How to find the closest element to a given key value in a binary search tree?

Question

Given a bst with integer values as keys how do I find the closest node to that key in a bst ? The BST is represented using a object of nodes (Java). Closest will be for eg 4

Answer 1

Here is the working solution in java which uses the characteristics of BST and additional integer to store minimum difference

public class ClosestValueBinaryTree {
        static int closestValue;

        public static void closestValueBST(Node22 node, int target) {
            if (node == null) {
                return;
            }
            if (node.data - target == 0) {
                closestValue = node.data;
                return;
            }
            if (Math.abs(node.data - target) < Math.abs(closestValue - target)) {
                closestValue = node.data;
            }
            if (node.data - target < 0) {
                closestValueBST(node.right, target);
            } else {
                closestValueBST(node.left, target);
            }
        }
    }

Run time complexity - O(logN)

Space time complexity - O(1)

Answer 2

Traverse takes O(n) time. Can we proceed it in top-bottom? like this recursive code:

Tnode * closestBST(Tnode * root, int val){
    if(root->val == val)
        return root;
    if(val < root->val){
        if(!root->left)
            return root;
        Tnode * p = closestBST(root->left, val);
        return abs(p->val-val) > abs(root->val-val) ? root : p;
    }else{
        if(!root->right)
            return root;
        Tnode * p = closestBST(root->right, val);
        return abs(p->val-val) > abs(root->val-val) ? root : p;
    }   
    return null;
}

Answer 3

void closestNode(Node root, int k , Node result) {
    if(root == null) 
    {
       return;      //currently result is null , so it  will be the result
    }
    if(result == null || Math.abs(root.data - k) < Math.abs(result.data - k) )
    {
      result == root;
    }
    if(k < root.data)
    {
    closestNode(root.left, k, result)
    } 
    else 
    {
        closestNode(root.right, k, result);
    }

}

Answer 4

Below one works with different samples which I have.

public Node findNearest(Node root, int k) {
    if (root == null) {
        return null;
    }
    int minDiff = 0;
    Node minAt = root;
    minDiff = Math.abs(k - root.data);

    while (root != null) {
        if (k == root.data) {
            return root;
        }
        if (k < root.data) {
            minAt = updateMin(root, k, minDiff, minAt);
            root = root.left;
        } else if (k > root.data) {
            minAt = updateMin(root, k, minDiff, minAt);
            root = root.right;
        }

    }
    return minAt;
}

private Node updateMin(Node root, int k, int minDiff, Node minAt) {
    int curDif;
    curDif = Math.abs(k - root.data);
    if (curDif < minDiff) {
        minAt = root;
    }
    return minAt;
}

Answer 5

It can be solved in O(log*n*) time.

• If the value in a node is same as the given value, it's the closest node;
• If the value in a node is greater than the given value, move to the left child;
• If the value in a node is less than the given value, move to the right child.

The algorithm can be implemented with the following C++ code:

BinaryTreeNode* getClosestNode(BinaryTreeNode* pRoot, int value)
{
    BinaryTreeNode* pClosest = NULL;
    int minDistance = 0x7FFFFFFF;
    BinaryTreeNode* pNode = pRoot;
    while(pNode != NULL){
        int distance = abs(pNode->m_nValue - value);
        if(distance < minDistance){
            minDistance = distance;
            pClosest = pNode;
        }

        if(distance == 0)
            break;

        if(pNode->m_nValue > value)
            pNode = pNode->m_pLeft;
        else if(pNode->m_nValue < value)
            pNode = pNode->m_pRight;
    }

    return pClosest;
}

You may visit my blog for more details.

Answer 6

Traverse the tree as you would to find the element. While you do that record the value that is closest to your key. Now when you didn't find a node for the key itself return the recorded value.

So if you were looking for the key 3 in the following tree you would end up on the node 6 without finding a match but your recorded value would be 2 since this was the closest key of all nodes that you had traversed (2,7,6).

                 2
              1      7
                   6   8