Open AccessContext-Aware Dynamic Object Relationship Modeling
Author(s) -
Toyoda Kentaroh,
Rachel Gan Kai Ying,
Tan Puay Siew,
Allan N. Zhang
Publication year - 2022
Publication title -
international journal of modeling and optimization
Language(s) - English
Resource type - Journals
ISSN - 2010-3697
DOI - 10.7763/ijmo.2022.v12.798
Subject(s) - computer science , discriminative model , context (archaeology) , semantics (computer science) , similarity (geometry) , object (grammar) , focus (optics) , ontology , measure (data warehouse) , data mining , artificial intelligence , information retrieval , machine learning , image (mathematics) , paleontology , philosophy , physics , epistemology , optics , biology , programming language
Finding relationships in the data is essential for object modeling. However, existing methods generally focus on pre-defined static relationships using semantics and ontology, which is inappropriate when we are interested in dynamic relationships between objects that appear in data sources (e.g. log files). In this paper, we propose two novel methods to dynamically extract contextual relationships that appear in heterogeneous data sources. Our method detects contexts (e.g. time and location) in a given data source and quantifies the similarities between objects based on the detected contexts. Specifically, our methods consist of (i) a fast and accurate context detection method with carefully engineered discriminative features and (ii) a similarity measure that takes into account contexts. We evaluated our context detection method with an open dataset to show its detection accuracy and speed.