optimization - Subset data with dynamic conditions in R -

i have dataset of 2500 rows bank loans. each bank loan has outstanding amount , collateral type. (real estate, machine tools.. etc)

i need draw random selection out of dataset example sum of outstanding amount = 2.5million +-5% , maximum 25% loans same asset class.

i found function optim, asks function , looks constructed optimization portfolio of stocks, more complex. there easy way of achieving this?

i created sample data set illustrate question better:

dataset <- data.frame(balance=c(25000,50000,35000,40000,65000,10000,5000,2000,2500,5000)                       ,collateral=c("real estate","aeroplanes","machine tools","auto vehicles","real estate",                                     "machine tools","office equipment","machine tools","real estate","auto vehicles")) 

if want example 5 loans out of dataset sum of outstanding balance = 200.000 (with 10% margin) , not more 40% allowed same collateral type. (so maximum 2 out of 5 in example)

please let me know if additional information necessary. many thanks, tim

this function made works:

pick_records <- function(df,size,bal,collat,max.it) {   <- 1   j <- 1   while ( == 1 ) {     s_index <- sample(1:nrow(df) , size)     print(s_index)     output <- df[s_index,]     out_num <- lapply(output,as.numeric)     tot.col <- sum(as.numeric(out_num$collateral))     if (sum(out_num$balance) < (bal*1.1) &           sum(out_num$balance) > (bal*0.9) &           all(  table(out_num$collateral)/size  <= collat)   ) {       return(output)       break     }     print(j)     j <- j + 1     if ( j == max.it+1) {       print('no solution found')       break}        } }   > <- pick_records(dataset,5,200000,0.4,20) >   balance       collateral 3   35000    machine tools 7    5000 office equipment 4   40000    auto vehicles 5   65000      real estate 2   50000       aeroplanes 

where df dataframe, size number of records want , max.it number of maximum iterations find solution before returning no solution found error, bal limit balance , collat same collateral. can change please.

let me know if don't part of it.