Social sites research through CourseRank

Abstract

Social sites such as FaceBook, Orkut, Flickr, MySpace and many others have become immensely popular. At these sites, users share their resources (e.g., photos, profiles, blogs) and learn from each other. On the other hand, higher education applications help students and administrators track and manage academic information such as grades, course evaluations and enrollments. Despite the importance of both these areas, there is relatively little research on the mechanisms that make them effective. Apart from being both a successful social site and an academic planning site, CourseRank provides a live testbed for studying fundamental questions related to social networking, academic planning, and the fusion of these areas. In this paper, we provide a system overview and our main research efforts through CourseRank.

Evaluation of set-based queries with aggregation constraints

“…With the exception of two fragments of SAC queries, namely, SAC queries with one count constraint and at most one groupby count constraint [2] and SAC queries with any number of sum constraints and at most one group-by sum constraint [5], most of the proposed techniques for evaluating more general SAC queries are heuristic solutions [1]. This raises the interesting open question of what is the complexity of evaluating SAC queries with more than one count constraint or SAC queries with a combination of sum and count constraints.…”

“…The space complexity of DP is O(K|S base | + Km ) to keep the matrices for the recurrence relations in the main memory. The time complexity of DP is the summation of the following three components: (1) partitioning S base based on B attribute (denoted by Tpart), (2) computing E matrix, and (3) computing D matrix. To partition S base based on the B attribute, Greedy augments the base query with an ORDER BY clause on B and evaluates this derived query to obtain the set of partitions of S base .…”

“…The usefulness of SAC queries is evidenced by the variants of SAC queries (with different restrictions on the number and types of constraints permitted) that have been studied: OPAC queries for business optimization problems [4], student course planning in the CourseRank project [2], and item recommendations in shopping applications [1].…”

“…Indeed, the evaluation of general SAC queries involving any combination of aggregation, group-by aggregation, and content constraints is an NP-complete problem. Although a number of polynomial [2] and pseudo-polynomial time [4,5] algorithms have been developed for evaluating simpler fragments of SAC queries, most of the proposed evaluation techniques are heuristic methods for more general SAC query fragments [1].…”

“…Currently, there are two fragments of SAC queries that are known to be amenable to efficient evaluation. First, SAC queries with one count constraint and at most one group-by count constraint, where the attribute(s) in the count and group-by constraint are the same, can be evaluated in polynomial time [2]. Second, SAC queries with any number of sum constraints and at most one group-by sum constraint can be evaluated in pseudo-polynomial time [5].…”