R function set.seed() does not work in Power BI R script visual

[mischelin11]

Hi

I need to compute and visualize cluster analysis from R in Power BI desktop. I use set.seed()function to prevent clusters from changing over time. I run this code first in Power BI Desktop and next time i publish it and run it in Power BI services. And here is the problem, because in Power BI desktop it make different (different order) clusters then in Power BI services. Is it possible that the problem is in the set.seed function?

A part of R code is:

set.seed(123)
km.res <- kmeans(scale(costs_noout_of), 4, nstart = 25)
dd <- cbind(costs_noout_all
       , cluster = km.res$cluster)
grid.table(ddply(dd, .(cluster), nrow))

And result is a table where is only numbers of clusters and counts of observations in that clusters, which i need in the following calculations.

Can anyone helps me?

[v-chuncz-msft]

@mischelin11,

If convenient, please elaborate and share a more complete example for a quick test.

[mischelin11]

 Hi @v-chuncz-msft,

I dont know, how to add more complete example for this issue.

I can only add two graphical results, first is from Power BI services and second is from Power BI Desktop. As you can see, these results are a little different.

Result from Power BI services:

Result from Power BI Desktop:

[v-chuncz-msft]

@mischelin11,

This seems to be caused by sorting. Check if you could specify the sort order in R.