> Use an array of size k as a min heap.
> With every element, compare it with min of heap, if found larger, insert in heap, else, ignore.
> Once the heap is full, if found larger than min of heap, delete the min and insert.

The Time Complexity should be at least $O(N)$ for scanning the items. If you’d only considered the $k$ sized heap then it should’ve been bound by terms of $k$, and $k$ here isn’t a constant, it might as well be $N$ for all it says.

Also, if I’m understanding correctly, you’re using a max heap. Finding $k$^{th} maximum element(considering the max-heap element to replace when a better alternative is to be inserted) takes at least $k/2$ time(scanning leaves) since it can be present anywhere within the leaves. Time complexity would be $O(N.K)$ if it would be the case.

The Correct Option would be D using a min-heap and finding deleting min elements whenever we find a element greater than min-heap’s min thus finding max. Here, Time Complexity would be $O(N.\log k)$. Every time an element greater than min of heap is found, it takes $O(\log k)$ time to insert.

> Also, if I’m understanding correctly, you’re using a max heap. Finding kth maximum elements takes at least k/2 time(scanning leaves) since it can be present anywhere within the leaves.

I Believe what OP has done is created a max heap in time O(N), and then extracted K roots. i.e KLogN, the min-heap method should work in NLogK as we are scanning n elements and performing insertion/deletion in LogK time.

But they can be copied right? Just take the n elements, copy them to an array, buildMaxHeap in O(N). I am not sure if I get your question correctly or maybe I am going wrong somewhere?