=Paper=
{{Paper
|id=Vol-3890/paper-16
|storemode=property
|title=Temporal data modeling evaluation in knowledge graphs: A healthcare use case
|pdfUrl=https://ceur-ws.org/Vol-3890/paper-16.pdf
|volume=Vol-3890
}}
==Temporal data modeling evaluation in knowledge graphs: A healthcare use case==
https://ceur-ws.org/Vol-3890/paper-16.pdf
Temporal Data Modelling Evaluation in Knowledge Graphs:
A Healthcare Use Case
Sepideh Hooshafza1 , Gaye Stephens1 , Mark A. Little1,2 and Beyza Yaman1
1 ADAPT Research Centre, School of Computer Science and Statistics, Trinity College Dublin, Dublin, Ireland
2 Trinity Health Kidney Centre, Trinity College Dublin, Dublin, Ireland
Abstract
Healthcare data, such as patients’ symptoms, laboratory test results, and various clinical measurements
are temporal in nature, and are associated with a time. Modelling temporal healthcare data could benefit
healthcare practitioners in healthcare decision making and support patient care. One method for
modelling data that has been used in academia and industry is RDF-based knowledge graphs (KGs).
Many approaches have been proposed to model temporal data in RDF-based KGs which have not been
evaluated systematically in the healthcare domain. In this paper, an evaluation framework is proposed
for the evaluation of temporal data modeling approaches in KGs and has been applied in a healthcare
use case.1
1. Introduction
Modelling temporal healthcare data supports medical professionals to comprehend disease
patterns, evaluate patient histories, identify relationships in clinical events, and make informed
predictions for improved patient care [1, 2]. A number of RDF-based approaches to modelling
temporal data in KGs exist, including “standard reification”, and “singleton property”[3,
4]. However, to the best of our knowledge, the current RDF based temporal data modelling
approaches have not been systematically evaluated using a healthcare use case. Existing
approaches have produced different results in terms of complexity, and performance which
require further evaluation and comparison [5]. In this study, an evaluation framework is
proposed to evaluate temporal data modelling approaches in KGs. The evaluation framework
components consist of six phases including data modeling approaches identification, use case
identification, KG creation, KG hosting, KG deployment, metrics identification and evaluation. The
evaluation framework was applied to a healthcare use case that addresses modeling medication
data for patients with the rare disease anti-neutrophil cytoplasmic antibody (ANCA) Associated
Vasculitis (AAV) in FAIRVASC [6, 7] project. The framework will guide data and knowledge
engineers in evaluating various temporal data modeling approaches within KGs.
2. Experiment
Two well-known approaches for adding temporal data to a KG were chosen including singleton
property and standard reification. The evaluation was performed based on the six phases of the
evaluation framework and the healthcare use case. The dataset contains a total of 600 patients.
The data included details regarding the medications utilized for patients, including both the start
and stop dates for each medication. Two ontologies were designed based on identified
approaches, RDF data was generated using the R2RML engine. RDF data were imported into a
triple store. A competency question was designed and RDF data was queried using SPARQL.
SWAT4HCLS 2024: The 15th International Conference on Semantic Web Applications and Tools for Health Care and
Life Science, February 26–29, 2024, Leiden, Netherlands
sepideh.hooshafza@adaptcentre.ie (S. Hooshafza); gaye.stephens@adaptcentre.ie (G. Stephens);
mark.little@adaptcentre.ie (M. Little); beyza.yaman@adaptcentre.ie (B. Yaman)
0000-0002-1061-1572 (S. Hooshafza); 0000-0001-5384-6139 (G. Stephens); 0000-0001-6003-397X (M. Little);
0000-0003-2130-0312 (B. Yaman)
© 2024 Copyright for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
The comparative analysis between the singleton property and standard reification approaches
provides insights into their performance and complexity within the realm of RDF data modeling.
In terms of modeling complexity, the standard reification approach is more complex than the
singleton property approach. While assessing performance metrics, there isn't a notable
difference between the two approaches, primarily attributed to the limited size of the dataset.
Table 1- Experiments results based on modelling complexity and performance.
Category Metric Singleton Standard
property reification
Modelling Number of Statements 22904 36735
Complexity Additional triple generation 13544 27375
Resource redundancy 21137 34479
Performance Data load time in triple store 0.5s 2s
Query length requirement to execute a particular task 9 9
Query response time 6.6 s 0.2 s
3. Conclusion
This study proposed an evaluation framework for evaluating temporal data modeling approaches
in KGs. The framework can guide data and knowledge engineers in evaluating various temporal
data modeling approaches within KGs. With this knowledge, they will be able to choose the
methods that will best meet their needs when modeling temporal health data in graph databases.
Acknowledgements
This work was funded by the ADAPT Centre for Digital Content Technology under the SFI
Research Centres Programme (Grant13/RC/2106-P2).
References
[1] N Poh, N., S. Tirunagari, and D. Windridge. Challenges in designing an online healthcare
platform for personalised patient analytics. in 2014 IEEE Symposium on Computational
Intelligence in Big Data (CIBD). 2014
[2] Combi, C., G. Cucchi, and F. Pinciroli, Applying object-oriented technologies in modeling and
querying temporally oriented clinical databases dealing with temporal granularity and
indeterminacy. IEEE Trans Inf Technol Biomed, 1997. 1(2): p. 100-27.
[3] Hernández, D., A. Hogan, and M. Krötzsch. Reifying RDF: What works well with wikidata?
2015.
[4] Nguyen, V., O. Bodenreider, and A. Sheth. Don't like RDF reification? Making statements
about statements using singleton property. 2014.
[5] Magkanaraki, A., et al. Benchmarking RDF schemas for the Semantic Web. in Lecture Notes
in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture
Notes in Bioinformatics). 2002
[6] Yaman, B., et al., Towards A Rare Disease Registry Standard: Semantic Mapping of Common
Data Elements Between FAIRVASC and the European Joint Programme for Rare Disease
[7] McGlinn, K., et al., FAIRVASC: A semantic web approach to rare disease registry integration.
Computers in Biology and Medicine, 2022. 145.