0

I have below code, which takes ages to run. I have already tried multiple things such as: limiting the amount of loops, using ifelse statements, and trying to declare as much thing as possible outside of the loop. However, it still takes a very long time.

What would be a good way to improve on this part of my code to improve its processing speed? Are there some things I'm not seeing?

z <- 625
numDays <- 365
k <- numDays * 96


#To estimate the size of the list
df <- rmarkovchain(n=365, object = mcList, t0= "home", include.t0 = TRUE)
allTheCars <- rep(list(df), z)

#an example of  df below:
Locations <- c("Home", "Bakery", "Grocery", "Home-Bakery", "Home-Grocery", "Bakery-Home", "Bakery-Grocery", "Grocery-Home", "Grocery-Bakery")
Iteration <- rep(seq(1:96), 365)

df <- data.frame(Iteration, sample(Locations, k, replace = TRUE))


#The loop takes a huge amount of time

for(y in 1:z){



  df <- rmarkovchain(n=365, object = mcList, t0= "Home", include.t0 = TRUE)
  df$Begin <- 0
  df[1,3] <- b
  df$Still <- ifelse(df$values == "Home", 1, 0)
  df$KM <- vlookup(df$values, averageDistance, lookup_column = 1, result_column = 2)
  df$Load <- ifelse(df$Still == 1, cp, 0)
  df$costDistance <- df$KM * 0.21
  df$End <- 0
  df[is.na(df)] <- 0
  df$reduce <- rep(seq(1:97), numDays)
  df <- df %>% filter(reduce != 97) 
  df$Load <- ifelse(df$reduce <= 69 | df$reduce >= 87, df$Load, 0)




      for(i in 1:k) {

      df[i,3] <- ifelse(df[i,3] < b, pmin(df[i,3] + df[i,6], b),  df[i,3])

      df[i,8] <- df[i,3] - df[i,7]
      j <- i + 1
      df[j,3] <- df[i,8]
      }

allDf[[y]] <- df

}

EDIT: After Minem's suggestion to look at Profvis I found out that the second for-loop takes by far the most amount of time, which now looks like this:

for(i in 1:k) {
      mainVector <- df[i,3]
      extra <- df[i,6]
      subtractingVector <- df[i,7]


      mainVector <- ifelse(mainVector < b, pmin(mainVector + extra, b),  mainVector )

      newMain <- mainVector - subtractingVector
      j <- i + 1
      df[j,3] <- newMain 
      }

Now the vectorization of the first three vectors takes some time and the last line of code, which integrates the calculated value back in the dataframe costs the most time. Is there anyway to improve upon this?

Edit 2: Reproducible example for all

library(dplyr)
library(markovchain)
library(expss)

matrixExample <- matrix(sample(runif(81, min = 0 , max =1), replace = FALSE ), nrow = 9, ncol = 9)
mcListLoop <- rep(list(matrixExample), 96)
mcList <- new("markovchainList", markovchains = mcListLoop)


distance <- runif(9, min = 5, max =10)
Locations <- c("Home", "Bakery", "Grocery", "Home-Bakery", "Home-Grocery", "Bakery-Home", "Bakery-Grocery", "Grocery-Home", "Grocery-Bakery")

averageDistance <- data.frame(cbind(distance, Locations))
JosvdBerg
  • 31
  • 7
  • 1
    What is `rmarkovchain`? Have you tried to profile your code? https://rstudio.github.io/profvis/ – minem Jun 09 '20 at 06:44
  • `mcList` is not defined... – minem Jun 09 '20 at 06:54
  • Thanks for the tip, I used profvis and somewhat improved my code already accordingly. It is ~3 times faster now. The things that are still taking is the final part of the loop df[j,3] <- df[i,8] . The df[i,8] has already been vectorized. Is there a faster way to assign values and replace values within a dataframe? – JosvdBerg Jun 09 '20 at 08:31
  • 1
    Can you supply reproducible example? Because, as I noted, you haven't supplied `mcList`... – minem Jun 09 '20 at 10:47
  • I have included a reproducible example including mcList and averageDistance. Apologies for the inconvenience, I unjustifiably thought you would not need those since I didn't think the problem to be in that part of the code. Thank you for help! – JosvdBerg Jun 09 '20 at 11:19
  • Still no definition of `mcList`. Reproducible should mean that we can copy-paste the code and it should run. https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example – minem Jun 09 '20 at 12:09
  • Sorry, I used the wrong name, it is good now. The list alltheCars was supposed to be mcList in the reproducible code example – JosvdBerg Jun 09 '20 at 12:14
  • Let us [continue this discussion in chat](https://chat.stackoverflow.com/rooms/215580/discussion-between-minem-and-josvdberg). – minem Jun 09 '20 at 12:17

0 Answers0