1

just started learning python. trying to change a columns data type from object to float to take out the mean. I have tried to change [] to () and even the "". I dont know whether it makes a difference or not. Please help me figure out what the issue is. thanks!!

My code:

df["normalized-losses"]=df["normalized-losses"].astype(float)

error which i see: attached as imageenter image description here

Aarushi Goyal
  • 55
  • 1
  • 5
  • Please format the code and the Traceback. Select the code and type `ctrl-k`. [Formatting posts](https://stackoverflow.com/help/formatting) ... [Formatting help](https://stackoverflow.com/editing-help) – wwii Aug 19 '18 at 13:31
  • I've edited my answer based on your screenshot. – Josh Friedlander Aug 19 '18 at 14:16

4 Answers4

3

Use:

df['normalized-losses'] = df['normalized-losses'][~(df['normalized-losses'] == '?' )].astype(float)

Using df.normalized-losses leads to interpreter evaluating df.normalized which doesn't exist. The statement you have written executes (df.normalized) - (losses.astype(float)).There appears to be a question mark in your data which can't be converted to float.The above statement converts to float only those rows which don't contain a question mark and drops the rest.If you don't want to drop the columns you can replace them with 0 using:

df['normalized-losses'] = df['normalized-losses'].replace('?', 0.0)
df['normalized-losses'] = df['normalized-losses'].astype(float)
Rik
  • 467
  • 4
  • 14
  • I am getting this error. what am I doing wrong here? – Aarushi Goyal Aug 19 '18 at 13:42
  • ValueError Traceback (most recent call last) in () ----> 1 df["normalized-losses"]=df["normalized-losses"].astype(float) ~\Anaconda3\lib\site-packages\pandas\util\_decorators.py in wrapper(*args, **kwargs) – Aarushi Goyal Aug 19 '18 at 13:43
  • ~\Anaconda3\lib\site-packages\pandas\util\_decorators.py in wrapper(*args, **kwargs) 175 else:176 kwargs[new_arg_name] = new_arg_value --> 177 return func(*args, **kwargs) 178 return wrapper 179 return _deprecate_kwarg ~\Anaconda3\lib\site-packages\pandas\core\generic.py in astype(self, dtype, copy, errors, **kwargs) 4995 # else, only a single dtype is given 4996 new_data = self._data.astype(dtype=dtype, copy=copy, errors=errors, -> 4997 **kwargs) 4998 return self._constructor(new_data).__finalize__(self) 4999 – Aarushi Goyal Aug 19 '18 at 13:53
  • Can you paste this error to the question by including the command you tried from this answer and the error message associated with that – Harikrishna Aug 19 '18 at 13:56
  • I have edited my post. Check if it still gives any errors – Rik Aug 19 '18 at 15:10
2

Welcome to Stack Overflow, and good luck on your Python journey! An important part of coding is learning how to interpret error messages. In this case, the traceback is quite helpful - it is telling you that you cannot call normalized after df, since a dataframe does not have a method of this name.

Of course you weren't trying to call something called normalized, but rather the normalized-losses column. The way to do this is as you already did once - df["normalized-losses"].

As to your main problem - if even one of your values can't be converted to a float, the columnwide operation will fail. This is very common. You need to first eliminate all of the non-numerical items in the column, one way to find them is with df[~df['normalized_losses'].str.isnumeric()].

Josh Friedlander
  • 10,870
  • 5
  • 35
  • 75
  • and more on that topic here: https://stackoverflow.com/questions/21771133/finding-non-numeric-rows-in-dataframe-in-pandas – Josh Friedlander Aug 19 '18 at 14:25
  • so in normalized-losses there are some entries which are missing and are represented by a ?. I need to fill those values by the average of normalized-losses. how can I fill in with averages if I cant find the average because of the missing values – Aarushi Goyal Aug 19 '18 at 14:49
  • 1
    @Aarushi you may wish to use `pd.to_numeric(df['column_name'], errors='coerce')` and then either `.dropna()` to get rid of invalid values or use `.fillna(0)` to introduce a value that'll count towards the number of items in a group but not it's sum... – Jon Clements Aug 19 '18 at 15:10
1

The "df.normalized-losses" does not signify anything to python in this case. you can replace it with df["normalized-losses"]. Usually, if you try 

df["normalized-losses"]=df["normalized-losses"].astype(float)

This should work. What this does is, it takes normalized-losses column from dataframe, converts it to float, and reassigns it to normalized column in the same dataframe. But sometimes it might need some data processing before you try the above statement.

Harikrishna
  • 1,100
  • 5
  • 13
  • 30
0

You can't use - in an attribute or variable name. Perhaps you mean normalized_losses?

Daniel Roseman
  • 588,541
  • 66
  • 880
  • 895