52 ISE Magazine | www.iise.org/ISEmagazine
Inside IISE Journals
research
What it takes to inform us about
lead contamination in water: A tale
of sensor technology and data science
Graphene field-effect transistors (GFETs) are promising
nanosensors that can detect a designated target substance,
such as lead contamination in water. A GFET reads the cur-
rents, which change when the GFET is exposed to the dif-
ferent amounts of the target substance.
In principle, one may estimate the substance amount us-
ing a transfer curve, which is the functional relationship
between the currents and the corresponding voltages. In
reality, it is not that simple, because GFETs have significant
device-to-device variations in their physical structures.
Moreover, GFETs are disposable and can only be used
once. As a result, the transfer curves obtained from the used
GFETs do not lend enough confidence when new GFETs
are used to estimate the contamination in water.
In their paper, “Landmark-Embedded Gaussian Process
with Applications for Functional Data Modeling,” assistant
professor Jaesung Lee from Texas A&M University, assis-
tant professor Chao Wang from the University of Iowa,
postdoctoral researcher Xiaoyu Sui from Northwestern
University, professor Shiyu Zhou from the University of
Wisconsin-Madison and professor Junhong Chen from the
University of Chicago and Argonne National Laboratory
propose a nonparametric statistical model that accounts for
the special random features of functional data (i.e., transfer
curve). Their method focuses on extracting the common
features in the curves’ shapes and locations that are obtained
from the same amount of the target substance. Modeling
the common shape and location features in the functional
data provides the key to unlock the riddle.
This month we highlight two articles in IISE Transactions. The first article tells a story that combines an emerging sensor
technology and some novel data science methodology for detecting with more accuracy and robustness the lead
contamination in water. A new nanosensor, i.e., the Graphene field-effect transistor, has the potential to detect a small
trace of lead in water. But to reach that potential, the authors pair with the sensor a two-layer Gaussian process model that
enables a good degree of transfer learning between old sensors and new ones and bridges the gap for the training data
to be used effectively. The second article looks at system reliability under performance sharing mechanism. Performance
sharing means that the performance surplus of some subsystems can be transmitted to compensate the reliability deficit of
other subsystems. To make performance sharing work, the authors make use of ISE methodologies like the finite Markov
chain imbedding approach and universal generating function technique. The insights from the study can guide engineers
in optimizing system design under cost constraints. These articles will appear in the November 2022 issue of IISE
Transactions (Volume 54, No. 11).
Jaesung Lee Chao Wang Xiaoyu Sui Shiyu Zhou Junhong Chen
October 2022 | ISE Magazine 53
Toward that objective, the authors established two lay-
ers of Gaussian processes that model the location features
and the shape features, respectively. As such, the two-layer
Gaussian processes characterize both random features of the
transfer curves in a unified model, thus enabling an accu-
rate inference of the target substance amount. The superi-
ority of the method is demonstrated in both numerical and
case studies using real GFET measurements. The two-layer
Gaussian process model is very flexible and can be used in
a broad range of applications beyond the detection of water
contamination.
CONTACT: Jaesung Lee; j.lee@tamu.edu; Wm Michael Barnes ’64 De-
partment of Industrial and Systems Engineering, Texas A&M University;
3131 TAMU, College Station, TX 77843-3131
Better system reliability
by performance sharing
A performance-sharing system consists of many compo-
nents or subsystems that perform to satisfy their own de-
mands, while their performance surplus can be transmitted
to compensate the reliability deficit of other components
or subsystems in the need to enhance the overall system
reliability.
Such a performance-sharing mechanism can be found in
power distribution systems and distributed computing sys-
tems. This kind of engineering system is usually operated
under a complicated environment in which external shocks
and a systemwide failure will cause enormous economic
losses and calamitous consequences. For this reason, model-
ing the operation of the performance-sharing systems and
evaluating its reliability under a shock environment require
a rigorous treatment.
In their work, “Reliability Assessment of Multi-State
Performance Sharing Systems With Transmission Loss and
Random Shocks,” doctoral student Congshan Wu from
Beijing Institute of Technology, associate professor Rong
Pan from Arizona State University and professor Xian
Zhao from Beijing Institute of Technology constructed a
reliability assessment model for performance-sharing engi-
neering systems. In their model, the system consists of mul-
tiple components that generate
a range of performances while
bearing random demand loads.
Performance can be shared
among all components in the
system via a common bus; the
goal is to satisfy each compo-
nents demand as much as pos-
sible. External shock processes
and transmission losses are
considered in the context of
actual engineering situations.
The authors evaluated system reliability by employing a
nite Markov chain imbedding approach and the univer-
sal generating function technique. They further conducted
sensitivity analyses to determine the key factors of system
reliability and presented a numerical example to demon-
strate the applicability of proposed approach.
The numerical results show that, to increase system reli-
ability, engineers can raise the performance level of each
component, expand the capacity of common bus and re-
duce transmission loss, or carry out maintenance activities.
These insights can be used to guide engineers in optimizing
system design under cost constraints.
CONTACT: Rong Pan; Rong.Pan@asu.edu; School of Computing and
Augmented Intelligence, Arizona State University, Tempe, AZ 85281
Machine learning tools assess
corporate supply chain strategy
from financial statement data
Evaluating a company’s supply chain strategy and finan-
cial performance pattern is a critical step for decisions such
as choosing supply chain partners. However, assessing a large
number of candidate companies can be a daunting task. Exist-
Rong Pan Xian Zhao
Congshan Wu
This month we highlight two articles from the
September 2022 issue of The Engineering
Economist (Volume 67, No. 3). In the first, the
author implements machine learning on target
companies’ financial performance patterns to
study their underlying supply chain strategies.
The approach produced a clear strategy map
to facilitate the supply chain partner selection
process. In the second article, the authors sought
to develop a model to construct a partial index
tracking portfolio for stocks. They proposed
stochastic dominance methods to define sequential
returns, classify the efficient stocks and facilitate
the decision in the process.
54 ISE Magazine | www.iise.org/ISEmagazine
research
ing approaches often demand input data that are not easily ac-
cessible. Moreover, in-depth analyses might not be necessary
for a first-round screening.
Thus, there is a need for a tool that can efficiently scan many
companies for easily accessible data and does not require deep
domain knowledge to operate. More importantly, the result
must be clearly comprehensible to facilitate the decision pro-
cess.
In terms of sourcing corporate data, the most accessible
source is the three financial statements companies release each
year. This raises three key questions:
1. Can the data be used not only to assess a company’s financial
performance pattern but also to decipher from the pattern
its supply chain strategy?
2. Among the many financial variables, which ones are salient
for the assessment process?
3. How could the result extracted from the multi-year multi-
variate financial time series be presented in an intelligible
way to users?
In the article, “On a Holistic View of Supply Chain Finan-
cial Performance and Strategic Position,” a machine learning
based approach takes the financial time series as input to cre-
ate a 2D supply chain strategy map on which each company
is represented by a point. The map renders a comprehensible
view on two aggregate measures and groups all companies
into several supply chain strategy clusters. Companies close to
each other on the map intuitively share a similar financial per-
formance pattern and supply chain strategy.
The approach relies on three machine learning modules. A
forecasting module smooths the time series to reduce noises,
a clustering module groups companies into supply chain strat-
egy clusters and a classification module helps select the feature
variables and check the bias-variance balance. Given a data
set of standard financial variables, this approach can select the
most relevant to create clearly distinguishable clusters of supply
chain strategies.
When tested on all U.S. manufacturers and traders listed on
the New York Stock Exchange and NASDAQ, the approach
produced a clear strategy map
with three clusters based on
only seven selected variables
found to be the most relevant.
Two of the clusters fit the
characteristics of Fishers ef-
ficient and responsive supply
chain strategies respectively,
and the third cluster is in-be-
tween the first two.
CONTACT: Chih-Yang Tsai; tsaic@
newpaltz.edu; Professor, School of
Business, State University of New York at New Paltz, 1 Hawk Drive, New Paltz,
NY 12561-2443
Developing modeling index tracking
portfolio based on stochastic
dominance for stock selection
How do we determine the number and candidate stocks in
the portfolio? We are looking for a model or method to ef-
ficiently construct a partial tracking portfolio and replicate
the return of the index. Determining the number of stocks,
selecting the stocks using sequence techniques and setting the
precise weights depending on the desired return or tracking
error are three main issues to establish one’s own tracking in-
dex portfolio.
Closely connected to the expected utility theory without
any restricted assumptions on return distributions, the sto-
chastic dominance methods are proposed to define sequential
returns, classify the efficient stocks and facilitate the decision
in the process. The optimal rule of the first-degree stochastic
dominance approach satisfies the investor’s wealth preference;
the second-degree stochastic dominance approach for all risk
averters for whom more is better and who are averse to risk;
and the third-degree stochastic dominance approach is for all
increasing risk averters, risk averters and wealth seekers. After
selecting the number and candidate stocks, the issue of weight-
ing is next for modeling an index tracking portfolio.
Portfolio models based on robust optimization techniques
have become a focal point for many researchers to overcome
weighting difficulties. Sequential least squares programming
was launched, minimizing the tracking error object and no
stock shorting constraints for each testing period to iterate the
optimal solution.
Authors Liangchuan Wu, Yuju Wang and Liang-Hong Wu
addressed this in their paper, “Modeling Index Tracking Port-
folio Based on Stochastic Dominance for Stock Selection.
The empirical study presented statistical tests on the perfor-
mance of the 40 scenarios in the testing periods with the con-
trast exchange-traded funds (HSBC FTSE 100 UCITS ETF
GBP) and traditional mean-variance portfolio. The mean-
variance portfolio is constructed in the same way as the pro-
posed stochastic dominance
portfolio, except that the ob-
jective function is substituted
from tracking error to in-
formation ratio for different
goals. The results show that
the model with stochastic
dominance methods can cre-
ate an index tracking portfo-
lio with low violations under
performance measures.
Empirical results indicate
Liangchuan Wu
Chih-Yang Tsai
October 2022 | ISE Magazine 55
that the proposed model precisely replicates the return of the
index and benefits investment managers who no longer need
to determine the exact utility function and can therefore re-
duce transaction costs. This model also supports investor de-
cision-making on the number of stocks in their portfolio and
the efficient stocks to select. Furthermore, the new model is
suggested to augment the index-tracking field using the sto-
chastic dominance approaches in portfolio selection and ap-
plied in real-life markets.
CONTACT: Yuju Wang; tammywang@smail.nchu.edu.tw; Institute of Technol-
ogy Management, National Chung Hsing University, No.250 Kuo-Kuang
Road, Taichung 402, Taiwan, R.O.C.
Yu Ding is the Mike and Sugar Barnes Professor of Industrial and
Systems Engineering at Texas A&M University and Associate Direc-
tor for Research Engagement at the Texas A&M Institute of Data Sci-
ence. He is editor-in-chief of IISE Transactions and a fellow of IISE.
Heather Nachtmann is the Earl J. and Lillian P. Dyess Endowed
Chair in Engineering and a professor of industrial engineering at
the University of Arkansas. She is editor-in-chief of The Engi-
neering Economist and a fellow of IISE.
The IISE Foundation and the IISE Scholarship Fund
offer you the chance to donate and help build
future generations of ISEs.
From the prestigious Frank & Lillian Gilbreth Award
to the $90,000-plus in scholarships bestowed in
2022, your major bequest, annual contribution,
one-time gift – or a combo of all three – are
tax-deductible opportunities to give a helping hand
to the ISEs who will continue your legacy.
Visit iise.org/PlannedGiving
For confidential assistance concerning donations
and scholarships, contact Donna Calvert at
dcalvert@iise.org or (770) 349-1108.
BUILD YOUR
LEGACY
IISE Transactions (link.iise.org/iisetransactions) is IISE’s flagship
research journal and is published monthly. It aims to foster exchange
among researchers and practitioners in the industrial engineering
community by publishing papers that are grounded in science and
mathematics and motivated by engineering applications.
The Engineering Economist (link.iise.org/engineeringeconomist) is a
quarterly refereed journal published jointly by IISE and the American
Society of Engineering Education. Devoted to issues of capital
investment, its topics include economic decision analysis, capital
investment analysis, research and development decisions, cost
estimating and accounting, and public policy analysis.
About the journals
Yuju Wang Liang-Hong Wu