Pandas 数据框:使用线性插值重新采样

2022-01-11 00:00:00 python pandas datetime time-series

问题描述

我正在尝试使用一种相当基本的重采样方法来处理 pandas 数据框.我的数据框 df 由日期时间条目索引并包含价格

I am trying to get a fairly basic resampling method to work with a pandas data frame. My data frame df is indexed by datetime entries and contains prices

price datetime 2000-08-16 09:29:55.755000 7.302786 2000-08-16 09:30:10.642000 7.304059 2000-08-16 09:30:26.598000 7.304435 2000-08-16 09:30:41.372000 7.304314 2000-08-16 09:30:56.718000 7.304334

我想将其缩减为 5 分钟.使用

I would like to downsample this to 5min. Using

df.resample(rule='5Min',how='last',closed='left')

在我的数据中以 5 分钟的倍数取最左边的点；类似

takes the closest point to the left in my data of a multiple of 5min; similarly

df.resample(rule='5Min',how='first',closed='left')

将关闭点移到右侧.但是，我想采用左右点之间的线性插值，例如如果我的 df 包含两个连续的条目

takes the closes point to the right. However, I would like to take the linear interpolation between the point to the left and right instead, e.g. if my df contains the two consecutive entries

time t1, price p1 time t2, price p2

和

t1<t<t2 where t is a multiple of 5min

那么重新采样的数据框应该有条目

then the resampled dataframe should have the entry

time t, price p1+(t-t1)/(t2-t1)*(p2-p1)

解决方案

尝试创建两个单独的数据框，reset_index 它们(因此它们具有相同的数字索引)，fillna在他们身上，然后对 df1 和 df2 进行数学运算.例如:

try creating two separate dataframes, reset_index them (so they have the same numerical index), fillna on them, and then just do the math on df1 and df2. e.g:

df1 = df.resample(rule='5Min',how='last',closed='left').reset_index().fillna(method='ffill') df2 = df.resample(rule='5Min',how='first',closed='left').reset_index().fillna(method='ffill') dt = df1.datetime - df2.datetime px_fld = df1.price + ...

类似的东西应该可以解决问题.

something like that should do the trick.

相关文章