python pandas : How to get the index of the values from a series that have matching values in other series?

问题

I have these two series:

In [48]: serie1
Out[48]: 
0    A
1    B
2    C
3    A
4    D

In [49]: serie2
Out[49]: 
0    X
1    Y
2    A
3    Z
4    A
5    D
dtype: object

And for each value in serie1 I want to get the index/indexes from serie2. Is this possible without iterating over values? A possible solution would be to build a dataframe more or less like this:

       A      B      C      D
X    False  False  False  False                 
Y    False  False  False  False
A    True   False  False  False
Z    False  False  False  False
A    True   False  False  False
D    False  False  False  True

... and then get the index of the "True" values for each column

回答1:

I think serie2.index[(array(serie2)=='A').flatten()] may work. 'A' is the value you want to find index for.

Or this, which may be less readable: serie2.index[(serie2=='A')[0]]

回答2:

1) For the boolean table of matches: if you want a crosstabulation (only shows unique values, no repetitions), and then convert it to boolean:

serie1 = pd.Series(['A','B','C','A','D'])
serie2 = pd.Series(['X','Y','A','Z','A','D'])

pd.crosstab(serie2,serie1) > 0 

col_0      A      B      C      D
row_0                            
A      False  False   True   True
X       True  False  False  False
Y      False   True  False  False
Z       True  False  False  False

(Note that the row-index is automatically sorted by value, so not the order in which the values appear in serie1. You may be able to override that by playing with .reorder_levels(...))

2) As for the indices of matches, to get them as a dict of arrays...

serie2.groupby(serie1).indices

{'A': array([0, 3]), 'C': array([2]), 'B': array([1]), 'D': array([4])}

# ... or as a list of arrays...
serie2.groupby(serie1).indices.values()

[array([0, 3]), array([2]), array([1]), array([4])]

# Here are alternatives with list comprehensions which are probably less efficient than `Series.groupby()` 
>>> [ np.flatnonzero(serie2.apply(lambda i2: i2==i1)) for i1 in serie1 ]
[array([2, 4]), array([], dtype=int64), array([], dtype=int64), array([2, 4]), array([5])]

>>> [ np.flatnonzero(serie2.apply(lambda i2: i2==i1)).tolist() for i1 in serie1 ]
[[2, 4], [], [], [2, 4], [5]]

来源：https://stackoverflow.com/questions/21665162/python-pandas-how-to-get-the-index-of-the-values-from-a-series-that-have-match

标签

python

pandas

series