IEEE Systems, Man and Cybernetics Magazine - April 2020 - 3

Editorial

by Saeid Nahavandi

A Cluster
of Thoughts

Data! Data! Data! I can't make
bricks without clay.
-Sherlock Holmes

e are living in the era of
big data, where high-variety data sets are generated
in large volumes with high frequency. State-of-the-art machine learning
and artificial intelligence techniques,
such as deep learning, can then be
used with these big data sets to develop predictive models and decisionmaking tools. Only the data with
high veracity must be processed and
used to generate insights and make
decisions. Exploratory data analysis
(EDA) is a crucial step before statistical and machine learning techniques
are applied for model development. By
preprocessing, transforming, and visualizing the data, the analyst becomes
intimately familiar with the data sets,
making EDA the first step in answering research questions.
EDA, at the high level, is the innovative practice of applying quantitative
and visual techniques to understand a
data set and build an intuition for its
structure. In reality, it is a prerequisite
for developing solid, reliable, and parsimonious predictive models. There
are many techniques supporting EDA
efforts, such as numeric summaries,
finding missing values, aggregations,
distributions, dimensionality reduction, correlations, and visualization
(univariate, bivariate, and multivariate plots).
A frequently applied method for performing EDA and gaining deep familiarity with the data is clustering, also known
as unsupervised classification. From a
non--t echnical perspective, the purpose of clustering is to group similar
objects into the same cluster. Many
Digital Object Identifier 10.1109/MSMC.2020.2973436
Date of current version: 20 April 2020

The a r ticle by
clustering techniques
Dheeraj Kumar and
have been proposed
Ja mes C. Bezdek,
over the past few decades. Beginning with
-"Visual Approaches
k-nearest -n eighbors,
for Exploratory Data
Analysis," provides
clustering methods
have advanced into
a comprehensive sur--
state-of-the-art artifivey of the visual ascial intelligence techsessment of tendenniques, such as fuzzy
cy (VAT) algorithm,
logic and deep learnwhich deals with a
ing. Despite decades
key question regardof progress, however,
ing the existence and
Saeid Nahavandi
there are still fundaquantity of data clusmental issues with how clustering
ters. The survey also provides demethods are applied for EDA and decitailed information on how VAT can
sion making.
be extended and applied to big data
In this issue of IEEE Systems,
problems. The article by Ludmila
Man, a n d C y b e r n e t i c s Magazine,
I. Kuncheva, "Prototy pe Classifiwe include a number of articles providers and the Big Fish," deals with the
ing insights into the history and
problem of prototy pe (instance)
development of cluss e l e c t i o n for detering techniques. The
velopi ng cla s s i article by Thomas
fiers (cluster ing
From a nontechniA. Runkler, "Sequenbased on a distance
tial Cluster Estimametric). It provides
cal perspective,
tion," proposes a
a historic a l v i e w
the purpose of
novel method as an
o n pr og ress made
clustering is to
extension of alterwithin the last three
group similar
nating cluster estide c a de s and dis--
mation by finding
cusses how these
objects into the
clusters sequentialoriginal metho d s
same cluster.
ly. Specifically, it inhave evolved through
troduces sequential
the time. The main
cones and Poisson
conclusion is that
c lu s t e r i n g methods, which are
random search methods achieve
shown to outperform nonsequential
the best results fo r t h e select ion
clustering methods for data with
of i n st a nce s to sat i sf y p e r fo rmany clusters.
mance criteria.

Ap ri l 2020

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE

IEEE Systems, Man and Cybernetics Magazine - April 2020

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - April 2020