Convert float64 type DataFrame to float in Python

Question

I created a multi-indexed DataFrame wherein I used groupby with mean.

On checking data type of any particular value

(using ChkVlu = df.loc['Index1'].loc['index2'].loc['requiredcolumn'])

I get data type to be numpy.float64. I require the data type to native python float.

If I use ChkVlu = float(ChkVlu) then only one value at a time gets converted from float64 to float, however on using df = df.astype(float) for entire DataFrame I still get data type to be numpy.float64.

I used df = df.astype(float, copy=False) as well as df = df.astype(float, copy=True) but still getting float64 rather than float.

Please help me in converting float64 type DataFrame to float.

Edit1 : Posting the code here where AnnualData.csv has index 1 and index2 as 1st 2 columns and year-month combination as next set of columns in format of 2001-01, 2001-02, 2001-03 ....... 2016-09, 2016-10, 2016-11 with numerical data in each row. I am converting data in quarters with column names having q1, q2 and so on with mean value of each quarter.

 df = pd.read_csv(‘AnnualData.csv')
 df.set_index(['index1, 'index2'],inplace = True)
 def quarters(col):
  if col.endswith(("01", "02", "03")):
   final_col = col[:4] + "q1"
  elif col.endswith(("04", "05", "06")):
   final_col = col[:4] + "q2"
  elif col.endswith(("07", "08", "09")):
   final_col = col[:4] + "q3"
  else:
   final_col = col[:4] + "q4"
  return final_col  
df = df.groupby(quarters, axis = 1).mean()
ChkVlu = df.loc['index1'].loc['index2'].loc['requiredcolumn']
type(ChkVlu)

What is the reason for your preference on float over np.float64? — Blownhither Ma
– Blownhither Ma, Commented Oct 20, 2018 at 7:57
Hi pygo, I tried that too, but still data type remains float64. I tried pd.to_numeric(df) also but seems it works only on list or 1D array and not on 2D DataFrame — Kilo_Papa
– Kilo_Papa, Commented Oct 20, 2018 at 8:09
Can you update your post with the DataFrame you have to reproduce it to get an answer? — Karn Kumar
– Karn Kumar, Commented Oct 20, 2018 at 8:21

charlie6411 · Accepted Answer · 2018-10-20 08:47:44Z

1

You can use to_numeric which will give you a float32

or np.float32(x)

then x.item() will give you python float

answered Oct 20, 2018 at 8:47

charlie6411

1067 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

Karn Kumar Over a year ago

@ charlie6411, what about df1 = np.float(x) rather np.float32(x).

charlie6411 Over a year ago

I think its depreciated, or will be soon

Kilo_Papa Over a year ago

Should x be a dataframe or a single value, because the suggested method seems to work on single value while error stating 'can only convert an array of size 1 to a Python scalar' is received if attempted on dataframe

charlie6411 Over a year ago

to_numeric will work on series, remember dataframe is just a collection of series

Kilo_Papa Over a year ago

I tried df['requiredcolumn'] = pd.to_numeric(df['requiredcolumn']) for one column to check the same, but still output is numpy.float64

|

Collectives™ on Stack Overflow

Convert float64 type DataFrame to float in Python

1 Answer 1

7 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Related