在不同活動的序列中計數回合長度

Question

我有一個數據集，我需要計算非常大的數據集（65400行！）的回合長度。當我們收集數據時，我們記錄了我們的動物在5分鐘時間內每分鐘和每秒所做的動作。我的示例數據如下：

structure(list(date = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L), .Label = "02/04/2015", class = "factor"), minute = c(1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), sec = 1:11, activity = structure(c(2L, 
2L, 2L, 1L, 3L, 3L, 3L, 2L, 2L, 2L, 2L), .Label = c("N", "S", 
"U"), class = "factor"), day_time = structure(c(1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = "14:45", class = "factor")), .Names = c("date", 
"minute", "sec", "activity", "day_time"), class = "data.frame", row.names = c(NA, 
-11L)) 

df 
     date minute sec activity day_time 
1 02/04/2015  1 1  S 14:45 
2 02/04/2015  1 2  S 14:45 
3 02/04/2015  1 3  S 14:45 
4 02/04/2015  1 4  N 14:45 
5 02/04/2015  1 5  U 14:45 
6 02/04/2015  1 6  U 14:45 
7 02/04/2015  1 7  U 14:45 
8 02/04/2015  1 8  S 14:45 
9 02/04/2015  1 9  S 14:45 
10 02/04/2015  1 10  S 14:45 
11 02/04/2015  1 11  S 14:45 

我們需要計算的是每次回合的長度（以秒爲單位），以保持5分鐘的會話分離。在這種情況下，我們所期望的輸出將是

structure(list(date = structure(c(1L, 1L, 1L, 1L), .Label = "02/04/2015", class = "factor"), 
     minute = c(1L, 1L, 1L, 1L), activity = structure(c(2L, 1L, 
     3L, 2L), .Label = c("N", "S", "U"), class = "factor"), day_time = structure(c(1L, 
     1L, 1L, 1L), .Label = "14:45", class = "factor"), bout_length = c(3L, 
     1L, 3L, 4L)), .Names = c("date", "minute", "activity", "day_time", 
    "bout_length"), class = "data.frame", row.names = c(NA, -4L)) 

desired output 

      date minute activity day_time bout_length 
    1 02/04/2015  1  S 14:45   3 
    2 02/04/2015  1  N 14:45   1 
    3 02/04/2015  1  U 14:45   3 
    4 02/04/2015  1  S 14:45   4 

我已經試過（RLE）沒有成功，因爲我必須指定不同的分會議。謝謝你的幫助

Answer 1

我終於找到了方向。這將是很好，如果有一個短途的方式

bouts1 <- as.data.frame(lapply(df, as.character), stringsAsFactors = FALSE) 
bouts1<-head(do.call(rbind, by(bouts1, df$session, rbind, "empty")), -1) 
rownames(bouts1) <- seq(length=nrow(bouts1))# this renames row names in my dataframe 
diffs <- bouts1$activity[-1L] != bouts1$activity[-length(bouts1$activity)]#tells us where the activity is diff 
idx <- c(which(diffs), length(bouts1$activity)) 
bout.len<-diff(c(0, idx)) 
trial <- c(which(diffs=="TRUE"))#these are the row IDs that contain activities that change 
new.bouts<-bouts1[row.names(bouts1) %in% trial,]# extracting the rows in the data where activity changes 
temp<-bouts[65400,]#took the last observation from the main data set. 
new.bouts <- rbind(new.bouts,temp)#these 2 commands I had to create another line so that we get where to put e value 
new.bouts$bout.len<-bout.len 
new.bouts<-new.bouts[!new.bouts$Date=="empty",]