Recommenddit: A Recommendation Service for Reddit Communities

摘要

Reddit.com is a popular website consisting of many online forums, or "subreddits", each centered on a particular topic, many of which are closely related. However, two communities that might appear to be similar may in fact have little overlap in their user-base or content, while apparently dissimilar subreddits may actually represent topics that appeal to substantially overlapping audiences. At present, there is a lack of effective methods for identifying these relations. In this paper, we propose and implement an automated method of identifying such relations using an association rule mining algorithm. Given the output from this algorithm, we develop a web-based application named Recommenddit that a user can query to retrieve subreddits closely associated with a given subreddit.