pass_fun1 <- function(transfer,gps){

 nrow_transfer <- nrow(transfer)

 for(i in 1:nrow_transfer){

gps_temp <- subset(gps, LINE == transfer[i,]$b_line & gps_time >= transfer[i,]$alight_time

  & gps_time <= (transfer[i,]$b_ic_time+30) & ln_stop_id == transfer[i,]$b_board_stop)

transfer[i,]$pass <- length(unique(gps_temp$vehicleID))

 }

 return(transfer)

}

pass_fun2 <- function(transfer,gps){

 nrow_transfer <- nrow(transfer)

 for(i in 1:nrow_transfer){

gps_temp1 <- subset(gps,LINE == transfer[i,]$b_line)

gps_temp2 <- subset(gps_temp1, gps_time >= transfer[i,]$alight_time)

gps_temp3 <- subset(gps_temp2, gps_time <= (transfer[i,]$b_ic_time+30))

gps_temp4 <- subset(gps_temp3, ln_stop_id == transfer[i,]$b_board_stop)

transfer[i,]$pass <- length(unique(gps_temp4$vehicleID))

 }

 return(transfer)

}

上面两个函数执行的是同样的功能，但是pass_fun2的效率要比pass_fun1高很多，因为前者拆分成了4次判断，而后面的每次判断都是在前一次判断结果基础上做出的，也就是说是在全体样本的越来越小的子集上进行的，效率相对更高。