category split

library(magrittr)
library(plyr)
library(dplyr)

set.seed(100)
numCate <- 5000
N <- 1e6
cates <- sample(paste0("A", 1:numCate), N, TRUE)
time <- sample(1:300, N, TRUE)
dat <- data.frame(cates, time, stringsAsFactors = FALSE) %>% tbl_df
st <- proc.time()
dat_out <- ddply(dat, .(cates), function(sub_df){
  sub_df %>% arrange(time) %>% filter(diff(c(0, time)) > 1)
}) %>% tbl_df
proc.time() - st
# user  system elapsed
# 6.13    0.00    6.21

library(foreach)
library(doSNOW)
cl <- makeCluster(8, type = "SOCK")
registerDoSNOW(cl)

st <- proc.time()
dat_out <- ddply(dat, .(cates), function(sub_df){
  sub_df %>% arrange(time) %>% filter(diff(c(0, time)) > 1)
}, .parallel=TRUE) %>% tbl_df
proc.time() - st
# user  system elapsed
# 2.85    0.58    4.32
stopCluster(cl)