want to use as index.
So, when you create the new series with index=..., you will probably have all NaN.

In order to put just the values of y column into the new series, you have to take only its values using to_numpy()

series = pd.Series(df.y.to_numpy(), index=df.ds)
print(series)

ds
2021-10-14 18:12:00.000    22811316
2021-10-14 18:42:00.000    22700704
dtype: object