方法#1
这是一个被this post
启发一个量化的方法 -
def group_duplicate_index(df):
a = df.values
sidx = np.lexsort(a.T)
b = a[sidx]
m = np.concatenate(([False], (b[1:] == b[:-1]).all(1), [False]))
idx = np.flatnonzero(m[1:] != m[:-1])
I = df.index[sidx].tolist()
return [I[i:j] for i,j in zip(idx[::2],idx[1::2]+1)]
采样运行 -
In [42]: df
Out[42]:
param_a param_b param_c
1 0 0 0
2 0 2 1
3 2 1 1
4 0 2 1
5 2 1 1
6 0 0 0
In [43]: group_duplicate_index(df)
Out[43]: [[1, 6], [3, 5], [2, 4]]
方法2
对于整数编号dataframes,我们可以每行降低到一个标量每能让我们有一个1D
阵列工作,给我们一个更好的性能之一,像这样 -
def group_duplicate_index_v2(df):
a = df.values
s = (a.max()+1)**np.arange(df.shape[1])
sidx = a.dot(s).argsort()
b = a[sidx]
m = np.concatenate(([False], (b[1:] == b[:-1]).all(1), [False]))
idx = np.flatnonzero(m[1:] != m[:-1])
I = df.index[sidx].tolist()
return [I[i:j] for i,j in zip(idx[::2],idx[1::2]+1)]
运行测试
其他方法(ES) -
def groupby_app(df): # @jezrael's soln
df = df[df.duplicated(keep=False)]
df = df.groupby(df.columns.tolist()).apply(lambda x: tuple(x.index)).tolist()
return df
计时 -
In [274]: df = pd.DataFrame(np.random.randint(0,10,(100000,3)))
In [275]: %timeit group_duplicate_index(df)
10 loops, best of 3: 36.1 ms per loop
In [276]: %timeit group_duplicate_index_v2(df)
100 loops, best of 3: 15 ms per loop
In [277]: %timeit groupby_app(df) # @jezrael's soln
10 loops, best of 3: 25.9 ms per loop