Visual-aided ontology-based ranking on multidimensional data: a case study in academia
| DOI | https://doi.org/10.1108/DTA-03-2017-0014 |
| Published date | 02 July 2018 |
| Pages | 366-383 |
| Date | 02 July 2018 |
| Author | Evangelia Triperina,Georgios Bardis,Cleo Sgouropoulou,Ioannis Xydas,Olivier Terraz,Georgios Miaoulis |
| Subject Matter | Library & information science,Librarianship/library management,Library technology,Information behaviour & retrieval,Metadata,Information & knowledge management,Information & communications technology,Internet |
Visual-aided ontology-based
ranking on multidimensional data:
a case study in academia
Evangelia Triperina
Department of Informatics and Computer Engineering, University of West Attica,
Athens, Greece and
XLIM Laboratory, University of Limoges, Limoges, France
Georgios Bardis, Cleo Sgouropoulou and Ioannis Xydas
Department of Informatics and Computer Engineering, University of West Attica,
Athens, Greece
Olivier Terraz
XLIM Laboratory, Universite de Limoges, Limoges, France, and
Georgios Miaoulis
Department of Informatics and Computer Engineering, University of West Attica,
Athens, Greece and
XLIM Laboratory, University of Limoges, Limoges, France
Abstract
Purpose –The purpose of this paper is to introduce a novel framework for visual-aided ontology-based
multidimensional ranking and to demonstrate a case study in the academic domain.
Design/methodology/approach –The paper presents a method for adapting semantic web technologies
on multiple criteria decision-making algorithms to endow to them dynamic characteristics. It also showcases
the enhancement of the decision-making process by visual analytics.
Findings –The semantic enhanced ranking method enables the reproducibility and transparency of ranking
results, while the visual representation of this information further benefits decision makers into making
well-informed and insightful deductions about the problem.
Research limitations/implications –This approach is suitable for application domains that are ranked
on the basis of multiple criteria.
Originality/value –The discussed approach provides a dynamic ranking methodology, instead of focusing
only on one application field, or one multiple criteria decision-making method. It proposes a framework that
allows integration of multidimensional, domain-specific information and produces complex ranking results in
both textual and visual form.
Keywords Ranking, Ontology, Visual analytics, ELECTRE III, Semantic web,
Multiple criteria decision making
Paper type Research paper
1. Introduction
The typical context for multiple criteria decision making (MCDM) methodologies arises
when a set of alternative options has to be evaluated against a set of criteria each
contributing to the final outcome. These alternatives are systematically represented as
vectors of values according to a common set of attributes (dimensions) and subsequently
evaluated, upon these attribute values, against a set of criteria pertaining to the decision
maker’s (DM) preferences. The evaluation may be desired to lead to: elicitation of the best
among the available options; ranking of the available options; grouping of the options in
similarity groups; or definition, in the constructive sense, of the optimal option(s) (Doumpos
and Zopounidis, 2014). The latter case is covered by multiple objective optimization, where
the alternative options are considered to include all valid combinations of attribute values,
Data Technologies and
Applications
Vol. 52 No. 3, 2018
pp. 366-383
© Emerald PublishingLimited
2514-9288
DOI 10.1108/DTA-03-2017-0014
Received 2 March 2017
Revised 19 March 2018
24 April 2018
Accepted 27 May 2018
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/2514-9288.htm
366
DTA
52,3
thus requiring the identification of the optimal combination conforming to DM’s
desired outcome. The former three cases fall within the area of multiple criteria decision
analysis (MCDA), where the options are predefined as a set of concrete, usually
reality-bound, instances, representing a typically small subset of the possible attribute
values’combinations Chen et al. (2011).
Rankings, in particular, provide insights about the order of a group of similar objects and
have many applications, such as academic, economic and sports related. In academic
strategic planning for example, a policy maker may require past performances of the
university’s departments (the academia case will be used as a core example throughout this
work). Normally, the policy maker would have to access records related to educational,
research and collaboration activities of each department and correlate these performances to
decide on future actions. Usually, this kind of information is scattered across multiple
platforms and websites and calls for considerable human time and effort to accumulate.
All relevant information could be provided through an integrated system, facilitating its
exploration and proces sing by the user. The compa rison mechanisms of obj ect
performances on selected criteria range from simple aggregation techniques, following
the classical approach of utility function, to complex decision analysis tools represented by
the family of outranking methods. The former techniques are applied in problems that
demand the handling of a small set of criteria or vast numbers of alternative options,
whereas the latter are applied when the final decision depends on multiple variables and
requires the consideration of potentially contradicting partial evaluations.
Ranking processes should satisfy several requirements, such as efficiency, transparency,
personalization and easier comprehension by DMs. The motivation for our work emanates
from these requirements and therefore addresses ranking problematic by implementing a
visual, dynamic method based on an outranking algorithm. More specifically, efficiency is
facilitated by ranking algorithms, which in our case, due to the multivariate nature of the
problem, should support multiple criteria and the possibility of ambiguous evaluations,
while transparency is ensured by ontologies that ease data sharing, openness and
interoperability. Personalization is achieved through the modular design of the approach, in
which the various facets can be used independently or in combination, to deliver ranking
profiles that are representative of the application domain. The easier comprehension of the
ranking procedure and outputs is accomplished by visual analytics, which exploit the
human ability to process larger amount of data in a visual form (Thomas and Cook, 2006).
The main novelty of this method is the integration of ontologies and visual analytics with
the outranking method to provide a framework that supports the entire decision-making
process. The paper is structured as follows: after the introduction, a detailed literature
review, focusing on the involved scientific fields, is presented. The proposed framework and
the developed system are described in detail in the subsequent sections and specific case
studies on academia and World Development Indicators (WDI) are outlined in Section 5.
Conclusions and future work considerations are discussed in the final section.
2. Literature review
Following the formalism in Vincke (1992), the available options in any decision making task
constitute a set of objects G, which is assumed to be finite and stable during the process.
One of three, mutually exclusive, relations must necessarily apply between any two
members of G: preference (P), indifference (I) or incomparability (R). To ensure a valid
preference structure, where the chosen decision-making mechanism will be applied, the
aforementioned relations must fulfill the following: ∀x,y∈G,xPy⇒(yPx)’;∀x∈G,xIx;
∀x,y ∈G,xIy⇒yIx;∀x∈G,(xRx)’;∀x, y ∈G,xRy⇒yRx. These conditions do not imply
transitivity per se, thus allowing xPy,yPzand zPxto hold concurrently in an otherwise
valid preference structure. Rankings represent the summation of the DM’s preferences
367
Visual-aided
ontology-based
ranking
To continue reading
Request your trial