[pandas] Data type conversion (astype etc...)

import pandas as pd
df = pd.DataFrame({'date' : ['5/11/21', '5/12/21', '5/13/21', '5/14/21', '5/15/21'],
                   'sales' : ['10', '15', '20', '25', '30'], 'visitors' : ['10', '-', '17', '23', '25'],
                   'temp.' : ['24.1', '24.3', '24.8', '25', '25.4']})

# we need to check data type for each column (to change/edit data)
print(df.dtypes)

# try to change the value without data conversion
df['edited sales'] = df['sales'] + 1

In the last sentence, the program won't return its calculated result. It is because df['sales'] is in object type. To calculate, the data should be converted into numeric type (int, float, etc.)

df = df.astype({'sales':'int'})
# now we can calculate/edit/change the data
df['edited sales'] = df['sales'] + 1
print(df)

# However when there is missing value in the data, the program will not convert the data
# therefore, we need to coerce the program to convert the data.
df['visitors'] = pd.to_numeric(df['visitors'], errors='coerce')

 

As you can see the result, the missing value is changed from '-' to 'NaN'

And then we need one more step to edit/calculate.

# convert it to 0 / other values first
df.fillna(0, inplace=True)
# then convert that data type of the column to int type
df = df.astype({'visitors':'int'})
print(df)

df['visitors+1'] = df['visitors'] + 1
print(df)

As you can see, after we finished to change the datatype, users can change the data.

 

extra) change the format of data into datetime format

df['date'] = pd.to_datetime(df['date'], format="%m/%d/%y")
print(df)

Then it would be converted into a datetime format.