python - Pandas 系列到二维数组

Question

因此，我使用了 Put a 2d Array into a Pandas Series 中的答案将 2D numpy 数组放入 pandas 系列。简而言之，就是

a = np.zeros((5,2))
s = pd.Series(list(a))

现在，将 pandas 系列转换回二维数组的最便宜的方法是什么？如果我尝试 s.values，我会得到具有 object dtype 的数组数组。

到目前为止，我尝试了 np.vstack(s.values)，但它当然会复制数据。

Answer 1

我相信你需要:

a = np.array(s.values.tolist())
print (a)
[[ 0.  0.]
 [ 0.  0.]
 [ 0.  0.]
 [ 0.  0.]
 [ 0.  0.]]

---

a = np.zeros((50000,2))
s = pd.Series(list(a))

In [131]: %timeit (np.vstack(s.values))
10 loops, best of 3: 107 ms per loop

In [132]: %timeit (np.array(s.values.tolist()))
10 loops, best of 3: 19.7 ms per loop

In [133]: %timeit (np.array(s.tolist()))
100 loops, best of 3: 19.6 ms per loop

但是如果转置差异很小(但是 caching ):

a = np.zeros((2,50000))
s = pd.Series(list(a))
#print (s)

In [159]: %timeit (np.vstack(s.values))
The slowest run took 23.31 times longer than the fastest. This could mean that an intermediate result is being cached.
10000 loops, best of 3: 55.7 µs per loop

In [160]: %timeit (np.array(s.values.tolist()))
The slowest run took 7.20 times longer than the fastest. This could mean that an intermediate result is being cached.
10000 loops, best of 3: 49.8 µs per loop

In [161]: %timeit (np.array(s.tolist()))
The slowest run took 7.31 times longer than the fastest. This could mean that an intermediate result is being cached.
10000 loops, best of 3: 62.6 µs per loop