堆叠和整形DataFrame( pandas )的切片而无需循环 [英] Stacking and shaping slices of DataFrame (pandas) without looping

问题描述

我具有以下形式的DataFrame:

I have a DataFrame of the following form:

     var1           var2        var3        day
0   -0.001284819    0.00138089  1.022781    1
1   -0.001310201    0.001377473 1.022626    1
2   -0.001330947    0.001374873 1.022477    2
3   -0.0013596      0.001430423 1.022385    2
4   -0.001361913    0.00144389  1.02228     3
5   -0.001371761    0.001440607 1.022161    3
6   -0.001393955    0.00143729  1.022017    4
7   -0.001431099    0.00143434  1.021908    4
8   -0.001466792    0.00143334  1.021749    5
9   -0.001491       0.00143159  1.021602    5

每个变量var1到var3是一个时间序列，每天记录两个样本.我正在尝试将此DataFrame转换为如下形式:

Each variable var1 to var3 is a time series and two samples are recorded every day. I'm trying to transform this DataFrame into something like this:

     day   1             2             3             4             5
0   var1  -0.001284819  -0.001330947  -0.001361913  -0.001393955  -0.001466792
1         -0.001310201  -0.0013596    -0.001371761  -0.001431099  -0.001491
2   var2   0.00138089    0.001374873   0.00144389    0.00143729    0.00143334
3          0.001377473   0.001430423   0.001440607   0.00143434    0.00143159
4   var3   1.022781      1.022477      1.02228       1.022017      1.021749
5          1.022626      1.022385      1.022161      1.021908      1.021602

现在每列都是一天，并且在每一天之内，每个变量仍然与原始时间序列中的顺序相同.

Now each column is a day, and within each day, each variable is still in the same order as in the original time series.

我一直在尝试找到一种方法来使用这种透视，堆叠和此类操作来避免循环，但是到目前为止我还无法管理.

I've been trying to figure out a way of achieving this result using pivoting, stacking and operations of this sort to avoid looping but so far I haven't be able to manage.

对于如何解决此问题的任何建议/提示/想法，将不胜感激:)

Any suggestions/hints/ideas on how to solve this would be very much appreciated :)

推荐答案

不一定是最漂亮的，但过去我做过类似的事情

It's not necessarily the prettiest, but in the past I've done things like

df = pd.read_csv("vd.csv", sep="\s+")
d2 = pd.melt(df, id_vars="day")
d2["sample"] = d2.groupby(["variable", "day"])["day"].rank("first")
d3 = d2.pivot_table(index=["variable", "sample"], columns="day")

给出

>>> d3
                    value                                        
day                     1         2         3         4         5
variable sample                                                  
var1     1      -0.001285 -0.001331 -0.001362 -0.001394 -0.001467
         2      -0.001310 -0.001360 -0.001372 -0.001431 -0.001491
var2     1       0.001381  0.001375  0.001444  0.001437  0.001433
         2       0.001377  0.001430  0.001441  0.001434  0.001432
var3     1       1.022781  1.022477  1.022280  1.022017  1.021749
         2       1.022626  1.022385  1.022161  1.021908  1.021602

[6 rows x 5 columns]

(虽然说实话，我认为Andy的方法比较笨拙.我将其保留在这里，因为过去在更困难的情况下，melt-modify-pivot模式对我来说非常有用.)

(Although to be honest, I think Andy's way is slicker. I'll leave this here though because the melt-modify-pivot pattern has proved pretty useful for me in the past in harder cases.)

这篇关于堆叠和整形DataFrame( pandas )的切片而无需循环的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！