Enhancement of HCB Tree for Improving Retrieval Performance and Dynamic Environments
Sung Wan Kim
Abstract
Open-access reader
Sung Wan Kim
Abstract
Open-access reader
이진 트라이를 이진 비트열로 압축하여 표현하는 CB 트리는 키가 늘어남에 따라 이진 비트열이 길어지게 되어 검색 시간이 증가하며 잦은 키 삽입/삭제 연산에 비효율적이다. 작은 분할 트라이들을 계층적 구조로 표현한 HCB 트리가 제안되었으나 비트열 시프트 처리를 근본적으로 해결할 수 없으며 자식 혹은 부모 트리 참조를 위해 별도의 자료 구조를 탐색해야 하는 부담이 있다. 본 논문에서는 각 분할 트리를 포화 이진 트라이 형태로 표현하고 레벨 순위에 따라 분할 트리 번호를 할당하여 검색 성능을 향상 시키는 한편 키의 삽입/삭제 시에 시프트 연산이 발생하지 않도록 하였다. 시 공간 복잡도를 사용한 성능 평가에서 검색 시에는 제안 방법과 HCB 트리 방법이 CB 트리에 비해 우수한 것으로 나타났으며, 키 삽입/삭제는 제안 방법이 가장 높은 성능을 보여주었다. 공간 사용량은 제안 방법이 CB 트리 방법에 비해 71~89%의 공간만을 요구하여 가장 좋은 성능을 보였다. CB tree represents the binary trie by a compact binary sequence. However, retrieval time grows fast since the more keys stored in the trie, longer the binary sequences are. In addition it is inefficient for frequent key insertion/deletion. HCB tree is a hierarchical CB tree consisting of small binary tries. However it can not avoid shift operations and have to scan an additional table to refer child or parent trie. In order to improve retrieval performance and avoid shift operations when keys are inserted or deleted, we in this paper represent each separated trie by a full binary trie and then assign the unique identifier to it. Finally the theoretical evaluations show that both the proposed approach and HCB tree provides better than CB tree for key retrieval. The proposed approach shows the highest performance in case of key insertion/deletion and moreover requires only 71%~89% of storage as compared with CB tree.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
이진 트라이를 이진 비트열로 압축하여 표현하는 CB 트리는 키가 늘어남에 따라 이진 비트열이 길어지게 되어 검색 시간이 증가하며 잦은 키 삽입/삭제 연산에 비효율적이다. 작은 분할 트라이들을 계층적 구조로 표현한 HCB 트리가 제안되었으나 비트열 시프트 처리를 근본적으로 해결할 수 없으며 자식 혹은 부모 트리 참조를 위해 별도의 자료 구조를 탐색해야 하는 부담이 있다. 본 논문에서는 각 분할 트리를 포화 이진 트라이 형태로 표현하고 레벨 순위에 따라 분할 트리 번호를 할당하여 검색 성능을 향상 시키는 한편 키의 삽입/삭제 시에 시프트 연산이 발생하지 않도록 하였다. 시 공간 복잡도를 사용한 성능 평가에서 검색 시에는 제안 방법과 HCB 트리 방법이 CB 트리에 비해 우수한 것으로 나타났으며, 키 삽입/삭제는 제안 방법이 가장 높은 성능을 보여주었다. 공간 사용량은 제안 방법이 CB 트리 방법에 비해 71~89%의 공간만을 요구하여 가장 좋은 성능을 보였다. CB tree represents the binary trie by a compact binary sequence. However, retrieval time grows fast since the more keys stored in the trie, longer the binary sequences are. In addition it is inefficient for frequent key insertion/deletion. HCB tree is a hierarchical CB tree consisting of small binary tries. However it can not avoid shift operations and have to scan an additional table to refer child or parent trie. In order to improve retrieval performance and avoid shift operations when keys are inserted or deleted, we in this paper represent each separated trie by a full binary trie and then assign the unique identifier to it. Finally the theoretical evaluations show that both the proposed approach and HCB tree provides better than CB tree for key retrieval. The proposed approach shows the highest performance in case of key insertion/deletion and moreover requires only 71%~89% of storage as compared with CB tree.
Key concepts: Tree (set theory), Computer science, Environmental science, Mathematics, Mathematical analysis