<p>你可能会发现这也很有用。</p>
<pre><code>keys = range(horizon + 1)
pd.concat([df.shift(-i) for i in keys], axis=1, keys=keys)
0 1 2 3
PRICE PRICE PRICE PRICE
0 100 200.0 300.0 400.0
1 200 300.0 400.0 500.0
2 300 400.0 500.0 600.0
3 400 500.0 600.0 700.0
4 500 600.0 700.0 800.0
5 600 700.0 800.0 900.0
6 700 800.0 900.0 1000.0
7 800 900.0 1000.0 NaN
8 900 1000.0 NaN NaN
9 1000 NaN NaN NaN
</code></pre>
<p>如果将<code>concat</code>分配给<code>df_c</code></p>
<pre><code>keys = range(horizon + 1)
df_c = pd.concat([df.shift(-i) for i in keys], axis=1, keys=keys)
df_c.apply(lambda x: pd.Series([x[0].values, x[1:].values]), axis=1)
0 1
0 [100.0] [200.0, 300.0, 400.0]
1 [200.0] [300.0, 400.0, 500.0]
2 [300.0] [400.0, 500.0, 600.0]
3 [400.0] [500.0, 600.0, 700.0]
4 [500.0] [600.0, 700.0, 800.0]
5 [600.0] [700.0, 800.0, 900.0]
6 [700.0] [800.0, 900.0, 1000.0]
7 [800.0] [900.0, 1000.0, nan]
8 [900.0] [1000.0, nan, nan]
9 [1000.0] [nan, nan, nan]
</code></pre>