Suggestion on page 432

[baifanhorst]

On this page, the AUTO MPG dataset is loaded by the following codes:

url = 'http://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data'
column_names = ['MPG', 'Cylinders', 'Displacement', 'Horsepower', 'Weight',
'Acceleration', 'Model Year', 'Origin']

df = pd.read_csv(url, names=column_names,
na_values = "?", comment='\t',
sep=" ", skipinitialspace=True)

However, there is one more column in the data file, which contains car names. These names are written as "abc", with the quote marks explicit in the file. I tried to modify the above codes, including adding a new entry to the 'column_names' list, and adding additional options in 'pd.read_csv' such as quotechar = '"' and escapechar='', but still failed to load the car names. On the other hand, the following code, which is simpler, can load those car names:

df = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data', delim_whitespace=True, header=None, names=['mpg', 'cylinders', 'displacement', 'horsepower', 'weight', 'acceleration', 'model year', 'origin', 'car name'])

I suggest in the future this code may replace the original one.

Also, I found the option comment='\t' confusing in the original codes. I examined the data file and cannot find any comment line beginning with '\t'. However, without this option, pd.read_csv would fail to read the data. Later I realized that in front of each car name, there is a tab. So the function of "comment='\t'" is to ignore the car name. I am not sure whether my understanding is correct.

[baifanhorst]

I missed the option 'na_values='?' in the new code that I suggested. It should be:

df = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/auto-mpg/auto-mpg.data',
delim_whitespace=True,
header=None,
names=['mpg', 'cylinders', 'displacement', 'horsepower', 'weight', 'acceleration', 'model year', 'origin', 'car name'],
na_values='?')

Without the option, a following code that drops rows containing '?' won't work.