将数据帧转换为R中的TS [英] Transforming a dataframe into a TS in R
问题描述
我一直在尝试将整合在一起的数据帧转换为时间序列，但是由于某种原因它无法正常工作。
I've been trying to transform a dataframe I put together into a Time Series, but for some reason it doesn't work. I am very new to R.
x<-Sales_AEMBG%>%
+ select(Ecriture.DatEcr, Crédit, Mapping)
> names(x)<-c("Dates","Revenue","Mapping")
> str(x)
Classes ‘tbl_df’, ‘tbl’ and 'data.frame': 15167 obs. of 3 variables:
$ Dates : POSIXct, format: "2016-01-02" "2016-01-02" "2016-01-02" "2016-01-02" ...
$ Revenue: num 124065 214631 135810 225293 57804 ...
$ Mapping: chr "E.M 1.5 L" "E.M 1.5 L" "E.M 1.5 L" "E.M 1.5 L" ...'
当我尝试查看数据时，这就是我所拥有的
When I try to look at the data, here's what I have
> head(x)
# A tibble: 6 x 3
Dates Revenue Mapping
<dttm> <dbl> <chr>
1 2016-01-02 00:00:00 124065. E.M 1.5 L
2 2016-01-02 00:00:00 214631. E.M 1.5 L
3 2016-01-02 00:00:00 135810. E.M 1.5 L
4 2016-01-02 00:00:00 225293. E.M 1.5 L
5 2016-01-02 00:00:00 57804. E.M 1.5 L
6 2016-01-02 00:00:00 124065. E.M 1.5 L
当然，我尝试了as.ts函数
Of course, I tried the as.ts function
> x_xts <- as.ts(x)
Warning message:
In data.matrix(data) : NAs introduced by coercion
> is.ts(x)
[1] FALSE
但它一直告诉我数据帧仍未被识别为TS。
But it keeps telling me that my dataframe is still not recognized as a TS.
您有什么建议？
谢谢
推荐答案
我在其中添加了一些观察结果
I've added a few more observations to your data.
# A tibble: 12 x 3
Dates Revenue Mapping
<dttm> <dbl> <chr>
1 2016-01-02 00:00:00 124065 E.M 1.5 L
2 2016-01-02 00:00:00 214631 E.M 1.5 L
3 2016-01-03 00:00:00 135810 E.M 1.5 L
4 2016-01-03 00:00:00 225293 E.M 1.5 L
5 2016-01-05 00:00:00 57804 E.M 1.5 L
6 2016-01-05 00:00:00 124065 E.M 1.5 L
7 2016-01-02 00:00:00 24065 E.M 1.5 M
8 2016-01-02 00:00:00 14631 E.M 1.5 M
9 2016-01-03 00:00:00 35810 E.M 1.5 M
10 2016-01-03 00:00:00 25293 E.M 1.5 M
11 2016-01-05 00:00:00 7804 E.M 1.5 M
12 2016-01-05 00:00:00 24065 E.M 1.5 M
首先，您需要按天（日期）和产品类型（您的映射变量？），并转换为更广泛的数据格式：
First you need to sum the sales by day (Dates) and product type (your Mapping variable?), and pivot into a wider data format:
library(dplyr)
library(tidyr)
x.sum <- x %>%
group_by(Mapping, Dates) %>%
summarise(Revenue=sum(Revenue)) %>%
pivot_wider(id_cols=Dates, names_from=Mapping, values_from=Revenue)
# A tibble: 3 x 3
Dates `E,M 1.5 L` `E,M 1.5 M`
<dttm> <dbl> <dbl>
1 2016-01-02 00:00:00 338696 38696
2 2016-01-03 00:00:00 361103 61103
3 2016-01-05 00:00:00 181869 31869
请注意，我已故意省略1月4日
Note that I've deliberately omitted Jan 4.
如果您的时间序列数据缺少日期，例如股票价格，周末金融市场休市，则使用 as.ts （或 ts ）功能不起作用。如果没有丢失的日期，则将数据转换为时间序列对象（ ts）的正确方法是指定要转换的列（ x.sum [，2：3 ] ）以及该系列的开始日期（2016年1月2日）和频率（每天）。
If your time series data has missing days, such as stock prices where financial markets are closed on the weekends, then using the as.ts (or ts) function won't work. If there are no missing days, then then correct way to convert the data into a time series object ("ts") is to specify the column(s) to convert (x.sum[,2:3]) and the start (January 2, 2016) and frequency (daily) of the series.
x.ts <- ts(x.sum[,2:3], start=c(2016, 2), frequency=365)
开始时要小心，因为第二个参数取决于指定的频率。在这里，365表示每日，因此 2表示2016年的第2天。如果频率为每月，则 2表示2016年的第2个月。
Be careful with the start as the second argument depends on the specified frequency. Here, 365 means daily, so the "2" means day 2 of year 2016. If the frequency was monthly, the "2" would mean month 2 of year 2016.
但是正如我提到的， ts 不会忽略任何丢失的日子。因此，对于这种化妆数据，如果绘制时间序列，则会得到错误的信息。
But as I mentioned, ts doesn't ignore any missing days. So for this make-up data, if you plotted the time series, then you will get the wrong information.
在这种情况下，其他软件包如 xts 和 zoo 可用于简化工作。
In this case, other packages such as xts and zoo can be used to simply the work.
library(xts)
x.xts <- xts(x.sum[,2:3], order.by=x.sum$Dates)
plot(x.xts) # Correct results.
Other answers about time series can be found here and here.
这篇关于将数据帧转换为R中的TS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！
