classifying a series to a new column in pandas

Question

I want to be able to take my current set of data, which is filled with ints, and classify them according to certain criteria. The table looks something like this: ``` [in]> df = pd.DataFrame({'A':[0,2,3,2,0,0],'B': [1,0,2,0,0,0],'C': [0,0,1,0,1,0]}) [out]> A B C 0 0 1 0 1 2 0 0 2 3 2 1 3 2 0 0 4 0 0 1 5 0 0 0 ``` I'd like to classify these in a separate column by string. Being more familiar with R, I tried to create a new column with the rules in that column's definition. Following that I attempted with .ix and lambdas which both resulted in a type errors (between ints & series ). I'm under the impression that this is a fairly simple question. Although the following is completely wrong, here is the logic from attempt 1: ``` df['D']=( if ((df['A'] > 0) & (df['B'] == 0) & df['C']==0): return "c1"; elif ((df['A'] == 0) & ((df['B'] > 0) | df['C'] >0)): return "c2"; else: return "c3";) ``` for a final result of: ``` A B C D 0 0 1 0 "c2" 1 2 0 0 "c1" 2 3 2 1 "c3" 3 2 0 0 "c1" 4 0 0 1 "c2" 5…

Accepted Answer

I can think of two ways. The first is to write a classifier function and then `.apply` it row-wise:

```
>>> import pandas as pd
>>> df = pd.DataFrame({'A':[0,2,3,2,0,0],'B': [1,0,2,0,0,0],'C': [0,0,1,0,1,0]})
>>> 
>>> def classifier(row):
...         if row["A"] > 0 and row["B"] == 0 and row["C"] == 0:
...                 return "c1"
...         elif row["A"] == 0 and (row["B"] > 0 or row["C"] > 0):
...                 return "c2"
...         else:
...                 return "c3"
...     
>>> df["D"] = df.apply(classifier, axis=1)
>>> df
   A  B  C   D
0  0  1  0  c2
1  2  0  0  c1
2  3  2  1  c3
3  2  0  0  c1
4  0  0  1  c2
5  0  0  0  c3
```

and the second is to use advanced indexing:

```
>>> df = pd.DataFrame({'A':[0,2,3,2,0,0],'B': [1,0,2,0,0,0],'C': [0,0,1,0,1,0]})
>>> df["D"] = "c3"
>>> df["D"][(df["A"] > 0) & (df["B"] == 0) & (df["C"] == 0)] = "c1"
>>> df["D"][(df["A"] == 0) & ((df["B"] > 0) | (df["C"] > 0))] = "c2"
>>> df
   A  B  C   D
0  0  1  0  c2
1  2  0  0  c1
2  3  2  1  c3
3  2  0  0  c1
4  0  0  1  c2
5  0  0  0  c3
```

Which one is clearer depends upon the situation. Usually the more complex the logic the more likely I am to wrap it up in a function I can then document and test.

classifying a series to a new column in pandas

Solution

Problem