Question

我极力尝试重现经典的Pandas滚动联接示例，其中quotes数据与trade数据合并。

在此处查看https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.merge_asof.html

以下是data.table格式的数据：

trades <- data.table(time = c('2016-05-25 13:30:00.023',
                              '2016-05-25 13:30:00.038',
                              '2016-05-25 13:30:00.048',
                              '2016-05-25 13:30:00.048',
                              '2016-05-25 13:30:00.048'),
                     ticker = c('MSFT','MSFT','GOOG','GOOG','AAPL'),
                     price = c(51.95,51.95,720.77,720.92,98.0),
                     quantity = c(75,155,100,100,100))
> trades
                      time ticker  price quantity
1: 2016-05-25 13:30:00.023   MSFT  51.95       75
2: 2016-05-25 13:30:00.038   MSFT  51.95      155
3: 2016-05-25 13:30:00.048   GOOG 720.77      100
4: 2016-05-25 13:30:00.048   GOOG 720.92      100
5: 2016-05-25 13:30:00.048   AAPL  98.00      100

和引号

quotes <- data.table(time = c('2016-05-25 13:30:00.023',
                              '2016-05-25 13:30:00.023',
                              '2016-05-25 13:30:00.030',
                              '2016-05-25 13:30:00.041',
                              '2016-05-25 13:30:00.048',
                              '2016-05-25 13:30:00.049',
                              '2016-05-25 13:30:00.072',
                              '2016-05-25 13:30:00.075'),
                     ticker = c('GOOG','MSFT','MSFT','MSFT','GOOG','AAPL','GOOG','MSFT'),
                     bid = c(720.50, 51.95, 51.97, 51.99, 720.5,97.99,720.5,52.01),
                     ask = c(270.93,51.96,51.98,52.00,720.93,98.01,720.88,52.03))
> quotes
                      time ticker    bid    ask
1: 2016-05-25 13:30:00.023   GOOG 720.50 270.93
2: 2016-05-25 13:30:00.023   MSFT  51.95  51.96
3: 2016-05-25 13:30:00.030   MSFT  51.97  51.98
4: 2016-05-25 13:30:00.041   MSFT  51.99  52.00
5: 2016-05-25 13:30:00.048   GOOG 720.50 720.93
6: 2016-05-25 13:30:00.049   AAPL  97.99  98.01
7: 2016-05-25 13:30:00.072   GOOG 720.50 720.88
8: 2016-05-25 13:30:00.075   MSFT  52.01  52.03

我想做的是按照以下方式合并交易数据和报价数据

对于每笔交易，请尽可能匹配最接近的上一个报价
匹配的报价必须在10毫秒内
完全匹配应该不发生。

输出（与Pandas教程中的输出相同）应该

                      time ticker  price quantity   bid   ask
1: 2016-05-25 13:30:00.023   MSFT  51.95       75    NA    NA
2: 2016-05-25 13:30:00.038   MSFT  51.95      155 51.97 51.98
3: 2016-05-25 13:30:00.048   GOOG 720.77      100    NA    NA
4: 2016-05-25 13:30:00.048   GOOG 720.92      100    NA    NA
5: 2016-05-25 13:30:00.048   AAPL  98.00      100    NA    NA

实际上，您可以看到唯一可能的报价匹配是针对2016-05-25 13:30:00.038的第二笔交易，因为封闭的（先前的）报价发生在2016-05-25 13:30:00.030处，所以这是在10毫秒之内（并非完全相同）匹配）。

尽管进行了试验，但我无法在data.table中重现此内容。有任何想法吗？谢谢！

Answer 1

您还可以将this idiom与滚动连接结合使用，这与@sindri_baldur提出的建议相似但不完全相同：

library(lubridate)
library(data.table)

quotes[, time := as.POSIXct(time, format="%Y-%m-%d %H:%M:%OS", tz = "GMT")]
trades[, time := as.POSIXct(time, format="%Y-%m-%d %H:%M:%OS", tz = "GMT")]

match_inexact <- function(q_time, t_time, bid, ask) {
  exact <- q_time == t_time # exact matches get NA
  bid[exact] <- NA_real_
  ask[exact] <- NA_real_
  list(bid, ask)
}

trades[, c("bid", "ask") := quotes[.SD,
                                   match_inexact(x.time, i.time, x.bid, x.ask),
                                   on = .(ticker, time),
                                   roll = lubridate::dmilliseconds(10L)]]

要注意的重要事项： time是为连接指定的最后一列，因为这是data.table将尝试滚动值的列。

Answer 2

这是一件完成工作的东西（快速又肮脏）：

# Format as POSIXct*
quotes[, time := as.POSIXct(time, format="%Y-%m-%d %H:%M:%OS", tz = "GMT")]
trades[, time := as.POSIXct(time, format="%Y-%m-%d %H:%M:%OS", tz = "GMT")]

# Match the nearest time (in the right direction) for each ticker and add as column
trades[quotes, on = .(time > time, ticker), qtime := i.time]
# Remove if not within time limit (10 millsecs)
trades[(time - qtime) > 0.01, qtime := NA_real_]
# Now perform an equi-join after removing timestamp that was too distant
trades[, c("bid", "ask") := quotes[trades, on = .(time = qtime), .(bid, ask)]]
trades[, !"qtime"] # drop this temporary column

#                   time ticker  price quantity   bid   ask
# 1: 2016-05-25 13:30:00   MSFT  51.95       75    NA    NA
# 2: 2016-05-25 13:30:00   MSFT  51.95      155 51.97 51.98
# 3: 2016-05-25 13:30:00   GOOG 720.77      100    NA    NA
# 4: 2016-05-25 13:30:00   GOOG 720.92      100    NA    NA
# 5: 2016-05-25 13:30:00   AAPL  98.00      100    NA    NA

*构建了POSIXct向量在双矢量的顶部，其中的值表示自1970-01-01起的秒数

从亚历克西斯（Alexis）的帖子中学习到的是使用roll参数的更简洁的版本。

trades[, c("qtime", "bid", "ask") := quotes[.SD, roll = 0.01, on = .(ticker, time), .(x.time, bid, ask)]]
trades[time == qtime, c("bid", "ask") := NA_real_][, qtime := NULL]

Answer 3

另一种可能的非等额联接方法，即使用该10ms窗口内的最新报价：

options(digits.secs=3) #see https://stackoverflow.com/a/43475068/1989480
library(data.table)

quotes[, time := as.POSIXct(time, format="%Y-%m-%d %H:%M:%OS", tz = "GMT")]
trades[, time := as.POSIXct(time, format="%Y-%m-%d %H:%M:%OS", tz = "GMT")][,
    c("start", "end") := .(time-0.01, time)]

trades[, c("bid", "ask") :=
        quotes[trades, on=.(ticker, time>=start, time<end), mult="last", .(bid, ask)]
    ][, c("start", "end") := NULL]

输出：

                      time ticker  price quantity   bid   ask
1: 2016-05-25 13:30:00.023   MSFT  51.95       75    NA    NA
2: 2016-05-25 13:30:00.038   MSFT  51.95      155 51.97 51.98
3: 2016-05-25 13:30:00.048   GOOG 720.77      100    NA    NA
4: 2016-05-25 13:30:00.048   GOOG 720.92      100    NA    NA
5: 2016-05-25 13:30:00.048   AAPL  98.00      100    NA    NA

如何将报价和交易数据与data.table（滚动联接）进行匹配

3 个答案: