开满天机
df.A.values转换df.A为 numpy 数组。为了回答这个问题，让我们创建一个测试数据框：>>> from pandas import util>>> df= util.testing.makeDataFrame()>>> df.head() A B C DBMdjymcTHC -0.684721 1.622097 -2.525634 1.6272900e7Mekvkf7 0.003399 0.152074 -0.095163 -0.276664q0E6te3rF9 1.639105 -1.935913 1.733587 -0.729493w7d1NGfq1p -0.496669 -1.182373 -0.950125 2.201667RPqDHEGhxs -1.169309 0.608857 -0.748978 0.270510您的代码给出以下输出：>>> df.A.values[:, None]array([[-0.68472066], [ 0.00339929], [ 1.63910531], [-0.49666918], [-1.16930896], [ 0.18225299], [ 0.88957142], [ 0.97299314], [ 0.67984743], [ 1.11192848], [-1.43273161], [-0.59633832], [ 0.81591342], [ 1.26188783], [ 0.08789735], [-0.37412069], [ 0.15285941], [-0.14208735], [ 0.37897237], [ 0.49208469], [ 0.86949863], [-0.98972967], [ 0.66001405], [-1.69139314], [ 1.18512158], [ 1.47981638], [ 1.21812138], [ 0.82375357], [-0.4896989 ], [ 0.53701562]])让我们检查一下shape：>>> df.A.values[:, None].shape(30, 1)如果你没有None：>>> df.A.values[:]array([-0.68472066, 0.00339929, 1.63910531, -0.49666918, -1.16930896, 0.18225299, 0.88957142, 0.97299314, 0.67984743, 1.11192848, -1.43273161, -0.59633832, 0.81591342, 1.26188783, 0.08789735, -0.37412069, 0.15285941, -0.14208735, 0.37897237, 0.49208469, 0.86949863, -0.98972967, 0.66001405, -1.69139314, 1.18512158, 1.47981638, 1.21812138, 0.82375357, -0.4896989 , 0.53701562])是shape：>>> df.A.values[:].shape(30,)因此，它本质上是向数组添加一个维度/轴numpy，并在我的例子中创建一个维度数组(30x1)。您的代码相当于：>>> df.A.values.reshape(-1,1)>>> df.A.values.reshape(-1,1).shape(30, 1)