问题
I want to convert N columns into one series. How to do it effectively?
Input:
0 1 2 3
0 64 98 47 58
1 80 94 81 46
2 18 43 79 84
3 57 35 81 31
Expected Output:
0 64
1 80
2 18
3 57
4 98
5 94
6 43
7 35
8 47
9 81
10 79
11 81
12 58
13 46
14 84
15 31
dtype: int64
So Far I tried:
print df[0].append(df[1]).append(df[2]).append(df[3]).reset_index(drop=True)
I'm not satisfied with my solution, moreover it won't work for dynamic columns. Please help me to find a better approach.
回答1:
You can also use Series class and .values attribute:
pd.Series(df.values.T.flatten())
Output:
0 64
1 80
2 18
3 57
4 98
5 94
6 43
7 35
8 47
9 81
10 79
11 81
12 58
13 46
14 84
15 31
dtype: int64
回答2:
you need np.flatten
pd.Series(df.values.flatten(order='F'))
out[]
0 64
1 80
2 18
3 57
4 98
5 94
6 43
7 35
8 47
9 81
10 79
11 81
12 58
13 46
14 84
15 31
dtype: int64
回答3:
You can use unstack
pd.Series(df.unstack().values)
回答4:
Here's yet another short one.
>>> pd.Series(df.values.ravel(order='F'))
>>>
0 64
1 80
2 18
3 57
4 98
5 94
6 43
7 35
8 47
9 81
10 79
11 81
12 58
13 46
14 84
15 31
dtype: int64
回答5:
Use pd.melt() -
df.melt()['value']
Output
0 64
1 80
2 18
3 57
4 98
5 94
6 43
7 35
8 47
9 81
10 79
11 81
12 58
13 46
14 84
15 31
Name: value, dtype: int64
回答6:
df.T.stack().reset_index(drop=True)
Out:
0 64
1 80
2 18
3 57
4 98
5 94
6 43
7 35
8 47
9 81
10 79
11 81
12 58
13 46
14 84
15 31
dtype: int64
来源：https://stackoverflow.com/questions/53317743/how-to-convert-dataframe-into-series