Publication Type

Conference Proceeding Article

Version

acceptedVersion

Publication Date

3-2024

Abstract

Similarity joins on multidimensional data are crucial operators for recommendation purposes. The classic ��-join problem finds all pairs of points within �� distance to each other among two ��-dimensional datasets. In this paper, we consider a novel and alternative version of ��-join named community similarity based on user profile joins (CSJ). The aim of CSJ problem is, given two communities having a set of ��-dimensional users, to find how similar are the communities by matching every single pair of users (a user can be matched with at most one other user) having an absolute difference of at most �� per dimension. Each dimension in each user vector stores a counter that measures the number of user preferences to a concrete general topic. CSJ uses low �� and applies the strict condition per dimension so as to really find similar user profiles among two communities. CSJ applies to a number of cases in which the popular community detection and community search problems do not suit. This happens since CSJ treats existing communities as brands of a specific commercial value and does not search to form general communities as prior works do; these two community types semantically differ. Also, CSJ does not rely its execution on social or/and physical links among community users, instead, it only focus on the similarity of user profiles. We deploy a suite of 6 methods to solve CSJ; 3 approximate and 3 exact algorithms. We evaluate our solutions to meaningful case studies of real and synthetic datasets having different characteristics. Our experimental results show interesting and diverse conclusions of CSJ applicability to realistic scenarios.

Discipline

Databases and Information Systems

Research Areas

Data Science and Engineering

Publication

Proceedings 27th International Conference on Extending Database Technology ( EDBT 2024 ) Paestum, Italy, March 25-28

Volume

27

First Page

572

Last Page

583

ISBN

2367-2005

Identifier

10.48786/edbt.2024.49

Publisher

EDBT

City or Country

Italy

Copyright Owner and License

Authors

Additional URL

https://doi.org/10.48786/edbt.2024.49

Download

Share

COinS