如何从时间序列重采样中获取列内类别的计数

本文介绍了如何从时间序列重采样中获取列内类别的计数的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！ 问题描述

我是数据帧的新手，正在努力寻找如何实现以下目标的方法:

I'm new to data frames and am struggling to figure out how to accomplish the following:

我已经有一个像这样的时间序列的数据框:

I have a dataframe already as a time series like so:

timestamp source 2017-06-18 10:43:54 two 2017-06-20 03:38:23 three 2017-06-18 07:37:02 one 2017-06-07 16:49:51 two 2017-06-15 22:36:10 two 2017-06-07 16:49:51 two 2017-06-18 22:36:10 two

我正在尝试1)每天重新采样，2)获得当天每种类别的百分比.像这样:

I am trying to 1) resample into daily and 2) get a % of each category for that day. Like so:

timestamp One Two Three 2017-06-18 33% 66% 0% 2017-06-20 0% 0% 100% 2017-06-07 0% 100% 0% 2017-06-15 0% 100% 0%

我可以完成一些基本工作，例如，每天重新采样来源"的数量，但并没有将其细分为类别.

I can accomplish basic things like, get a count of 'source' resampled to daily, but it doesn't break it down into categories.

有人可以帮我指出正确的方向吗?非常感谢.

Can anyone help point me in the right direction? Greatly appreciated.

推荐答案

groupby + value_counts + unstack

groupby + value_counts + unstack

(df.groupby(df.timestamp.dt.date).source.value_counts(normalize=True)*100).unstack().fillna(0) source one three two timestamp 2017-06-07 0.000000 0.0 100.000000 2017-06-15 0.000000 0.0 100.000000 2017-06-18 33.333333 0.0 66.666667 2017-06-20 0.000000 100.0 0.000000

pivot_table

pivot_table

df2 = df.pivot_table(index=df.timestamp.dt.date, columns='source', aggfunc='size') df2 = df2.divide(df2.sum(1), axis=0).fillna(0)*100

pd.crosstab

pd.crosstab(df.timestamp.dt.date, df.source, normalize='index')*100