Studies in Informatics and Control, 30(1) 5-18, March 2021 ISSN: 1220-1766 eISSN: 1841-429X

An Overview of Cloud-Based Decision Support

Applications and a Reference Model
Shashidhar KAPARTHI*, Arti MANN, Daniel J. POWER
University of Northern Iowa, College of Business Administration, Cedar Falls, IA 50614, USA
[email protected] (*Corresponding author), [email protected], [email protected]
Abstract: Major shifts have occurred in the past 70 years in the deployment of decision support applications. Now it is possible
to deploy innovative and traditional Decision Support Systems (DSS) and decision algorithms using Cloud architectures and
technologies. Cloud deployment with hand-held, mobile-device platforms is encouraging decision support experimentation
and innovation. This survey article emphasizes the-how, especially the architecture, the-what, the-who, the-when, and the-
why of building Cloud-based decision support and decision automation. To assist in conceptualizing cloud-based decision
support and analytics, and to facilitate easier understanding and communication, a hierarchical reference model is presented
using multiple perspectives. Additionally, the proposed decision support reference architecture model is used to assess cloud-
based decision support systems from the prior literature.
Keywords: Business analytics, Cloud computing architecture, Cloud-based decision support reference model, Decision
support systems.

1. Introduction
Cloud technology has evolved from distributed, characteristics, and user characteristics. Given the
grid, and utility computing. Cloud computing is complexity associated with the above-mentioned
a distributed computing paradigm that enables domains, conceptualizing and understanding
access to virtualized resources and tools including cloud-based decision support and analytics
computers, networks, storage, development are important tasks. This article proposes a
platforms, and applications (Mell & Grace, 2009). reference model that uses a “divide and conquer”
Cloud computing is one of the fastest-growing philosophy. Multiple perspectives or lenses are
innovative technologies and it is predicted that applied to create the proposed reference model.
the worldwide public cloud service market Cloud-based decision support and analytics are
will grow from $182.4B in 2018 to $331.2B in examined using an actor lens, a primary driver
2022, attaining a compound annual growth rate lens, and an infrastructure lens. Using multiple
(CAGR) of 12.6% (Columbus, 2019). Under perspectives allows deeper understanding and
the umbrella of cloud computing, the three most facilitates future innovative and creative decision
common service models are Infrastructure as a support uses.
Service (IaaS), Platform as a Service (PaaS), and
Software as a Service (SaaS). These three service The remainder of this article is organized as
models offer a number of potential advantages follows: Section 2 summarizes background on
including reduced costs, omnipresent services, decision support architectures. Section 3 discusses
collaborative support, on-demand access to the literature review methodology, defines
computing resources, and simplified operations cloud-based decision support, and discusses
(Armburst et al., 2010). The potential value of the cloud-based decision support and analytics
cloud has encouraged firms to deploy a variety of architectures in the context of the prior literature.
cloud-based services supporting novel business Section 4 lays out the different perspectives or
models. Today, implementing cloud-based lenses and presents the proposed reference model.
computing is one of the leading strategies for Section 5 categorizes the prior literature based on
creating competitive differentiation of a firm. the proposed reference model and Section 6 draws
major conclusions.
Cloud-based decision support and analytics have
been used in several domains including financial 2. Background on Decision
services, e-business, and supply chains. Within Support Architectures
these domains, there is a large variation in the
nature and type of the provided decision support, During the past several decades, the cost of
and significant variations in data collection computer and information technology has
techniques, data collection devices, modelling decreased exponentially. Data storage cost per
techniques, infrastructure, output devices, output GB has decreased from over $1 million per GB

https://doi.org/10.24846/v30i1y202101 ICI Bucharest © Copyright 2012-2021. All rights reserved

6 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

in 1966 to $0.02 per GB in 2017 (Lucas, 2017). dialogue management, database management,
Further, the cost of bandwidth has gone down from model base management, and DSS architecture.
$9.01 per Mbs in 2006 to $0.89 per Mbs in 2016 They identified four architectures: 1) DSS
(NCTA, 2019). During this time frame computer Network, 2) DSS Bridge, 3) DSS Sandwich, and
systems evolved from stand-alone mainframe 4) DSS Tower. The DSS Network has multiple
computers to personal computers, to networked dialogue, modelling and database components
computers, to inter-networked computers, and that are interconnected. The Bridge has a standard
mobile devices. Mainframe computers also interface with local dialogue and modelling
evolved into client-server systems, then web- components that link to remote modelling and
based systems, and now cloud-based systems. database components. A Sandwich architecture
has a single dialogue and database component,
Computing evolution has resulted in several but with multiple model components linked by
generations of architectures for supporting the architecture. Finally, the Tower includes more
decision-making and analytics in organizations. An vertical components or tiers with data extraction
architecture for a computerized decision support tools integrating diverse data base components.
system documents the plan for deploying the The rest of the Tower architecture is similar to the
components of the envisioned DSS. These include Network architecture structure.
Mainframe, Time-sharing, Distributed, Web-based,
Mobile-based, and Cloud-based architectures. Generation 4: Client Server/Web-based
Table 1 summarizes different generations of
architectures identified by Power (2018). Power & Kaparthi (1998) identified six DSS
architectures subtypes: 1) distributed dialogue,
Table 1. Decision Support Architecture Generations 2) remote dialogue, 3) distributed model, 4)
Generation Category distributed data, 5) remote data, and 6) stand-
1.0 Mainframe alone. Distributed dialogue is basically a thin-
2.0 Time-Sharing client web architecture. The remote dialog is a
Distributed: network, bridge, sandwich, more traditional thick-client application. In the
3.0 distributed model, the application software on
and tower
Client Server/Web-based: distributed the client expands, and model capabilities are
4.0
dialogue, remote dialogue, distributed distributed. The distributed data architecture
model, distributed data, remote data, and requires accessing data across the network. With a
stand-alone
remote data architecture, some data is downloaded
4.5 Mobile-based
to the client for faster processing. Finally, a stand-
5.0 Cloud-based
alone architecture has the entire DSS application
on a stand-alone computer with no provision for
Generation 1: Mainframe
network access to server-based components.
Mainframe hardware was rudimentary using
Generation 4.5: Mobile
vacuum tubes, the operating systems were
elementary, the software was specialized, According to Power (2013), mobile decision-
and decision support was included as part of support uses wireless technology and portable
operational tasks like payroll, and users needed computing devices to provide interactive
technical expertise. computer-based systems to help decision-
Generation 2: Time-Sharing makers use data, documents, knowledge, and
models to identify and solve problems, complete
Time-Sharing was a multi-programming and decision process tasks, and make decisions.
multitasking architecture that provided a Mobile decision support also includes alerts and
computing resource to many users concurrently. information pushed to a user based on location.
This architecture was used for many tasks
including decision support in the 1960s and 1970s. Generation 5: Cloud

Generation 3: Distributed Cloud-based DSS uses a cloud computing

architecture to deliver decision support
Sprague & Carlson (1982) defined four applications (Power, 2012). A well-designed cloud
components of a DSS technology framework as that uses redundancy can be reliable, secure, and

https://www.sic.ici.ro
 An Overview of Cloud-Based Decision Support Applications and a Reference Model 7

suitable for business applications. Using a cloud titles that included the 3 terms “cloud, decision,
to deliver decision support can reduce capital and support. In addition to relevant articles for
expenses and operating costs (Irvine, 2012). this study, the search also resulted in articles
Other benefits claimed are agility, accessibility to that were describing decision support needed
software and device and location independence. for migration to cloud computing environments.
After removing such articles from further
3. Literature Review analysis and tabulation, there were a total of 30
articles relevant to this analysis. Table 2 below
A search in the ABI/INFORM database of summarizes the reviewed articles.
periodicals was conducted for all the articles with

Table 2. Articles Reviewed

Notation: Cloud-based Decision Support provided: D: Data-driven; M: Model-driven; C: Communication-
driven; K: Knowledge-driven; O: Document-driven

Authors D M C K O Domain/ Environment

1 Jayalakshmi & Gomathi (2018) x Healthcare
2 Mercier-Laurent et al. (2018) x Industrial Design and Energy
3 Wan et al. (2018) x Material Handling
4 Rupnik et al. (2019) x x Agriculture
5 Heilig et al. (2017) x Logistics
6 Singh (2017) x Water Management
7 Eskelinen et al. (2017) x x x Environmental Monitoring
8 Hill & Böse (2017) x Transportation
9 Bock et al. (2017) x Search & Rescue
10 Choi & Lambert (2017) x Risk Analysis
11 Gupta et al. (2017) x Healthcare
12 Helo & Hao (2017) x x Manufacturing
13 Lee (2017) x x x Retail
14 Kumar et al. (2016) x Manufacturing
15 Chapron (2015) x Wildlife Management
16 Lee et al. (2016) x x Franchise Inventory
17 Lauras et al. (2015) x Crisis Management
18 Depeige & Doyencourt (2015) x x General
19 Mackay et al. (2015) x x x Environmental Policy
20 Benatia et al. (2015) x x Urban Planning
21 O’Flaherty & Heavin (2015) x x Consumer Behavior
22 Guo et al. (2015) x Manufacturing
23 Chichernea (2014) x x Smart City Management
24 Bara et al. (2014) x x Energy
25 Tsiligiridis et al. (2014) x x Pest Control
26 Peng & Shiang (2014) x Sustainable Supply Chain
27 Guo et al. (2014) x Manufacturing
28 Thomas et al. (2013) x Hospital Administration
29 Demirkan & Delen (2013) x x x x x General
30 Bellini et al. (2012) x Healthcare

ICI Bucharest © Copyright 2012-2021. All rights reserved

8 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

Next, the articles were read and categorized by rescue scenarios, that supports path planning in
content. The four article categories are 1) real post-disaster structures. Their system utilizes point
time data focused, 2) visualization based, 3) end cloud data acquired from unmanned robots and
user focused, and 4) components of cloud-based generates assessments of automatically generated
architectures. Some of the major articles in each paths. As data uncertainty and a priori unknown
category are summarized below. information make fully automated systems
impractical, they present the incident commanders
3.1 Real Time Data Focused with a set of viable access paths, based on varying
risk factors, in a 3D environment combined with
Eskelinen et al. (2017) proposed a system where visual analysis tools enabling informed decision-
modern sensor technologies with connections to making and trade-offs. They describe visualization
cloud-based data collection services were used design considerations for the system and
for monitoring and collecting environmental data. decision support.
They applied a four-stage innovation process for
industry, which included context definition, idea 3.3 End User Focused
generation, and selection, and produced multi-
criteria decision support data to help the design Chapron (2015) argues that research in wildlife
of the business model. According to the authors, management increasingly relies on quantitative
environmental monitoring typically combines population models but it is a challenge to have
many data sampling sources (e.g., from water, air, end-users benefit from this research. The author
or soil) with multiple methods of measurement. then proposes a new approach, ’wildlife in the
Continuous monitoring has made it possible to cloud,’ to enable active learning by practitioners
collect data almost constantly in real-time, with from cloud-based ecological models whose
multiple variables being measured simultaneously complexity remains invisible to the user. This
and stored on cloud-based infrastructure. concept is illustrated by presenting an online
decision-support tool for moose management in
Guo et al. (2014) address the order tracking and areas with predators in Sweden. A cloud-app is
allocation problem in an apparel manufacturing designed through which users can compare the
environment with multiple plants. A cloud-based effects of alternative management decisions that
intelligent decision system was developed to tackle can be used for adjusting their hunting strategy.
this problem and track remote production orders. It Further, the dynamic nature of cloud-apps opens
employed computational intelligence techniques to the door to different ways of learning, informed
generate effective order allocation solutions each by ecological models from which both the users
plant. In their system, radio frequency identification and the researchers can benefit.
and cloud computing technologies were used to
capture real-time production records. The system O’Flaherty & Heavin (2015) report a design for a
implementation resulted in reductions in production predictive decision support system for customer
retention. They argue that the power of propensity
costs and increases in production efficiency.
scoring based on historical data delivered as
Thomas et al. (2013) implemented and hosted a service via cloud computing heralds the
a prototype real-time bed-assignment solution availability of capabilities that previously required
as a cloud-based application for Mount Sinai teams of quantitative specialists to deliver. This
Medical Centre in New York. The authors argue experimental predictive analytics platform
that bed management is an important function of was developed in conjunction with a statistical
any hospital and has a major impact on patient software company. In their system, analysis of
care, patient flow, patient, and staff satisfaction, historical data gives an indication of the predicted
and ultimately on the hospital’s operating margin. future behaviour of magazine subscribers. The key
Further, a key challenge in bed management is stakeholders who could use the system include
optimizing the bed-assignment process. the retention manager, the data administrator,
and the account manager. The retention manager
3.2 Visualization Based develops the company’s overall retention strategy
with the support provided by the system while
Bock et al. (2017) proposed a visualization system working closely with the account managers, who
for incident commanders in urban search and are responsible for implementing the strategy.

https://www.sic.ici.ro
 An Overview of Cloud-Based Decision Support Applications and a Reference Model 9

3.4 Components of Cloud-Based search and rescue operations, wildlife management,

Architectures consumer/buyer behavioural prediction,
manufacturing, and hospital administration. The data
Russell et al. (2010) develop a mathematical model collection techniques, the data collection devices,
for assessing the availability of a cloud-based the modelling techniques, the infrastructure, the
DSS. They draw distinctions between decision output devices, the output characteristics, the users,
resource availability and system reliability. They the interactions and the nature of decision support
conceptualize a hierarchical relationship between are varied too. Given such complexity, a systematic
several resources needed like data from some reference model is needed for conceptualizing and
storage-system, a network communication medium, understanding these systems. The major focus of
an output device, collaborating system users, or even this article is to propose an understandable reference
decision-makers themselves. They view a decision architecture model with commonly occurring
resource as a hierarchical structure of interacting elements and capabilities.
functions. This hierarchical structure may have
multiple levels and when the resource is a computing 4. Proposed Reference Model for
device, these levels would encompass hardware, Cloud-Based Decision Support
firmware, and software. The underlying concept is and Analytics
that a decision-resource, via its hierarchy, relates to
a number of sub-resources at lower levels of the Previous examples of systems and architectures
hierarchy. The functions in a resource hierarchy demonstrate the complexity of conceptualizing
act and interact together to provide the resource and understanding cloud-based decision support
at the top of the hierarchy. Upper-level functions and analytics. To enable the conceptualization of
depend on lower levels to ensure their reliability and such a complex phenomenon, a reference model
availability. They argue the real benefit of the model that uses a “divide and conquer” philosophy is
may be realized when availability information is proposed. Multiple perspectives or lenses are
extended to the decision maker. applied in the present reference model. Cloud-
based decision support and analytics are examined
Liu et al. (2011) identify three services as part
using an actor lens, a primary driver lens, and
of an overall DSS, a data service, an information
an infrastructure lens. Multiple perspectives
service and an analytics service and several
allow deeper understanding and facilitate future
other components. An application programming
innovative and creative uses. Further, each of these
interface acts as a mechanism to populate source
perspectives is approached in a hierarchical fashion
systems with raw data and to pull operational
with several levels of increasing detail and focus.
reports. Operational transaction systems run day-
to-day business operations and provide source The three lenses capture the breadth, and the levels
data for the data warehouse and DSS environment. capture the depth of the concepts. In the present
An enterprise application integration/staging area nomenclature, a prefix is used to represent the lens,
provides an integrated common data interface or the perspective being used; for example, A. is
and interchange mechanism for real-time and used for the actor lens. The levels are numbered
source systems. Extract, transform, and load 0, 1, 2, 3... where level 0 is the highest level that
(ETL) processes move source data into the data provides a birds-eye view of cloud-based decision
warehouse and move data from one location to support and level 1 would zoom in to the next
another. Metadata management is part of the data level of detail and so on.
service and it describes the meaning and structure
of business data, as well as how it is created, 4.1 Actor Lens
accessed and used. In addition, there is provision
for data marts in the cloud that support specific Design, implementation, and use of cloud-
decision and analytical needs and provide business based decision support systems involve multiple
units more flexibility, control, and responsibility. organizations and individuals or actors. This
reference model from the actors’ perspective is
As seen in Table 1 and the above examples, cloud- shown in Figure 1. The Level 0 conceptualization
based decision support and analytics have been used depicts the relationships, interactions, and
in several domains of applications including but coordination between all these actors. This model
not limited to logistics, environmental monitoring, is structured similarly to the reference model for

ICI Bucharest © Copyright 2012-2021. All rights reserved

10 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

cloud computing from the National Institute of Often, the provider has a portal for managing
Standards and Technology, U.S. Department of the service with tools for provisioning compute
Commerce (Liu et al., 2011). A Cloud-Based and store. Management tasks like monitoring and
decision support and analytics consumer enters billing are also available through these portals.
a contractual relationship with a provider. For Tools for migrating data and legacy applications
example, an organization (consumer) uses Amazon as well as educational materials are included.
Web Services (provider) for the design and use User and identity management with associated
of a system. Optionally, the interaction could be credentials, rights, and security provisions allow
with an intermediary who provides value-added teams of developers and users proper access.
services. For example, an organization uses a SaaS Federated identity management allows users
application offered by an intermediary who in turn from consumer organizations single-sign-on
contracts with a provider. convenience. Billing can be customized to match
the transfer pricing mechanisms prevalent in
The consumer (A.1) is one of many types: an
the consumer organization. In the ZDnet survey
individual (A.1.1), a team, an organization or a
of 2019 top cloud providers, Dignan (2019)
group of organizations. Further, an individual’s
estimates Amazon Web Services - AWS annual
task within an organization could be operational
revenue run rate based on latest quarter at $29.72
(A.1.1.1), tactical or strategic. The nature of
the provided decision support varies based on billion, Microsoft Azure annual revenue run rate
the type of task being performed. The teams or at $11 billion, Google cloud platform’s annual
groups of users are operating synchronously revenue run rate at $4 billion and Alibaba’s annual
or asynchronously and are located at the same revenue run rate at $3.85 billion. Workday with
place or are at different places. Again, the nature an annual cloud revenue run rate of $3 billion
of decision support, the technology, and the is an example of a cloud intermediary that
interactions would vary among these use cases. allows Amazon Web Services - AWS hosting.
The domain within the organization provides In addition to hybrid on-premises/off-premises
useful context. Financial applications use a infrastructure, the use of multiple public providers
lot of numbers whereas a marketing use case is growing. According to the survey, 97 percent
involves more text processing like customer of respondents reported their companies to use
reviews. Lastly, a supply chain is an example of Amazon Web Services - AWS, but 35 percent
a consumer that involves a group of participating also said they actively use Azure too. Twenty-
organizations. The provider could be an external four percent use Amazon Web Services - AWS and
organization, a consortium or an internal private Google Cloud Platform together.
cloud. A hybrid provider is another possibility.

Figure 1. Reference Model - Actor Lens

https://www.sic.ici.ro
 An Overview of Cloud-Based Decision Support Applications and a Reference Model 11

4.2 Primary Decision Driver Lens trends with colours, fonts, animations, and other
visual elements. Graphical tools allow a simplified
Decision support research has a rich history display of data with minimal clutter to facilitate
spanning several decades. Power (2001) reviewed business insight.
the DSS research and classified the decision
support provided as being primarily driven by Correlations between outcomes of interest
one or more of several factors. He classified and independent variables can be examined to
decision support systems into data-driven, model- facilitate hypothesis formulation and mental model
driven, communication-driven, knowledge- building. Efficiency and speed of visualization
driven and document-driven systems. Using the are additional benefits provided by massively
lens of the primary driver for the provision of parallel processing in cloud clusters. Good visuals
decision support, Figure 2 presents the proposed enable meaningful storytelling. By allowing easy
reference model for cloud-based decision integration of data from multiple sources with
support and analytics. There is also a possibility limited technical database skills requirements,
of overlap between data-driven, model-driven, users are able to create visualizations and
communication-driven, knowledge-driven and get insight with little help from Information
document-driven features. Technology - IT personnel.

The exponential growth in the amount of data Model-driven decision support (D.2) is primarily
being collected and stored has been the subject of facilitated by use of mathematical and other types
mainstream media, practitioners, and researchers of models. Spreadsheets (Kaparthi & Power, 2002)
recently. Big data and analytics are common allow mathematical models to be created and users
terms now. Cloud-based data-driven systems can conduct what-if analyses and goal seeking to
(D.1) are especially suited for storing large help facilitate decision-making. For example,
amounts of data with on-demand, pay-as-you- once a model is constructed, a manager could ask
go storage and services. Application software to a question like “what if I increase my advertising
visualize this data provides decision support to spend by 10% on Facebook, how might it impact
users by allowing management reporting, data the sales?” Model-driven decision support relies
exploration and exception identification, ABC on research contributions from the disciplines of
analysis, executive dashboards with KPI tracking, management science and operations research.
as well as a graphical presentation of data and Linear programming, nonlinear programming,

Figure 2. Reference Model – Primary Decision Driver Lens

ICI Bucharest © Copyright 2012-2021. All rights reserved

12 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

integer, and binary programming as well as multi- media, companies can now access an enormous
criteria decision-making have many applications amount of text with mentions, comments, and
in business organizations. reviews of their products and other happenings.
Customers’ emails and chats are other important
Handling uncertainty is done through stochastic sources. Document-driven decision support is
modelling as well as Monte-Carlo simulations. provided by several tools that can search through
Computer-aided design and modelling, simulation this large amount of unstructured data to find
of effects of operating conditions through finite documents of interest.
element methods, and augmented reality and
virtual reality allow for visualization of products Automated summarizations, classification,
and facilities prior to them being built. Automated sentiment analysis, natural language processing,
decision algorithms can also be deployed in the and bibliometric & lexical analysis allow
cloud. Cloud-based systems allow for real-time decision-makers to listen to their customers
optimization of models to enable innovative use and other stakeholders. Word clouds and other
cases like when Uber solves multiple shortest path visualizations provide insight into the situation.
problems when a customer requests a ride to find Optical Character Recognition (OCR) is easily
the closest drivers available. implemented in the cloud. For example, Google
Cloud Vision Application Programming Interface
A major way in which knowledge-driven decision – API has a Detect Text OCR feature. Domain-
support (D.3) is different from the others is specific tools are also being designed to deal
that the system can suggest or recommend a with vocabulary related to certain professions.
course of action to managers. These systems For example, Amazon Comprehend Medical is
have specialized problem-solving expertise. useful for extracting complex medical information
This expertise consists of knowledge about a from unstructured text. This service can identify
particular domain, understanding of problems medical information, such as medical conditions,
within that domain, and skill at solving some of medications, dosages, strengths, and frequencies
these problems. A future step in knowledge driven from a variety of sources like doctor’s notes,
DSS is to augment decision maker intelligence clinical trial reports, and patient health records.
and problem-solving capabilities by providing
artificial cognitive systems (Filip, 2020; Siddike Communications-driven decision support (D.5) has
et al., 2018). its roots in research about group decision support
systems. It includes communication, collaboration
A related concept is Data Mining. It refers to a and decision support technologies that groups of
class of analytical applications that search for people use together (Filip et al., 2017). Cloud-
hidden patterns in a database. Data mining is based systems are well suited for communication-
similar to machine learning and is the process of
driven decision support due to the need for
sifting through large amounts of data to produce
coordination, collaboration, planning, assessment,
data content relationships. Analytics applies these
decision-making, consensus generation between
data mining concepts and other tools.
team members in geographically distributed global
According to Power et al. (2018), business organizations. Such systems allow users the use
analytics is a “systematic thinking process that of communications as well as decision models.
applies qualitative, quantitative, and statistical
Support for electronic communication, messaging,
computational tools and methods to analyse
scheduling, document sharing, and other group
data, gain insights, inform, and support decision-
productivity activities enhance decision support.
making. Any particular analysis may use a variety
As mentioned in the subsection about the actor
of techniques including diagnostic, predictive,
lens, teams or groups of users may be operating
prescriptive, and optimization models.”
synchronously or asynchronously and are located
Document-driven decision support (D.4) builds at the same place or are at different places. Massey
on document management systems that were (2008) lists tools that are relevant to the present
designed to make organizations go paperless context too. Support for synchronous same-place
by tracking, storing, archiving, and retrieving situations is provided through brainstorming,
documents based on keywords, indexing, and voting, prioritization, and presentation tools.
content. With the rapid growth in web and social Support for synchronous different-place situations

https://www.sic.ici.ro
 An Overview of Cloud-Based Decision Support Applications and a Reference Model 13

is provided through audio-video conferencing, S.1 Data Sources

chat, instant messaging, screen sharing and
co-editing. Editing a shared Google document Large volumes of data come from a variety
synchronously is an example of such support of sources over the Internet to the cloud
provided by cloud-based systems. infrastructure. These sources include both external
and internal sources. The external sources are
Workflow and e-signatures formalize business generally public data extracted from various
processes. For example, Slack is a cloud- sources outside of the organization such as
based collaboration hub that has gained a lot of websites, mobile devices, audio and video, third-
popularity recently. Asynchronous situations can party aggregators, machines and sensors, and
benefit through such tools too. social media. Internal data is retrieved from all
functional areas of an organization.
4.3 Services Lens
S.2 Integration Services
Next, the service lens perspective for Cloud-based
decision support and analytics proposed in this The cloud service provider uses integration
paper is presented. This proposed reference model services to organize data and information
is partially based on the Cloud Standards Customer collected from various sources into data
Councils’ (CSCC) reference architecture for the repositories. Cloud integration solutions include
traditional analytics production environment the discovery of data sources, cleansing,
(Beradi et al., 2015). The present reference transforming and monitoring data. Different
model takes into consideration the provider methods such as Extract Transform Load - ETL,
cloud, enterprise network, various deployment event-based processing, capturing the change of
options, analytical services, and multi-channel data with replication, and streaming computing
delivery. As shown in Figure 3, there are five main are used. The primary goal is to create and
components of the reference model as well as the organize data repositories (using enterprise
various deployment options and data sources. The metadata definitions) where analytics can be
following sections describe the main components performed to deliver actionable insights.
of the proposed model.

Figure 3. Reference Model - Services Lens

ICI Bucharest © Copyright 2012-2021. All rights reserved

14 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

S.3 Storage Services (e.g., SaaS, PaaS, IaaS) and utilization choices
(fully cloud-based, hybrid mix of cloud and legacy
Storage services in the cloud are provided through system, fully legacy system-based, etc.).
Data-as-a-Service (DaaS), Infrastructure-as-a-
Service (IaaS) and Replication-as-a-Service. DaaS 5. Application of the Proposed
builds a layer of abstraction so that organizations Reference model to Past Works
can access data without knowing where data
resides. IaaS provides processors, hardware and Benatia et al. (2015) propose a cloud-based
networking capabilities to store big data. This multiple criteria decision support system for
model is often implemented via Virtualization to selecting urban housing sustainable development
allow multiple users to work concurrently. The projects in order to help the decision-makers to
cloud provider offers a cost-effective, highly choose the best urban project with conflicting
scalable and efficient Replication-as-a-Service criteria. Their DSS is deployed on a CloudBees
for backing up data so that information can be Cloud platform and managed by the OpenStack
recovered anytime in case of a system failure or infrastructure. According to the authors, this cloud-
in the event of downtime. based DSS had significant advantages: it reduced
the deployment and processing time, ameliorated
S.4 Analytics Services the communication and the cooperation between
Using virtualized analytical and information the decision-makers, facilitated the accessibility,
services provide the ability for all users in an and decreased the cost. The DSS was built on the
organization to perform analysis on the datasets Cloud Computing architecture with three layers
that they are interested in analysing (Demirkan and included a multiple criteria decision-making
& Delen, 2013). Analytics services provide method as well as the procedure of negotiation
software tools, templates and deployed analytics in order to help the decision-makers to select
applications for use by an organization. Some of the best urban project. This example would be
the examples of such tools include a variety of categorized as incorporating both model-driven
software tools, analytics discovery and exploration and communication-driven cloud-based DSS
(for building new analytical models), dashboards, features within the urban planning domain.
and deployed analytics applications (for predictive Gupta et al. (2017) propose an intelligent decision
analytics, content analytics, planning, and support model that can aid medical experts in
forecasting) (Beradi et al., 2015). predicting heart disease based on the historical data
The core concept of information services is to of patients. Various machine learning algorithms
make consistent data available for any application. were implemented on the heart disease dataset
Information services provide a “single version of to predict accuracy for heart disease occurrence.
the truth” and make data accessible at all times and Naïve Bayes had been selected as an effective
to all applications. Information services minimize model because it provided the highest accuracy
the cost and time by establishing a single, trusted followed by AdaBoost and boosted tree. Further,
source so that it is possible to set up a separate when ensembled, these 3 models increased
consumer of data in a separate application with the overall accuracy. The experimental results
little effort (Demirkan & Delen, 2013). were evaluated using over 10,000 instances that
clearly validated the maximum accuracy through
S.5 Multi-Channel Delivery Options ensembling and minimum execution time in a cloud
environment. This example might be categorized as
The cloud service provider delivers the business incorporating knowledge-driven cloud-based DSS
analytics data, results and solutions over multiple features within the healthcare domain.
channels, including stationary and mobile
channels, such as - desktop, laptops, mobile Hill & Böse (2017) present an innovative decision
phones, tablets, email, printers, etc. support system that simultaneously provides
predictive analytics to logistic nodes as well as
The deployment of analytics and BI in the cloud to collaborating truck companies. They suggest a
is complex. With cloud computing generally, flexible cloud-based service that incorporates an
complexity exists because of multiple deployment advanced forecasting engine based on artificial
models (e.g., public cloud, private cloud, hybrid intelligence capable of providing individual
cloud, and community cloud), service options predictions for users on all planning levels. Their

https://www.sic.ici.ro
 An Overview of Cloud-Based Decision Support Applications and a Reference Model 15

system provided forecasted truck arrival rates Helo & Hao (2017) propose a cloud-based
to the nodes and predicted truck gate waiting production planning system for sheet metal
times at the nodes to the truck companies based processing. According to the authors, cloud
on historical data, economic and environmental manufacturing has the potential to offer
impact factors. Based on the expected workloads, decision support as a service and medium of
the node personnel and machinery could be communication in production planning and
planned more efficiently. One would categorize control. Further, many production planning and
this example as a cloud-based DSS using model- control problems are essentially optimization
driven features in the transportation domain. problems, where the objective is to develop a
plan that meets the demand at minimum cost
Rupnik et al. (2019) describe a system named or maximum profit. Because the underlying
AgroDSS for farm management decision support optimization problem will vary in the different
that bridges the gap between agricultural systems business and operation phases, it is important to
and decision support methodology. The system think about it in a dynamic mechanism and in a
integrates into existing farm management number of interlinked sub-problems at the same
information systems and provides a cloud-based time and to solve these problems and produce
decision support toolbox, allowing farmers collaboration across the supply chain. This
to upload their own data, utilize several data application is categorized as a communication-
analysis methods and retrieve their outputs. The driven and model-driven cloud-based decision
implemented tools include predictive modelling support system in the manufacturing domain.
with an explanation, accuracy evaluation, time
series clustering and decomposition, and structural Lee (2017) suggests that in Retail 4.0, Omni-
change detection. They can help users make channels require seamless and complete
predictions for simulated scenarios and better integration of all available channels for purchasing
understand the dependencies within their domain. and that the diversification of channels not only
Further, the authors applied the AgroDSS system diversifies data sources but also rapidly generates
on a case study of pest population dynamics. One an enormous amount of data. Big data analytics
would categorize this system as a cloud-based can extract meaningful knowledge for decision-
DSS with data-driven and model-driven features making. The author then proposes a genetic
in the agriculture domain. algorithm-based optimization model to support
anticipatory shipping which is getting more
Jayalakshmi & Gomathi (2018) proposed a popular to ensure fast product delivery. The goal
system to report the progress of the elderly of the system is to predict when a customer will
in an appropriate manner with the help of the make a purchase and then begin shipping the
technology used in the healthcare system and product to the nearest distribution centres before
integrate the report of progress to the remote the customer places the orders online. The big
caretakers employing smartphones and videos. data generated from all channels are stored in the
This advancement in medical informatics has cloud. Future purchases are predicted by using
been done to monitor the health of the patients cluster-based association rule mining in terms of
and provide the details to the caretakers, who If-Then prediction rules. The algorithm considers
are nearby the remote areas. Medical informatics transportation cost, traveling distance and the
comprises a huge number of medical resources confidence of prediction rules in the optimization
to enhance storage, retrieval of health data, process. This application is a cloud-based DSS
and usage of these resources in healthcare. with data-driven, knowledge-driven and model-
Their system requires certain sensors which driven decision support features.
are of low cost, certain electronic devices, and
smartphone for the communication purpose. All the 30 studies that resulted from the search
Due to the condition monitoring of patients in described earlier in the second section are
real-time, the caretaker can provide suggestions presented in Table 2. As seen in Table 2, articles
regarding diagnosis through a video conference. relevant to cloud-based DSS started appearing in
Abnormalities are identified at an early stage the academic literature in 2012. Those articles
and increase the physical and mental health of described several domains of application. The most
patients. One would categorize this application frequently provided cloud-based decision support
as communication-driven cloud-based DSS in the was model-driven, then data-driven, followed
healthcare domain. by knowledge-driven, communication-driven

ICI Bucharest © Copyright 2012-2021. All rights reserved

16 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

and lastly document-driven. The most popular conceptualizing the depth of the involved concepts.
application domain has been manufacturing, and These systems were examined from an actor lens,
healthcare is the second. Over the time period in a primary-driver lens, and a services’ lens. The
the review, a broadening of domains as well as an components involved in each of these perspectives
increase in the number of applications are seen. were then explored in a hierarchical fashion by
zooming into each one at a time. Finally, several
6. Conclusion other studies from the prior literature were then
categorized according to the primary driver lens in
As computer and information technology evolved the proposed reference model.
so did the architectures for designing applications
that provide decision support. There have been five In addition to helping understand the complex
generations of such architectures, starting with the cloud computing phenomenon, this reference
first generation that included the use of a mainframe model can also be used to identify avenues for
computer, the second generation that required future decision support research. For example,
time-sharing, the third generation articulated by there is a need for design science research on
Sprague & Carlson (1982), the fourth generation document-driven cloud-based decision support in
that was client-server and web-based, generation law firms and other specialized settings. Another
4.5 that used mobile devices to generation 5.0 that example of a need is extending the limited
expanded the technology used to include cloud- research that examines business adviser bots that
based decision support and analytics. use natural language voice and/or text interfaces
to provide conversational decision support.
This article presents a multi-perspective hierarchical
reference architecture model for conceptualizing This hierarchical, three lens, decision support
cloud-based decision support and analytics. reference architecture model can maximize
the impact and facilitate the adoption and
Multiple perspectives were used for conceptualizing understanding of cloud-based decision support
the breadth and creating a hierarchical approach for and analytics.

REFERENCES
Amazon.com (2006). Announcing Amazon Elastic Support System for Selecting Urban Housing Projects,
Compute Cloud (Amazon EC2) – Beta. Available Human Systems Management, 34(2), 119-132.
at: <https://aws.amazon.com/about-aws/whats-
new/2006/08/24/announcing-amazon-elastic-compute- Berardi, T., Chessell, M., Gupta, M., Kak, A., Kreger,
cloud-amazon-ec2---beta/>, last accessed: 01/26/2021. H., Statchuk, C. & Schalk, K. (2015). Customer Cloud
Architecture for Big Data and Analytics, Version 1.1,
Armbrust, M., Fox, A., Griffith, R., Joseph, A. D., 1-20. Cloud Standards Customer Council.
Katz, R., Konwinski, A., Lee, G., Patterson, D.,
Rabkin, A. Stoica, I. & Zaharia, M. (2010). A View Bock, A., Svensson, Kleiner, A., Lundberg, J. &
of Cloud Computing, Communications of the ACM, Ropinski, T. (2017). A Visualization-Based Analysis
53(4), 50-58. System for Urban Search & Rescue Mission Planning
Support, Computer Graphics Forum, 36(6), 148-159.
Bara, A., Lungu, I., Oprea, S. V., Carutasu, G., DOI: 10.1111/cgf.12869
Botezatu, C. P. & Botezatu, C. (2014). Design
Workflow for Cloud Service Information System for Chapron, G. (2015). Wildlife in The Cloud: A New
Integration and Knowledge Management Based in Approach for Engaging Stakeholders in Wildlife
Renewable Energy, Journal of Information Systems Management, Ambio, 44, 550-556. DOI: 10.1007/
and Operations Management, 8, 230-237. s13280-015-0706-0

Bellini, P., Bruno, I., Cenni, D., Fuzier, A., Nesi, P. Chichernea, V. (2014). The Use of Decision
& Paolucci, M. (2012). Mobile Medicine: Semantic Support Systems (DSS) in Smart City Planning and
Computing Management for Health Care Applications Management, Journal of Information Systems &
on Desktop and Mobile Devices, Multimedia Tools Operations Management, 8(2), 1-14.
and Applications, 58(1), 41-79. DOI: 10.1007/s11042-
010-0684-y Choi, T. & Lambert, J. H. (2017). Advances in Risk
Analysis with Big Data, Risk Analysis, 37(8), 1435-
Benatia, I., Laouar, M. R., Bendjenna, H. & Eom, S. 1442. DOI: 10.1111/risa.12859
B. (2015). A Cloud-Based Multi-Criteria Decision

https://www.sic.ici.ro
 An Overview of Cloud-Based Decision Support Applications and a Reference Model 17

Columbus, L. (2019). Public Cloud Soaring Inter-Terminal Truck Routing Optimization, Flexible
To $331B by 2022 According to Gartner, Services and Manufacturing Journal, 29(3-4), 504-
Forbes. Available at: <https://www.forbes. 534. DOI: 10.1007/s10696-017-9280-z
com/sites/louiscolumbus/2019/04/07/public-
cloud-soaring-to-331b-by-2022-according-to- Helo, P. & Hao, Y. (2017). Cloud Manufacturing
gartner/?sh=1480d1de5739>, last accessed: System for Sheet Metal Processing, Production
01/26/2021. Planning & Control, 28(6-8), 524-537. DOI:
10.1080/09537287.2017.1309714
Demirkan, H. & Delen, D. (2013). Leveraging the
Capabilities of Service-Oriented Decision Support Hill, A. & Böse, J. W. (2017). A Decision Support
Systems: Putting Analytics and Big Data in Cloud, System for Improved Resource Planning and Truck
Decision Support Systems, 55(1), 412-412. Routing at Logistic Nodes, Information Technology
and Management, 18, 241-251. DOI: 10.1007/s10799-
Depeige, A. & Doyencourt, D. (2015). Actionable 016-0267-3
Knowledge as A Service (AKaaS): Leveraging Big
Data Analytics in Cloud Computing Environments, Irvine, M. (2012). IBM Cloud Offering Decoder.
Journal of Big Data, 2(1), 1-16. DOI: 10.1186/ Available at: <https://www.ibm.com/blogs/cloud-
s40537-015-0023-2 computing/2012/05/23/the-ibm-cloud-offering-
decoder/>, last accessed: 01/26/2021.
Dignan, L. (2019). Top Cloud Providers 2019: AWS,
Microsoft Azure, Google Cloud; IBM Makes Hybrid Jayalakshmi, M. & Gomathi, V. (2018). Pervasive
Move; Salesforce Dominates SaaS. Available at: health monitoring through video-based activity
<https://www.zdnet.com/article/top-cloud-providers- information integrated with sensor-cloud oriented
2019-aws-microsoft-azure-google-cloud-ibm-makes- context-aware decision support system, Multimedia
hybrid-move-salesforce-dominates-saas/>, last
Tools and Applications, 79, 3699-3712. DOI: 10.1007/
accessed: 01/26/2021.
s11042-018-6716-8
Eskelinen, T., Rsnen, T., Santti, U., Happonen, A. &
Kaparthi, S. & Power, D. J. (2002). Spreadsheets. In
Kajanus, M. (2017). Designing A Business Model
Bidgoli, H. (ed. in chief), Encyclopedia of Information
for Environmental Monitoring Services Using
Fast MCDS Innovation Support Tools, Technology Systems. Academic Press.
Innovation Management Review, 7(11), 36-46.
Kumar, A., Shankar, R., Choudhary, A. & Thakur,
Filip, F. G. (2020). DSS – A Class of Evolving L. S. (2016). A Big Data MapReduce Framework
Information Systems. In Dzemyda, G., Bernatavičienė, for Fault Diagnosis in Cloud-Based Manufacturing,
J. & Kacprzyk, J. (eds.), Data Science: New International Journal of Production Research, 54(23),
Issues, Challenges and Applications. Studies in 7060-7073. DOI: 10.1080/00207543.2016.1153166
Computational Intelligence, vol 869. Springer, Cham.
DOI: 10.1007/978-3-030-39250-5_14 Lauras, M., Benaben, F., Truptil, S. & Charles, A.
(2015). Event-Cloud Platform to Support Decision-
Filip, F. G., Zamfirescu, C. B. & Ciurea, C. (2017). Making in Emergency Management, Information
Computer-Supported Collaborative Decision-Making. Systems Frontiers, 17(4), 857-869. DOI: 10.1007/
Springer International Publishing. Print ISBN978-3- s10796-013-9475-0
319-47219-5.
Lee, C. K. H. (2017). A GA-based Optimisation
Guo, Z. X., Ngai, E., Yang, C. & Liang, X. (2015). Model for Big Data Analytics Supporting Anticipatory
An RFID-based Intelligent Decision Support Shipping in Retail 4.0, International Journal
System Architecture for Production Monitoring of Production Research, 55(2), 593-605. DOI:
and Scheduling in A Distributed Manufacturing 10.1080/00207543.2016.1221162
Environment, International Journal of Production
Economics, 159, 16-28. Lee, C. K. H., Choy, K. L., Ho, G. T. S. & Lin, C.
(2016). A Cloud-Based Responsive Replenishment
Guo, Z. X., Wong, W. K. & Guo, C. (2014). A Cloud- System in A Franchise Business Model Using A Fuzzy
Based Intelligent Decision-Making System for Order Logic Approach, Expert Systems, 33(1), 14-29. DOI:
Tracking and Allocation in Apparel Manufacturing, 10.1111/exsy.12117
International Journal of Production Research, 52(4),
1100-1115. Liu, F., Tong, J., Mao, J., Bohn, R. B., Messina,
J. V., Badger, M. L. & Leaf, D. M. (2011). NIST
Gupta, N., Ahuja, N., Malhotra, S., Bala, A. & Kaur, G. Cloud Computing Reference Architecture. Available
(2017). Intelligent Heart Disease Prediction in Cloud at: <https://www.nist.gov/publications/nist-cloud-
Environment Through Ensembling, Expert Systems, computing-reference-architecture>. DOI: 10.6028/
34(3): e12207. DOI: 10.1111/exsy.12207 NIST.SP.500-292

Heilig, L., Lalla-Ruiz, E. & Vo, S. (2017). Port-IO: Lucas, M. (2017). CW@50: Data Storage Goes
An Integrative Mobile Cloud Platform for Real-Time From $1m to 2 Cents Per Gigabyte. Available at:

ICI Bucharest © Copyright 2012-2021. All rights reserved

18 Shashidhar Kaparthi, Arti Mann, Daniel J. Power

<https://www.computerworld.com/article/3182207/ Power, D. J. (2018). What Is the Evolution of

cw50-data-storage-goes-from-1m-to-2-cents-per- Decision Support and Analytics Architectures?
gigabyte.html>, last accessed: 01/26/2021. Available at: <https://dssresources.com/faq/index.
php?action=artikel&id=449>, last accessed:
Mackay, E. B., Wilkinson, M. E., Macleod, C. J. 01/26/2021.
A., Beven, K., Percy, B. J., Macklin, M. G., Quinn,
P. F., Stutter, M. & Haygarth, P. M. (2015). Digital Power, D. J., Heavin, C., Mcdermott, J. & Daly, M.
Catchment Observatories: A Platform for Engagement (2018). Defining Business Analytics: An Empirical
and Knowledge Exchange Between Catchment Approach, Journal of Business Analytics, 1(1), 40-53.
Scientists, Policy Makers, And Local Communities, DOI: 10.1080/2573234X.2018.1507605
Water Resources Research, 51(6), 4815-4822. DOI:
10.1002/2014WR016824 Power, D. J. & Kaparthi, S. (1998). The Changing
Technological Context of Decision Support Systems.
Massey, A. P. (2008). Collaborative Technologies. In. Berkeley, D., Widmeyer, G. R., Brezillon, P.
In Burstein, F. & Holsapple, C. (eds.), Handbook & Rajkovic, V. (eds.), Context-Sensitive Decision
on Decision Support Systems, Vol. 17. New York: Support Systems. London: Chapman and Hall.
Springer-Verlag.
Rupnik, R., Kukar, M., Vračar, P., Košir, D., Pevec,
Mell, P. & Grance, T. (2009). The NIST Definition D. & Bosnić, Z. (2019). AgroDSS: A decision support
of Cloud Computing, Technical Report 15. National system for agriculture and farming, Computers
Institute of Standards and Technology, Information Electronics Agriculture, 161, 260-271. DOI: 10.1016/j.
Technology Laboratory. compag.2018.04.001

Mercier-Laurent, E., Rousseaux, F. & Haddad, R. Russell, S., Yoon, V. & Forgionne, G. (2010). Cloud-
(2018). Preventing and Facing New Crisis and Risks Based Decision Support Systems and Availability
in Complex Environments, International Journal of Context: The Probability of Successful Decision
Management & Decision Making, 17(2), 148-170. Outcomes, Information Systems and E- Business
DOI: 10.1504/IJMDM.2018.092570 Management, 8, 189-205. DOI: 10.1007/s10257-010-
0126-4
NCTA (2019). The Shrinking Cost of a Megabit.
Available at: < https://www.ncta.com/whats-new/the- Siddike, M. A. K., Spohrer, J., Demirkan, H. &
shrinking-cost-of-a-megabit>, last accessed: 01/26/2021. Kohda, J. (2018). People’s Interactions with Cognitive
Assistants for Enhanced Performances. In Proceedings
O’Flaherty, B. & Heavin, C. (2015). Positioning of the 51st Hawaii International Conference on System
Predictive Analytics for Customer Retention, Journal Sciences (pp. 1640–1648).
of Decision Systems, 24(1), 3-18.
Singh, S. K. (2017). Conceptual Framework of a
Peng, T. & Shiang, W. (2014). Developing A Cloud Cloud-Based Decision Support System for Arsenic
Carbon Footprint Decision Support System for Supply Health Risk Assessment, Environment Systems &
Chain, International Journal of Electronic Business Decisions, 37(4), 435-450. DOI: 10.1007/s10669-
Management, 12(3), 200-213. 017-9641-x

Pick, J. B., Turetken, O., Deokar, A. V. & Sarkar, Sprague Jr., R. H. & Carlson., E. D. (1982). Building
A. (2017). Location Analytics and Decision Effective Decision Support Systems. Englewood
Support: Reflections on Recent Advancements, Cliffs, N.J.: Prentice-Hall, Inc.
Decision Support Systems, 99(1-8). DOI: 10.1016/j.
dss.2017.05.016 Thomas, B. G., Bollapragada, S., Akbay, K., Toledano,
D., Katlic, P., Dulgeroglu, O. & Yang, D. (2013).
Power, D. J. (2001). Supporting Decision-Makers: Automated Bed Assignments in A Complex and
An Expanded Framework. In Harriger, A. (ed.), Dynamic Hospital Environment, Interfaces, 43(5),
eProceedings of the 2001 Informing Science 435-448.
Conference, Krakow, Poland, ISSN 1535-0703 (pp.
431-436). DOI: 10.28945/2384 Tsiligiridis, T., Pontikakos, C. & Perdikis, D. (2014).
Architectural Issues of a Location-Aware System
Power, D. J. (2012). What Are Benefits of Decision Applied in Fruit Fly E-Monitoring and Spraying
Support Applications in The Cloud?, Decision Control, AGRIS on-Line Papers in Economics and
Support News. Available at: <http://dssresources.com/ Informatics, 6(4), 195-207.
faq/index.php?action=artikel&id=243>, last accessed:
01/26/2021. Wan, J., Tang, S., Hua, Q., Li, D., Liu, C. & Lloret,
J. (2018). Context-Aware Cloud Robotics for Material
Power, D. J. (2013). Mobile Decision Support and Handling in Cognitive Industrial Internet of Things,
Business Intelligence: An Overview, Journal of IEEE Internet of Things Journal, 5(4), 2272-2281.
Decision Systems, 22(1), 4-9. DOI: 10.1109/JIOT.2017.2728722

https://www.sic.ici.ro