BFS에 기반한 포스트 랭킹: 알고리즘 및 성능 평가

초록

Existing post ranking algorithms assume that the ranking of a post is related to both of the relevant posts and bloggers connected by blog-actions such as trackbacks, scraps, and comments. For this reason, existing algorithms factor in distant related posts/bloggers, which negatively influences ranking accuracy. In this paper, to eliminate the effect of less related posts and bloggers in ranking, we propose a novel post ranking algorithm, PBFS. This algorithm selects the relevant posts and bloggers in the manner of the breadth first search and additionally considers weights by the relatedness between a target post and the other posts/bloggers. We evaluate the accuracy of ranking algorithms by performing extensive experiments using real-world blog data. The experimental results reveal that our algorithm outperforms the existing algorithms.

키워드