Find Mode Value From A List Of Float Data

Find the frequency of each item!!!then find the maximum of frequencies(u can use any good sort algorithm), the number corresponding to this frequncies will be mode.

all u need to be carefull is that u can keep frequency and its candidate bounded so that when u find frequency u know what the number is and when u know number u also know what its frequency is in the list;)

RanjanNa at 2007-7-14 18:48:55 >

I'd make a binary search tree with nodes that contain the float and frequency.. Go down the tree with each float. If you find it, increment the frequency. If you dont, create a new node with that float and frequency=1. Insert time=log(n) for any one float, and you can get your frequencies sorted by float in linear time.

dforbua at 2007-7-14 18:48:55 >

This code should be tested! The suggestion of using a BST (which by the way could have a very bad runtime, unless it was balanced) or other object data structures is inefficient.This problem does not require anything as complex. This code runs in O(nlogn) time.

float[] data = ...;

java.util.Arrays.sort(data);

int longest = 0;

float mode = 0;

int i = 1;

while (i < data.length) {

int temp = 1;

while (i < data.length) {

if (data[ i ] == data[i-1]) {

temp++;

i++;

}

else break;

}

if (temp > longest) {

longest = temp;

mode=data[i-1];

}

i++;

}

rkippena at 2007-7-14 18:48:55 >

> The suggestion of using a

> BST (which by the way could have a very bad runtime,

> unless it was balanced) or other object data

> structures is inefficient.This problem does not

> require anything as complex. This code runs in

> O(nlogn) time.

Well, I was gonna come in and blast you for doggin my idea, but the more I think about it, you really do have nice simple efficient code there, I like it...

But, to defend my original reply, the only inefficiency I can see with the tree solution is all the node objects that need to be generated, they will take the extra cpu to instantiate, and a little extra memory since I'm holding frequencies as well as the float.. If you have BST code already, then it really isn't any harder to code than your array implementation.

There's a lot of benefits from having your data in tree form that you don't have in arrays. A tree could still be a better idea if you will need to add/remove/search nodes later. A tree is actually very easy to keep balanced, even if your data starts off sorted. I suppose whether to use arrays or trees would depend on how you need to operate on this data later. If it's a one time calculation, and you're just gonna dump the data when you're done, then your simple code will be a better answer.

dforbua at 2007-7-14 18:48:55 >