Java random number with high probability

Question

I have an array char[] Select = {'A','B','C','D','E','F','G','H','I','J'} and each element in this array has different probability to be selected. For example,

int[] Weight =  {10,30,25,60,20,70,10,80,20,30};

My requirement is to select 5 elements from this array and the element with high weight values has higher probability to be selected and these 5 elements should be different.

My plan is first sum the weight

int[] weightSum = {10, 40, 65, 125, 145, 215, 225, 305, 325, 355}

Then I use Random to generate a random number k in the range of [0,355]. Then looking for the first element which is bigger than k in the weightSum[]. This process are repeated 5 times.

The problem is that the element with high probability could be selected multiple times. I try to remove the duplicate elements at each iteration. The duplicates are removed, but element with high weight values are not selected.

How to solve this problem?

thanks.

shouldn't you look for the *last* element *smaller* than `k`? — Paulo Guedes, Aug 26 '11 at 19:19

score 3 · Answer 1 · answered Aug 26 '11 at 19:20

3

Not sure I understand correctly, but how about something like this:

after first selection you remove the selected element from char[] Select
you also remove the corresponding weight from int[] Weight
regenerate int[] weightSum
repeat the whole process

answered Aug 26 '11 at 19:20

bpgergo

15,669
5
44
68

1

If s77 does not want repeats, the simplest thing would be after selecting a value set the corresponding weight to 0. – emory Aug 26 '11 at 19:33

score 1 · Answer 2 · answered Aug 26 '11 at 19:21

1

I guess each time you remove the duplicates, you must also update your weightSum array.

answered Aug 26 '11 at 19:21

Paulo Guedes

7,189
5
40
60

ratchet freak · Accepted Answer · 2011-08-26T22:48:33.653

don't maintain the cumulative sum or adjust it each time: (requires O(n) for each selection though)

char[] Select = {'A','B','C','D','E','F','G','H','I','J'};
int[] Weight = {10,30,25,60,20,70,10,80,20,30};
int sum = 355;
for(int a=0;i<5;i++){
    int rand = (int)(Math.random()*sum);
    int s=0;//temp cumulative sum
    int i=0;
    while( (s+=Weight[i])<rand)i++;
    result.add(Select[i]);

    sum-=Weight[i];//total weight is lower now
    Weight[i]=0;//if weight is 0 it will never be selected

}

edit: fixed so I don't subtract 0 from sum

score 0 · Answer 4 · answered Aug 26 '11 at 19:14

I'm not really understanding your problem, but your algorithm sounds right: you should be doing something like storing each generated value in a list (based on random number generator), but first check to see if that number already exists in the list before adding it. Repeat until the list has 5 numbers.

score 0 · Answer 5 · answered Aug 26 '11 at 19:20

My statistics memory is a bit fuzzy, but I think what you want to do is remove the element from consideration after it's been selected. In other words, after selecting the entry, remove that entry from weightSum and subtract its Weight from all subsequent entries and the range of the random number. Might be easier to manage if you work with ArrayLists instead of primitive arrays.

Java random number with high probability

5 Answers5