Big Data Research 36 (2024) 100445

Contents lists available at ScienceDirect

Big Data Research

journal homepage: www.elsevier.com/locate/bdr

Interactive big data visualization and analytics

One of the major challenges of the Big Data era is the need to support Pechlivanoglou, and Manos Papagelis introduces a novel heat-map
the analysis of a great amount and variety of massive datasets by non- visualization, termed OL-HeatMap, designed for identifying and dis­
corporate data analysts with little or no support and expertise on data playing the precise density of overlapping rectangles. The proposed
processing such as research scientists, data journalists, policy makers, approach not only reveals valuable insights, such as the actual position
SMEs and individuals. These datasets are made accessible in a raw and size of the formed overlapping rectangle. The authors exploit a state-
format (e.g., plain text, json) and are not being loaded or indexed in a of-the-art computational geometry method based on the sweep-line
database. Furthermore, they are dynamic, noisy, and heterogeneous in paradigm. The adopted sweep-line algorithm enables the implementa­
nature. The level of difficulty in transforming a data-curious user into tion of fast and exact density-based visualization Furthermore, the au­
someone who can access and analyze this data is even more burden­ thors present two evaluation metrics that quantitatively assess the
some. The purpose of visual data exploration is to facilitate information accuracy of grid-based overlap visualizations. An extensive evaluation,
perception and manipulation, knowledge extraction and inference by incorporating both synthetic and real datasets, showcases the accuracy
non-expert users. The visualization techniques, used in a variety of and efficiency of the proposed methods.
modern systems, provide users with intuitive means to interactively In "ViewSeeker: An Interactive View Recommendation Framework",
explore the content of the data, identify interesting patterns, infer cor­ Xiaozhong Zhang, Xiaoyu Ge, Panos K. Chrysanthis, and Mohamed A.
relations and causalities, and support sense-making activities that are Sharaf explore the problem of context-aware view recommendation by
not always possible with traditional data traditional data analysis introducing an Interactive View Recommendation framework named
techniques. ViewSeeker. The framework engages in a human-in-the-loop approach,
In the Big Data era, several challenges arise in the field of data interacting with the user to identify the most suitable utility function for
visualization and analytics. First, the modern exploration and visuali­ the current analysis context. ViewSeeker provides two forms of adap­
zation systems should offer scalable data management techniques in tation: utility function tuning, constituting its initial and optional phase,
order to efficiently handle billion objects datasets, limiting the system and utility function integration, constituting its subsequent phase. In the
response in a few milliseconds. Besides, nowadays systems must address utility function tuning phase, ViewSeeker collaborates with the user to
the challenge of on-the-fly scalable visualizations over large and dy­ fine-tune the parameters of a utility function. This ensures that the pa­
namic sets of volatile raw data, offering efficient interactive exploration rameters accurately capture the data analysis context. The utility func­
techniques, as well as mechanisms for information abstraction, sampling tion integration phase employs active-learning techniques to select
and summarization for addressing problems related to visual informa­ informative example views for labeling and to predict the contribution
tion overplotting. Further, they must encourage user comprehension by of each utility function within a multi-objective utility function. Exten­
offering customization capabilities to different user-defined exploration sive experiments utilizing two real-world datasets have been conducted,
scenarios and preferences according to the analysis needs. Overall, the demonstrating the effectiveness and efficiency of the adaptation
challenge is to enable users to gain value and insights out of the data as methods.
rapidly as possible, minimizing the role of IT-expert in the loop. In “ExplorerTree: A Focus+Context Exploration Approach for 2D
This special issue aimed to publish work on multidisciplinary Embeddings”, Wilson E. Marcílio-Jr, Danilo M. Eler, Fernando V. Pau­
research areas spanning from Data Management and Mining to Infor­ lovich, José F. Rodrigues-Jr, and Almir O. Artero Artero introduces a
mation Visualization and Human-Computer Interaction. In addition to hierarchical exploration approach known as ExplorerTree. This meth­
the normal submissions, this special issue considered to invite some of odology tackles visual scalability challenges encountered when pre­
the best papers from the 3rd International Workshop on Big Data Visual senting dimensionality reduction results through scatter plots. The
Exploration and Analytics (BigVis 2020), held in conjunction with the approach adopts a hierarchical exploration paradigm enriched with
23rd International Conference on Extending Database Technology & Focus+context interaction. To mitigate visual clutter, the construction
23rd International Conference on Database Theory (EDBT/ICDT 2020). of the hierarchy incorporates a sampling selection algorithm and
In response to the call for papers, 22 submissions on different ap­ encoding strategies. The application of this approach in exploring im­
plications of visual data analysis techniques were received. From these, ages of convolutional filters, along with a user evaluation, illustrates the
9 articles were accepted after a two-stage review process supported by a effectiveness of the proposed methodology.
reviewer board of internationally renowned experts in the field. In "Intelligent Narrative Summaries: From Indicative to Informative
The article " OL-HeatMap: Effective Density Visualization of Multiple Summarization", Samira Ghodratnama, Amin Beheshti, Mehrdad
Overlapping Rectangles", by Niloy Eric Costa, Tilemachos Zakershahrak, and Fariborz Sobhanmanesh address introduce a generic

https://doi.org/10.1016/j.bdr.2024.100445

Available online 14 February 2024

2214-5796/© 2024 Published by Elsevier Inc.
D. Auber et al. Big Data Research 36 (2024) 100445

hierarchical personalized summarization framework named Narrative functional dependencies. Although many algorithms for discovering FDs
Summaries, empowering users to express preferences, resulting in the from static datasets, the article contributes to methods that continuously
generation of user-specific hierarchical summaries. The framework in­ monitor FDs that evolve in data streams. It presents a tool that allows
corporates two models: a semi-structured summarization approach and users to visualize FDs, explore and compare results based on their types,
a fully-structured summarization approach. In the semi-structured and employ quantitative measures to monitor how discovery results
summarization, the basic unit of representation is a sentence. The evolve. A user study has been conducted to assess the effectiveness of the
model employs objective functions to cluster sentences coherently and proposed visualization tool.
in a logical order. Furthermore, an objective function is applied during In "Graph Waves", James Abello and Daniel Nakhimovich address
the summarization phase to ensure that cluster summaries for a given problems related to the visualization of very large and complex net­
hierarchy level are logically distinct and also fit within a user-defined works. They present efficient algorithmic mechanisms to partition very
budget size. In the fully-structured summarization model, the basic large graphs into subgraphs for generating intuitive and interactive vi­
unit of representation is a concept. The model employs an objective sualizations of the meta-structure of complex networks. To achieve
function to hierarchically cluster concepts at different levels of detail. interactivity, a graph is processed into graph layers, called fixed points,
The proposed methods are assessed in comparison to state-of-the-art which are further decomposed into visual representations called graph
approaches using a news articles dataset. The results demonstrate that waves and fragments. This decomposition is used to create spanning
the generated summaries enhance users’ comprehension of the topic. views of fixed points in a graph. They provide illustrative examples on
In "Scaling the Growing Neural Gas for Visual Cluster Analysis", Elio publicly available data sets including social, web, and citation networks.
Ventocilla, Rafael M. Martins, Fernando Paulovich, and Maria Riveiro In "An Analytic Graph Data Model and Query Language for Exploring
present two methods that enhance growing neural gas algorithms within the Evolution of Science", Ke Li, Hubert Naacke and Bernd Amann
the context of visualizing clusters in large-scale, high-dimensional introduce a data model (called pivot topic graph) and a query language
datasets. The first method tackles overplotting and clutter issues by for the visualization and exploration of topic evolution networks rep­
avoiding connections that result in high-dimensional graphs. Conse­ resenting the research progress in scientific document archives. The
quently, the generated visualization becomes more accurate and model is independent of a particular topic extraction and alignment
meaningful. The second method addresses performance concerns related method and proposes a set of semantic and structural metrics for char­
to the time required to generate the results. This method parallelizes the acterizing and filtering meaningful topic evolution patterns. These
process by modeling and merging different parts of a dataset using the metrics are particularly useful for the visualization and the exploration
MapReduce model. The proposed methods are evaluated across nine of large topic evolution graphs. They also present a first implementation
datasets, utilizing various quantitative and qualitative metrics. The ex­ of their model on top of Apache Spark and experimental results obtained
periments reveal that the first method generates lower-dimensional from four real-world document archives.
graphs with reduced overplotting and clutter. The second method pre­ In closing, we would like to thank all authors who have submitted
serves visual quality while requiring less execution time. their work to this special issue, and the authors of the accepted papers
In "Visual Exploration of Anomalies in Cyclic Time Series Data with for their timely revisions. We sincerely thank the Editors-in-Chief of the
Matrix and Glyph Representations", Josef Suschniggab, Belgin Mutluac, Journal of Big Data Research, Themis Palpanas and Zhaohui Wu, for
Georgios Koutroulisa, Vedran Sabolc, Stefan Thalmannd, and Tobias hosting this special issue and supporting us during all stages. Addi­
Schrecke present an interactive glyph-based visual analytics approach tionally, we are very grateful to all reviewers for their time and valuable
for anomaly detection in multivariate time series. This approach con­ constructive comments.
siders sensor data collected by durability tests performed in industrial
settings (e.g. automotive sector) and displays the iterations of such tests David Aubera, Nikos Bikakisb,*, Panos K. Chrysanthisc,
as a collection of color-encoded cycle glyphs, which enable users to George Papastefanatosd, Mohamed Sharafe
visually identify conspicuous data through a matrix representation for a
University Bordeaux, France
drill-down and further comparative analysis. The article provides results b
Hellenic Mediterranean University, Greece
and discussion of a pair analytics evaluation, which has been conducted c
University of Pittsburgh, USA
in collaboration with the target user group and preliminary results on a d
ATHENA Research Center, Greece
visual interactive labelling concept for anomaly classification. e
United Arab Emirates University, UAE
In "Dependency Visualization in Data Stream Profiling", Bernardo
Breve, Loredana Caruccio, Stefano Cirillo, Vincenzo Deufemia, and *
Corresponding author.
Giuseppe Polese provide methods for the automatic detection and E-mail address: [email protected] (N. Bikakis).
visualization of functional dependencies (FDs) in streaming data and
more specifically functional dependencies and their extensions, relaxed