CODEWITHSUNDEEP
pythonreplacetimestamp
saltymisty73
saltymisty73

Reputation: 39

Replace Column values o Pandas timeseries

I want to replace a set of values from a series in pre-set timeseries dataframe e.g. 3 values from series a will replace 3 timestamp values. However, replace or .set_value command don't seems to be working.In this cases, timestamp is already in indexed form. Timestamp example series are below:

>  timestamp        Coloum_1
>  01/01/2010 00:00 21.17
>  01/01/2010 00:15 19.67
>  01/01/2010 00:30 17.95

   a = pd.Series([15.0,16.0,17.0])
    df.set_value("index=pd.DatetimeIndex(start='2010/01/1 00::00', periods=3, freq='900 sec')", 'Column_1', 'a')

Kinda new in python. Appreciate any support. Thanks.

Update

The output may look like the bottom one. Here the timestamp and Column1 data are in a separate excel file where the Python read it and extract for analysis. Timestamp are already in indexed form and only the column1 values needed to replace by a separate series value. However, it doesn't seems to replace the series values with a timestamp sequence.

In [42]: df4 = pd.DataFrame({'Timestamp': ['2010-01-01 00:00:00', '2010-01-01    00:15:00', '2010-01-01 00:30:00'], 'Column_1':[21.17, 19.17, 17.95]})
df4 = df4.set_index('Timestamp')
df4.head()

Out[42]:
Timestamp            Column_1
2010-01-01 00:00:00   21.17
2010-01-01 00:15:00   19.17
2010-01-01 00:30:00   17.95

In [43]:a = pd.Series([50.0,60.0,70.0])           
df4.replace("'2010-01-01 00:00:00', periods=3, freq='15 min'", 'Column_1',a.all) 
df4.head()

a = pd.Series([50.0,60.0,70.0]) 
df4.replace("'2010-01-01 00:00:00', periods=3, freq='15 min'", 'Column_1',a.all) 
df4.head() 

Out[43]:
Timestamp           Column_1
2010-01-01 00:00:00 21.17 
2010-01-01 00:15:00 19.17 
2010-01-01 00:30:00 17.95

But if I use only a single timestamp it is replacing the corresponding value.

In [50]:

df4.set_value('2010-01-01 00:00:00', 'Column_1', 50);
df4.head()


Out[50]:
Timestamp          Column_1
2010-01-01 00:00:00 50.00
2010-01-01 00:15:00 19.17
2010-01-01 00:30:00 17.95

Upvotes: 1

Views: 563

Answers (2)

saltymisty73
saltymisty73

Reputation: 39

Solved

In [68]:
df = pd.DataFrame({'Timestamp': ['2010-01-01 00:00:00', '2010-01-01 00:15:00', '2010-01-01 00:30:00', '2010-01-01 00:45:00'], 
                    'Column_1':[21.17, 19.17, 17.95, 25.0]})
df = df.set_index('Timestamp')
df.head()

Out[68]:
Timestamp           Column_1
2010-01-01 00:00:00  21.17
2010-01-01 00:15:00  19.17
2010-01-01 00:30:00  17.95
2010-01-01 00:45:00  25.00
4 rows × 1 columns
In [69]:

df.loc ['2010-01-01 00:00:00':'2010-01-01 00:15:00', 'Column_1'] = [50,60] 
df.head()

Out[69]:
Timestamp           Column_1
2010-01-01 00:00:00  50.00
2010-01-01 00:15:00  60.00
2010-01-01 00:30:00  17.95
2010-01-01 00:45:00  25.00

Upvotes: 2

Merlin
Merlin

Reputation: 25639

Try this: 

 a = pd.Series([15.0,16.0,17.0])
    #0    15.0
    #1    16.0
    #2    17.0
   # dtype: float64

   a =  pd.date_range('1/1/2010', periods=3, freq='900S')

  #DatetimeIndex(['2010-01-01 00:00:00', '2010-01-01 00:15:00',
  #             '2010-01-01 00:30:00'],
  #            dtype='datetime64[ns]', freq='900S')

print a 

   # DatetimeIndex(['2010-01-01 00:00:00', '2010-01-01 00:15:00',
    #               '2010-01-01 00:30:00'],
    #              dtype='datetime64[ns]', freq='900S')


df = pd.DataFrame(a, columns=['Column1'] ) 
# or df = pd.DataFrame(pd.date_range('1/1/2010', periods=3, freq='900S'), columns=['Column1'] ) 
  #               Column1
  #  0  2010-01-01 00:00:00
  #  1  2010-01-01 00:15:00
  #  2  2010-01-01 00:30:00 

s = pd.Series(range(3), name='Column1')
    #0    0
    #1    1
    #2    2
    #Name: Column1, dtype: int64

Edit:

df4['Column_1'] = [50.0,60.0,70.0]

Upvotes: 0

Related Questions