Architectures of analytics intelligent decision technologies systems (IDTS) for the COVID-19 pandemic
Abstract
This article presents a selective literature review of Analytics Intelligent Decision Technologies Systems (Analytics IDTS) developed to support decision-making in business and public organizations, with a particular focus on the global COVID-19 pandemic. We select Analytics IDTS published in 2019–2020 and evaluate them with an Analytics IDTS Design and Evaluation Framework. We include the types of Analytics IDTS, their decisional services, architectural capabilities, and support for phases in the decision-making process. Results are shown for 33 articles in the general Analytics domain and 71 articles in the focused Public Health domain applied to COVID-19, including how these Analytics IDTS were architected and utilized for decision making. Research in descriptive and predictive models is evident in Public Health COVID-19 research reflecting the lak of knowledge about the disease, while predictive and prescriptive models are the primary focus of the general Analytics domain. IDTS in all disciplines rely on Algorithmic decision services and Heuristic Analysis services. Higher-level decisional Synthesis and Hybrid services such as design, explanations, discovery, and learning associated with human decision-making are missing in most types of decision support, indicating that research in Machine Learning and AI has many growth opportunities for future research.
1.Introduction
The COVID-19 pandemic is a public health crisis that has had a concomitant effect on economic and social dimensions worldwide [1, 2]. Globally, as of April 7, 2021, there are over 132 million confirmed cases of COVID-19 reported to the World Health Organization [3]. In response to the ongoing pandemic, governments closed their borders, declared sudden or phased lockdowns in their countries, and implemented quarantine policies for social distancing and isolation, all of which have led to dramatic changes in how organizations across industries act and make decisions [4, 5, 6].
Organizations of different sizes in different industries and countries are confronted with many short-term and potentially long-term challenges, such as safety and health, rules and regulations, value chain and supply chain, the workforce, consumer demand, sales, and marketing [5]. Decision-makers and policymakers around the globe face an urgent need to reframe and leverage their decision-making strategies under the threat of pandemics like COVID-19. Multiple efforts on how computational technologies can help cope with the damage are currently underway in the research arena of intelligent decision support systems. In this paper, we investigate ways that Analytics Intelligent Decision Technologies Systems (Analytics IDTS) can help address critical issues in the context of a global crisis such as COVID-19 [7, 8, 9].
IDTS are defined as information systems utilizing intelligent technologies to enhance the capabilities of decision-makers in understanding a decision problem and selecting a sound alternative [10]. Traditional decision support systems are usually implemented with digital storage and information retrieval systems. Analytics IDTS are enhanced for optimal decision support with intelligent and analytics technologies, such as ontologies, fuzzy cognitive maps, case-based reasoning, agent-based systems, heuristic ruled-based systems, natural language interfaces, classic and modern data mining, and machine learning to offer additional support [11]. These capabilities have frequently been applied to healthcare and crisis management [10, 12, 13, 14]. Accordingly, the ongoing COVID-19 outbreak has revealed countless new themes, concepts, risks, heuristics, rules, and big data that demand updated and enhanced Analytics IDTS to support decision-makers effectively, efficiently, and ethically [9, 15, 16, 17, 18, 19, 20].
Motivated by the emerging challenges as well as the lack of practical transference of applying Analytics IDTS in the global pandemic, we conduct a selective literature review [21] of the applications and emerging trends of Analytics IDTS for the 2019–2020 time period to provide an updated Analytics IDTS Design and Evaluation Framework including the types of systems, their decisional services, architectural capabilities. We analyze the systems with a generic 5-phased decision-making process to report descriptive and quantitative findings for identified papers in the general Analytics domain and focused Public Health domain for COVID-19. We provide an analysis of how these Analytics IDTS were architected and used and then develop recommendations for applying Analytics IDTS to facilitate decision-making in the global COVID-19 pandemic. The results of this study provide frameworks that can assist Analytics IDTS researchers. Practitioners and designers are provided recommendations on using IDTS to cope with global pandemic challenges and crises.
The remainder of this paper is structured as follows: Section 2 provides a summary of the theoretical background of intelligent decision technologies and depicts an adapted Design and Evaluation Framework for applying relevant Analytics IDTS tools and capabilities. Section 3 reports a selective review to identify relevant studies during the 2019–2020 period addressing the themes of COVID-19 and Analytics IDTS. Section 4 reports the results of the review and a discussion of Analytics IDTS applications and implications and for supporting organizations in the global pandemic. Finally, Section 5 concludes with research limitations, recommendations, and conclusions.
2.Theoretical background
2.1Review of intelligent decision technologies systems (IDTS)
Decision Technologies Systems (DTS) are defined as any computer-based system designed to support several or all phases of a decision-making process [22, 23, 24, 25, 26, 27, 28]. DTS have their origin in Decision Support Systems (DSS), which emerged in the early 1970 s and evolved during the past half-century [23, 27, 28, 29, 30]. More recently, the rise of the Internet of Things (IoT) or the Industrial Internet together with accelerated access to large amounts of data enabled the data-intensive DTS trend. Intensively data-driven systems and methods, such as Analytics, have been proposed to improve new and real-time business decisional-making processes. According to Delen and Demirkan [30], Analytics refers to a set of traditional and advanced decision-making tools for transforming data to information and knowledge usable for decision-makers.
Such innovative development has led to the concept of Analytics DTS, described as “the scientific process of transforming data into insight for making better decisions” according to the INFORMS Society [31]. Similarly, Power et al. [32, p. 51] defined Analytics DST as “a systematic thinking process that applies qualitative, quantitative, and statistical computational tools and methods to analyze data, gain insights, inform, and support decision-making.” Specifically, when the core capabilities of the Analytics DTS integrate AI-based mechanisms, such systems can be referred to as Analytics Intelligent DTS or Analytics IDTS. These Analytics IDTS are designed to enhance generic DTS by incorporating more complete data representations, information, and knowledge models, and more intelligent processing algorithms than traditional systems [27, 33].
Table 1
Descriptive analytics IDTS | INT | DES | CHO | IMP | LEA |
---|---|---|---|---|---|
Executive information systems (classic methods) EIS
| • |
|
| • |
|
Data Warehouse-based business intelligence (advanced methods) DW/BI
| • |
|
| • |
|
Predictive analytics IDTS | |||||
Data mining (advanced methods) DM
|
| • | • |
|
|
Statistical business analytics (advanced methods) SBA
| • |
|
|
|
|
Prescriptive analytics IDTS | |||||
Decision support systems (classic methods) DSS
|
| • | • |
|
|
Expert systems/knowledge-based systems and knowledge management systems (classic methods) ES/KBS/KMS
| • | • | • | • | • |
Group decision support systems (classic methods) GDSS
| • | • | • | • | • |
Intelligent (optimization) DSS (advanced methods) i-DMSS
|
| • | • |
|
|
Analytics IDTS can be categorized into three groups based on the analysis methods they utilize: Descriptive, Predictive, and Prescriptive [30]. Descriptive systems primarily answer the question ‘what happened?’ by providing statistical and visual descriptions of historical data. Examples of these systems are Data Warehouse-based Business Intelligence tools (DW/BI) and Executive Information Systems (EIS) that provide standard, ad-hoc, on-demand, and/or interactive querying, reporting, and data visualization. In contrast, predictive Analytics IDTS attempt to answer the question ‘what will happen?’ by analyzing historical data to make predictions about the likelihood of future outcomes. Mathematical/statistical models such as linear and logistic regression, and machine learning models such as neural networks, help discover patterns and trends that suggest future states. Examples of predictive systems are Data Mining (DM) and Business Statistical Analytics DSS (BSA). Finally, prescriptive Analytics IDTS utilize mathematical/statistical models to answer the question ‘what should happen?’. For example, methods such as optimization can be applied to prescribe the best course of action when making tradeoffs between a goal and the constraints of a problem. In practice, some systems embed these types of support, which involve Analytics and Heuristic Optimization methods. For example, Knowledge Management Systems (KMS), Expert Systems or Knowledge-based Systems (ES/KBS), Decision Support Systems (DSS), Group Decision Support Systems (GDSS), and Intelligent DSS (i-DMSS).
Table 1, based on [45], summarizes the leading decision support characteristics provided by various types of classical and modern Analytics IDTS types and their explicit (using the symbol •) and implicit (using the character
Table 2
Task type | Generic services (inputs): Outputs | Generic intelligent task category |
---|---|---|
Algorithmic mechanisms | FIND (query, system): result-set | Retrieval |
ALERT (conditions, system): result-set | Triggering | |
APPLY-MADM (decision-data): decision-result | Calculation | |
WHAT-IF (variable-set, original-model): modified-model | Sensitivity analysis | |
GOAL-SEEKING (goal-variable, original-model): modified-model | Sensitivity analysis | |
IDENTIFY-CRITICAL-VARIABLES (variable-set, original-model): critical-variable-set | Sensitivity analysis | |
MAXorMIN (goals, constraints, model): result-set | Optimization | |
Heuristic analysis | CLASSIFY (data, system): system-pattern | Classification |
MONITOR (system, metrics-set): (system-variations, causal-links of variations) | Classification | |
INTERPRET (data, system): system-state-assessment) | Identification | |
PREDICT (system, events-set, time-period): future-system-state | Identification | |
Heuristic synthesis | CONFIGURE (parts, constraints, goals): system-structure | Design |
PLAN-SCHEDULE (activities, resources, constrains, goals): (states-sequence, system-structure) states-sequence | Design | |
FORMULATE-DESIGN (components, goals, constraints): system-structure | Complex design | |
Heuristic hybrid | EXPLAIN (data, system): system-cause-effect-links | Complex |
RECOMMEND (base system, required system): change-actions | Complex | |
CONTROL (system-state, goals): input-system-actions | Complex | |
DISCOVER (data, system): knowledge-structures | Complex | |
LEARN (system, knowledge-on-system): new-knowledge | Complex |
Figure 1.
2.2Analytics IDTS design and evaluation framework
The Analytics IDTS Design and Evaluation Framework has been derived from the intelligent DMSS Design and Evaluation Framework (IDEF-i-DMSS) reported in [27]. IDEF-i-DMSS was elaborated to integrate the DMSS literature with the Artificial Intelligence (AI) literature to improve designs of i-DMSS, as well as providing an architectural evaluation framework. The underlying theoretical premise of the IDEF-i-DMSS framework, adapted in this study as the Analytics IDTS Design and Evaluation Framework, proposes that decision-making phases and steps can be enhanced with decisional services supported by architectural capabilities implemented computational mechanisms. Figure 1 illustrates the framework.
Table 3
User interface capability levels | |
---|---|
I. Text and basic graphics/ charts | Action language structured commands or menus and as presentation language texts, graphics, and basic charts. |
II. Multimedia or advanced graphics/charts | Action language structured commands or menus and presentation language texts, graphics, advanced charts, sound, animations, and video. |
III. Advanced user interfaces | Action language, natural plain language and presentation language all previous issues enhanced by virtual or augmented reality environments. |
Data, information and knowledge capability levels | |
I. Databases | Plain files, simple data structures, or/and relational database schemes to represent data and information. |
II Multidimensional databases | Complex and highly-structured data structures or/and multidimensional database schemes to represent data and information. |
III. Numerical models | Structured data, information, and knowledge organized in numerical models, such as forecasting models, simulation models, statistical models, Bayesian networks, and neural layers. |
IV. Knowledge bases | Highly semi-structured data, information, and knowledge organized in knowledge chunks. Examples of these schemes are semantic networks, rules, fuzzy rules, frames, scripts, and cases. |
V. Distributed knowledge bases and big data | Network of highly ill-structured data, information, and knowledge organized in knowledge bases or big data distributed repositories. |
Processing capability levels | |
I. SQL methods | SQL actions: searching, adding, updating, deleting, and sorting using a crisp logic mechanism. Drilling-drown, rolling-up, slicing, and pivoting operations for multi-dimensional data warehouses. This level corresponds to descriptive analytics. |
II. MADM, numerical simulation, classic statistics, and optimization methods | Operations of ranking, estimation of distributions and parameters, discrete-event simulation, and optimization. This level corresponds to prescriptive analytics from a quantitative approach. |
III. Data mining and predictive analytics methods | Operations of classification, association, clustering, trend analysis, regression, and forecasting where problems are intensive on quantitative or numerical-based data. This level corresponds to predictive analytics. |
IV. Semi-structured problem-solving methods | Intelligent algorithms for complex analysis tasks such as classification, diagnosis, interpretation, and monitoring/control. Examples are rule-based systems (RBS), case-based reasoning (CBR) techniques, KMS/KBS mechanisms, and OKMS inference algorithms. This level corresponds to prescriptive analytics from a qualitative approach. |
V. Ill-structured problem-solving methods | Intelligent algorithms for complex synthesis tasks such as exploring, explanation, planning, design, and learning. Examples are agent-based systems (ABS) mechanisms, natural-language processing (NLP) mechanisms, and text mining (TM) mechanisms. This level corresponds to a new explanatory analytics from a qualitative approach. |
The Decision-Making Level shows the decision-making phases based generically on Simon [24]. The stages are Intelligence, Design, Choice, Implementation, and Learning. The literature [22, 23, 24, 25, 26] details the steps in the phases such as detecting the problem, gathering data, formulating the problem, classifying and building the model, validating and evaluating the model, performing sensitivity analysis, presenting results, task planning and monitoring, analyzing and synthesizing the outcome and process. Decisional services support DMP phases and steps by the second level, the Decisional Services Level. Four categories of decisional services are proposed: Algorithmic, Heuristic Analysis, Heuristic Synthesis, and Hybrid Services. Table 2 shows a suggested taxonomy of the decisional services at a high level of abstraction. Decisional services are the building blocks for designing an IDTS by selecting the types of services needed for support, and only the services required may be available. The third level, the Architectural Capability Level, includes the User Interface (UI), Data-Information-Knowledge (DIK), and Processing (P) capabilities provided by implementing computational components. The UI and DIK capabilities are based on the general and standard structure for a DMSS [34, 35]. The P capability is based on the levels of intelligence embedded in the computational mechanisms [36, 37].
Table 3 presents a description of UI, DIK, and P capabilities. Table 3 also reports the ordinal conceptual scales to measure the degree of UI capability, structure in the DIK capability, and the degree of intelligence embedded in the computational mechanisms in the IDTS. It must be noted that any support level usually includes or can include capabilities from the previous level. Finally, the fourth and lowest level, the Computational Level, refers to the algorithmic non-intelligent and the AI-based computational mechanisms to be used in a particular IDTS.
We propose that this IDEF-i-DMSS framework provides a conceptual tool to evaluate how Analytics IDTS have been architected and used conjointly with the Analytics IDTS Framework (see Table 1) to support decision-making phases in the context of issues from the COVID-19 pandemic.
3.Selective literature review of intelligent decision technologies for the 2019–2020 period
We conducted a selective literature review to identify relevant research studies reported during the 2019–2020 period addressing the themes of COVID-19 and Analytics Intelligent Decision Technologies Systems. A selective literature review method can be defined as a descriptive research approach and literature analysis research method [21]. This method pursues identifying the most relevant studies on a specific topic or a group of related issues to elaborate a descriptive landscape on the selected topics and highlight insights valuable to state these studies’ current achievements and limitations.
A selective literature review differs from a systematic literature review [38] and a mapping study [39]. A selective literature review relies on a reduced study sample rather than analyzing an exhaustive set of papers under the criteria. Also, it differs in purpose by focusing on specific research questions and extracting core findings rather than elaborating a broad classification of topics of interest to researchers.
We applied the following seven steps in this selective review method. (1) We defined the knowledge inquiry as researching the use and architectural design of Analytics Intelligent Decision Technologies Systems to support a decision-making process. (2) The selection criteria identified leading journals, defined as the top 10%, in the Analytics and Public Health domains. The final set of journals and number of articles is shown in the Appendix. (3) We used the search statement as ‘COVID-19’ plus any one of the terms ‘analytics’, ‘decision making’, ‘knowledge management’, ‘business intelligence, ‘simulation’, ‘modelling’, ‘optimization’, ‘intelligent’ for the timeframe 2019–2020. (4) We used Google Scholar to identify articles. A total of 325 and 299 articles were located, respectively, for the Analytics and Health domains. (5) Articles that did not address decision-making were excluded after applying the first inclusion-exclusion criteria that the article address COVID-19, 95 and 161 articles remained in the two domains. After applying the second criteria that the article explicitly addresses decision support, a final tally of 33 and 71 articles in the Analytics and Health domains, respectfully, were analyzed. (6) We downloaded the identified articles for detailed analysis. (7) The framework in Tables 4 and 5 was populated.
4.Results and discussion
4.1Descriptive results
Tables 4 and 5 show the evaluations for the 33 and 71 IDTS cases found in the general Analytics and focused Public Health domains, respectively. Tables 4 and 5 show the descriptive results on how these 33 and 71 IDT cases are classified as one of the three types of analytics (i.e., descriptive, predictive, and prescriptive) and the eight types of IDTS (EIS, DW/BI, SBA, DSS, ES/KBS/KMS, GDSS and Optimization i-DSS). Further, we show how they support one or several phases of the Generic Decision-Making Process, as well as how these they are architected and structured with User Interface (UI), Data, Information and Knowledge (DIK), and Processing capabilities levels. The description of each IDTS architectural capability level is reported in Table 3.
Tables 6 and 7 report the complimentary evaluations for the 33 and 71 IDTS cases. Tables 6 and 7 show the descriptive results on how these 33 and 71 IDTS cases, also classified by the three types of Analytics and the eight types of IDTS, provide one or several intelligent decisional services. The description of each IDTS intelligent decisional service is presented in Table 2.
4.2Discussion
The type of Analytics IDTS from the two domains, Analytics and Public Health, is shown for comparison in Table 8. As can be seen, Predictive Analytics is provided by more systems than Descriptive or Predictive, with approximately 40–60% of the total cases. Interestingly, Descriptive and Prescriptive Analytics support is different in the two domains, with Descriptive Analytics more critical in Public Health and Prescriptive Analytics necessary in the more general Analytics domain. The result is not surprising since Public Health publications have focused on a data-driven approach to aggregating Big Data from multiple sources over the past two years. The Analytics community focused on reporting, describing and visualizing data since COVID-19 was a new disease with many unknowns regarding how infection spread and potential mitigation. As more became known about the disease, predictive models were developed based on modeling of similar types of coronaviruses. More broadly, the Analytics domain is maturing and has focused more recently on AI methods that support human decision-making in the Prescriptive Analytics area. In addition, Descriptive Analytics is more domain-dependent since the statistical methods are well understood, and visualization methods have already been developed for Big Data.
In the Analytics domain, Table 4 shows that 61% of the models are predictive. Predictive models are generally focused on specific application areas such as finance, marketing, or operations management. Prescriptive models make up the next largest share of models at 36%. These models often explore machine learning and Big Data technologies. Descriptive models have the lowest percentage at 3% over the past two years.
Table 5 shows that 46% of Analytics IDTS are in Predictive analytics, focusing on SBA forecasting and regression models in the public health domain. This situation is consistent with the early utilization of a hybrid approach of statistical and disease transmission models such as those from the Health Metrics and Evaluation (IHME) global health research center at the University of Washington [40] to estimate the impact COVID-19. These models are grounded in real-time data and include human behavior and interven-
Table 4
UI capability level | DIK capability level | Processing capability level | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Types of analytics DMSS from analytics domains | |||||||||||||||
[l] | |||||||||||||||
graphics/charts |
[l] | ||||||||||||||
advanced graphics/charts |
[l] | ||||||||||||||
interfaces (VR,AR) |
[l](1) Databases | ||||||||||||||
databases |
[l](3) Numerical models |
[l](4) Knowledge bases | |||||||||||||
bases and big data |
[l](1) SQL methods |
[l] | |||||||||||||
classic statistics, | |||||||||||||||
and optimization methods |
[l] | ||||||||||||||
predictive analytics methods |
[l] | ||||||||||||||
solving methods (RBS, | |||||||||||||||
CBR, KBS, KMS, OKMS) |
[l] | ||||||||||||||
solving methods (MAS, | |||||||||||||||
NLP, text mining) |
[l] | ||||||||||||||
Descriptive | Executive information systems (EIS) | 0% 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Data warehouse-based business intelligence (DW/BI) | 3% 1 | 1 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | |
Predictive | Data mining (DM) | 30% 10 | 10 | 5 | 0 | 10 | 1 | 4 | 0 | 3 | 10 | 0 | 7 | 0 | 3 |
Statistical business analytics (SBA) | 30% 10 | 10 | 2 | 0 | 10 | 1 | 6 | 0 | 2 | 10 | 3 | 7 | 0 | 1 | |
Prescriptive | Decision support systems (DSS) | 9% 3 | 3 | 2 | 0 | 3 | 0 | 3 | 0 | 0 | 3 | 3 | 0 | 0 | 0 |
Expert systems/ knowledge-based systems and knowledge management systems (ES/KBS/ KMS) | 6% 2 | 2 | 0 | 0 | 2 | 0 | 1 | 1 | 0 | 2 | 1 | 0 | 0 | 1 | |
Group decision support systems (GDSS) | 9% 3 | 3 | 0 | 0 | 3 | 0 | 1 | 0 | 0 | 3 | 2 | 0 | 1 | 0 | |
Intelligent DSS (i-DMSS) | 12% 4 | 4 | 0 | 0 | 4 | 0 | 4 | 0 | 0 | 4 | 3 | 1 | 0 | 0 | |
Totals | 100% 33 | 100% 33 | 30% 10 | 0% 0 | 100% 33 | 9% 3 | 57% 19 | 3% 1 | 15% 5 | 100% 33 | 36% 12 | 45% 15 | 3% 1 | 15% 5 |
Table 5
UI capability level | DIK capability level | Processing capability level | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Types of analytics DMSS from analytics domains | |||||||||||||||
[l] | |||||||||||||||
graphics/charts |
[l] | ||||||||||||||
advanced graphics/charts |
[l] | ||||||||||||||
interfaces (VR, AR) |
[l](1) DATABASES | ||||||||||||||
databases |
[l](3) Numerical models |
[l](4) Knowledge bases | |||||||||||||
bases and big data |
[l](1) SQL METHODS |
[l] | |||||||||||||
classic statistics, | |||||||||||||||
and optimization methods |
[l] | ||||||||||||||
analytics methods |
[l] | ||||||||||||||
solving methods (RBS, | |||||||||||||||
CBR, KBS, KMS, OKMS) |
[l] | ||||||||||||||
solving methods (MAS, | |||||||||||||||
NLP, TEXT MINING) |
[l] | ||||||||||||||
Descriptive | Executive Information Systems (EIS) | 8% 6 | 6 | 3 | 0 | 6 | 0 | 3 | 0 | 0 | 6 | 3 | 0 | 0 | 0 |
Data warehouse-based business intelligence (DW/BI) | 24% 17 | 17 | 2 | 0 | 17 | 1 | 9 | 1 | 0 | 17 | 6 | 0 | 1 | 1 | |
Predictive | Data mining (DM) | 4% 3 | 3 | 0 | 0 | 3 | 0 | 3 | 0 | 0 | 3 | 1 | 0 | 0 | 0 |
Statistical business analytics (SBA) | 42% 30 | 30 | 10 | 0 | 30 | 0 | 30 | 0 | 0 | 30 | 29 | 1 | 2 | 0 | |
Prescriptive | Decision support systems (DSS) | 20% 13 | 13 | 9 | 0 | 13 | 0 | 13 | 1 | 0 | 13 | 13 | 0 | 1 | 0 |
Expert systems/ knowledge-based systems and knowledge management systems (ES/KBS/ KMS) | 1.5% 1 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 1 | 1 | |
Group decision support systems (GDSS) | 0% 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Intelligent DSS (i-DMSS) | 1.5% 1 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 1 | 1 | 0 | 0 | 0 | |
Totals | 100% 71 | 100% 71 | 33% 24 | 0% 0 | 100% 71 | 1% 1 | 83% 59 | 4% 3 | 0% 0 | 100% 71 | 74% 53 | 1% 1 | 7% 5 | 2% 2 |
Table 6
Algorithmic services | Heuristic analysis services | Heuristic synthesis services | Hybrid services | ||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Types of analytics DMSS from analytics domains | |||||||||||||||||||||
of analytics DSS |
Find |
Alert |
Apply-madm |
WHAT-IF |
Goal-seeking | ||||||||||||||||
variables |
MAXorMIN |
Classify |
Monitor |
Interpret |
Predict |
Configure |
Plan-schedule |
Formulate-design |
Explain |
Recommend |
Control |
Discover |
Learn | ||||||||
Descriptive | Executive information systems (EIS) | 3% 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Data warehouse-based business intelligence (DW/BI) | 3% 1 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Predictive | Data mining (DM) | 30% 10 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 10 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 |
Statistical business analytics (SBA) | 30% 10 | 0 | 2 | 1 | 0 | 0 | 1 | 0 | 4 | 0 | 0 | 10 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Prescriptive | Decision support systems (DSS) | 9% 3 | 1 | 0 | 1 | 3 | 0 | 0 | 0 | 0 | 0 | 1 | 2 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 |
Expert systems/ knowledge-based systems and knowledge management systems (ES/KBS/ KMS) | 6% 2 | 1 | 1 | 1 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | |
Group decision support systems (GDSS) | 9% 3 | 2 | 2 | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Intelligent DSS (i-DMSS) | 12% 4 | 0 | 0 | 0 | 0 | 0 | 0 | 4 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | |
Totals | 100% 33 | 12% 4 | 21% 7 | 12% 4 | 12% 4 | 0% 0 | 3% 1 | 12% 0 | 45% 15 | 6% 2 | 3% 1 | 39% 13 | 0% 0 | 0% 0 | 3% 1 | 9% 3 | 0% 0 | 0% 0 | 0% 0 | 0% 0 |
Table 7
Algorithmic services | Heuristic analysis services | Heuristic synthesis services | Hybrid services | ||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Types of analytics DMSS from analytics domains | |||||||||||||||||||||
of analytics DSS |
FIND |
Alert |
Apply-madm |
What-if |
Goal-seeking | ||||||||||||||||
variables |
MAXorMIN |
Classify |
Monitor |
Interpret |
Predict |
Configure |
Plan-schedule |
Formulate-design |
Explain |
Recommend |
Control |
Discover |
LEARN | ||||||||
Descriptive | Executive information systems (EIS) | 8% 6 | 6 | 6 | 0 | 0 | 0 | 0 | 0 | 0 | 6 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Data warehouse-based business intelligence (DW/BI) | 24% 17 | 17 | 17 | 0 | 0 | 0 | 0 | 0 | 0 | 17 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Predictive | Data mining (DM) | 4% 3 | 0 | 3 | 0 | 0 | 0 | 3 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Statistical business analytics (SBA) | 42% 30 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 8 | 0 | 0 | 20 | 2 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Prescriptive | Decision support systems (DSS) | 20% 13 | 0 | 0 | 0 | 13 | 0 | 0 | 0 | 0 | 0 | 0 | 13 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Expert systems/knowledge-based systems and knowledge management systems (ES/KBS/KMS) | 1.5% 1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | |
Group decision support systems (GDSS) | 0% 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
Intelligent DSS (i-DMSS) | 1.5% 1 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
TOTALS | 100% 71 | 33% 24 | 36% 26 | 0% 0 | 18% 13 | 0% 0 | 4% 3 | 1% 1 | 11% 8 | 32% 23 | 1% 1 | 46% 33 | 2% 2 | 0% 0 | 0% 0 | 0% 0 | 1% 0 | 0% 0 | 0% 0 | 0% 0 |
Table 8
Domain | Descriptive analytics | Predictive analytics | Prescriptive analytics |
---|---|---|---|
Analytics IDTS from analytics domain | 3% | 61% | 36% |
Analytics IDTS from public health domain | 32.5% | 46% | 21.5% |
Table 9
Analytics IDTS architectural capability dimensions | Type of support | Analytics IDTS from analytics domain | Analytics IDTS from public health domain |
---|---|---|---|
GUI capability level | Basic user interface | 100% | 100% |
Moderate user interface | 30% | 33% | |
Advanced user interface | 0% | 0% | |
DIK capability level | Databases | 100% | 100% |
Multidimensional databases | 9% | 1% | |
Numerical models | 57% | 83% | |
Knowledge bases | 3% | 4% | |
Distributed knowledge bases | 15% | 0% | |
Processing capability level | SQL methods | 100% | 100% |
MADM, NS, CS, OM | 36% | 74% | |
Intelligent predictive analytics (DM, ML) | 45% | 1% | |
Intelligent prescriptive analytics (RBS, CBR, KMS/KBS, OKMS) | 3% | 7% | |
Intelligent explanatory analytics (ABS, NLP, TM) | 15% | 2% |
tions such as government controls (e.g., masks, social distancing, closures) instituted to contain COVID-19 (http://www.healthdata.org/covid). Factors in these types of models include population density, mobility, mask usage, seasonal patterns of related diseases, deaths, hospitalizations, and Covid test results. A study by Friedman et al. [41] of public global forecast models showed an error of 7–13% at six weeks, a surprisingly good result given the complexities of modeling and giving impetus to efforts to pursue these types of models. Predictive models are being used for health system planning such as hospital resource planning and for policymaking such as mask mandates. More detailed models have been developed for diffusion through a specific community using factors such as network exposure and demographics to model how coronavirus could spread through a densely populated area such as a city [42]. Thus, these models provide insight into ‘what will happen’ using past data to develop scenarios such as most likely, worst case, or best case.
Table 10
Descriptive analytics | Predictive analytics | Prescriptive analytics | Explanatory analytics | |
---|---|---|---|---|
Types of analytics IDTS | Algorithmic services | Heuristic analysis services | Heuristic synthesis services | Hybrid services |
Analytics IDTS from analytics domain | 21% alerting services | 45% classification services | 3% formulation/design services | 9% explanation services |
Analytics IDTS from the public health domain | 36% alerting services | 46% prediction services | 2% configuration services | 1% recommending services |
Table 5 also shows descriptive analytics for COVID-19 analysis, with 32.5% of the studies using these models. Descriptive analytics uses Big Data to characterize the current and past state of factors related to the pandemic. For example, the Johns Hopkins University Coronavirus Research Center [43] provides data curated from many sources to show elements such as global and local deaths, hospitalizations, confirmed cases, and positivity ratio. These data provide a measure of ‘what has happened’.
The third type of modeling approach shows 21.5% for Prescriptive Analytics in Table 5. In general, these types of models utilize AI techniques such as intelligent agents. For example, a geo-social simulation, called location-based social networks, with intelligent agents that employ behaviors based on psychology and social science principles can explore different mitigation strategies to control disease spread [44]. This model allows the exploration of policies that minimize new infections while also minimizing the socio-economic costs of interventions. Optimization models have also been used to allocate resources in anticipation of demand. These models provide answers to ‘what should happen’.
In terms of Analytics IDTS architectural capabilities shown in Tables 4 and 5, support focuses on specific components such as text, basic and advanced graphics, and databases. Table 9 presents a summary of the results.
As seen in Tables 6 and 7, IDTS in all domains rely on decisional Algorithmic services and Heuristic Analysis services. Higher-level decisional Synthesis and Hybrid services such as design, explanations, discovery, and learning associated with human decision making are missing in current Analytics IDTS, indicating that research in Machine Learning and AI still has many growth opportunities. Table 10 reports a summary of the results.
5.Conclusions
This paper has investigated the current state of Analytics IDTS over the timeframe 2019–2020. To categorize associated COVID-19 research, we separated the field into two subsections: the general Analytics domain and the focused Public Health domain. Based on a review of the literature, we conclude that:
• Predictive models are the most widely researched models in both Analytics and Public Health domains;
• Predictive models are helpful in COVID-19 decision making even with the uncertainties inherent in a new problem domain;
• Descriptive models are beneficial in COVID-19 research due to the quantity and variety of Big Data reported from multiple global sources;
• Analytics IDTS currently provide support for lower- and middle-level decision-making capabilities using Algorithmic and Heuristic Analysis;
• Research opportunities for Analytics IDTS are evident in Machine Learning and AI to support higher-level thought processes.
Continued development of Analytics IDTS offers decision-making support in rapidly evolving pandemics and any data-rich environment. This research has shown that such systems can be envisioned as an architecture amenable to the types of decision support needed in the problem domain. Newer, adaptable decision technology systems with advanced machine learning and artificial intelligence capabilities offer the promise of improved decision-making delivered in near real-time that can positively impact human response to challenges such as global pandemics.
Acknowledgments
The authors thank their respective institutions for the academic support provided for this research.
References
[1] | Liu P, Zhong X, Yu S. Striking a balance between science and politics: Understanding the risk-based policy-making process during the outbreak of COVID-19 epidemic in China. Journal of Chinese Governance. (2020) ; 5: (2): 198-212. |
[2] | Verma S, Gustafsson A. Investigating the emerging COVID-19 research trends in the field of business and management: A bibliometric analysis approach. Journal of Business Research. (2020) ; 118: : 253-261. |
[3] | WHO (World Health Organization) [homepage on the Internet]. Geneva: Switzerland; 2021 [cited 2021 Apr 7]. Available from: https://covid19.who.int/. |
[4] | Chahrour M, Assi S, Bejjani M, Nasrallah AA, Salhab H, Fares M, Khachfe HH. A bibliometric analysis of COVID-19 research activity: A call for increased output. Cureus. 12: (3): E7357. doi: 10.7759/7357. |
[5] | Donthu N, Gustafsson A. Effects of COVID-19 on business and research. Journal of business research. (2020) ; 117: : 284-289. |
[6] | He H, Harris L. The impact of COVID-19 pandemic on corporate social responsibility and marketing philosophy. Journal of Business Research. (2020) ; 116: : 176-182. |
[7] | Rehfuess EA, Stratil JM, Scheel IB, Portela A, Norris SL, Baltussen R. The WHO-INTEGRATE evidence to decision framework version 1.0: Integrating WHO norms and values and a complexity perspective. BMJ Global Health. (2019) ; 4(Suppl 1): e000844. |
[8] | Portela A, Tunçalp Ö, Norris SL. Taking a complexity perspective when developing public health guidelines. Bulletin of the World Health Organization. (2019) ; 97: (4): 247-247A. |
[9] | O’Leary DE. Evolving information systems and technology research issues for COVID-19 and other pandemics. Journal of Organizational Computing and Electronic Commerce. (2020) ; 30: (1): 1-8. |
[10] | Papamichail KN, French S. Design and evaluation of an intelligent decision support system for nuclear emergencies. Decision Support Systems. (2005) ; 41: (1): 84-111. |
[11] | Ltifi H, Benmohamed E, Kolski C, Ben Ayed M. Adapted visual analytics process for intelligent decision-making: Application in a medical context. International Journal of Information Technology & Decision Making. (2020) ; 19: (01): 241-282. |
[12] | Zhuang ZY, Wilkin CL, Ceglowski A. A framework for an intelligent decision support system: A case in pathology test ordering. Decision Support Systems. (2013) ; 55: (2): 476-487. |
[13] | Jain S. Intelligent decision support for unconventional emergencies. in: Exploring Intelligent Decision Support Systems. Valencia-García R, Paredes-Valverde M, Salas-Zárate M, Alor-Hernández G, eds. Cham: Springer; (2018) . 199-219. |
[14] | Aggarwal L, Goswami P, Sachdeva S. Multi-criterion intelligent decision support system for COVID-19. Applied Soft Computing. (2021) ; 101: : 1-15. |
[15] | Moghadas SM, Pizzi NJ, Wu J, Yan P. Managing public health crises: The role of models in pandemic preparedness. Influenza and Other Respiratory Viruses. (2009) ; 3: : 75-79. |
[16] | Moberg J, Oxman AD, Rosenbaum S, Schünemann HJ, Guyatt G, Flottorp S, Glenton C, Lewin S, Morelli A, Rada G, Alonso-Coello P. The GRADE evidence to decision (EtD) framework for health system and public health decisions. Health Research Policy and Systems. (2018) ; 16: (1): 1-5. |
[17] | Phillips-Wren G, Pomerol JC, Neville K, Adam F. Supporting decision making during a pandemic: Influence of stress, analytics, experts, and decision aids. in: The Business of Pandemics: The COVID-19 Story. Liebowitz J, ed. Boca Raton: CRC Press. (2020) ; 187-212. |
[18] | Shearer FM, Moss R, McVernon J, Ross JV, McCaw JM. Infectious disease pandemic planning and response: Incorporating decision analysis. PLoS Medicine. (2020) ; 17: (1): 1-12. |
[19] | Squazzoni F, Polhill JG, Edmonds B, Ahrweiler P, Antosz P, Scholz G, Chappin É, Borit M, Verhagen H, Giardini F, Gilbert N. Computational models that matter during a global pandemic outbreak: A call to action. Journal of Artificial Societies and Social Simulation. (2020) ; 23: (2): 1-14. |
[20] | Aghazadeh Ardebili A, Padoano E. A literature review of the concepts of resilience and sustainability in group decision-making. Sustainability. (2020) ; 12: (7): 1-22. |
[21] | Glass RL, Ramesh V, Vessey I. An analysis of research in computing disciplines. Communications of the ACM. (2004) ; 47: (6): 89-94. |
[22] | Huber G. Managerial decision making. New York: Scott, Foresman and Co.; (1980) . |
[23] | Forgionne GA. Decision technology systems: A vehicle to consolidate decision making support. Information Processing & Management. (1991) ; 27: (6): 679-697. |
[24] | Simon HA. The new science of management decision. New York: Harper and Row. (1960) . |
[25] | Sage AP. Behavioral and organizational considerations in the design of information systems and processes for planning and decision support. IEEE Transactions on Systems, Man, and Cybernetics. (1981) ; 11: (9): 640-678. |
[26] | Howard RA. Decision analysis: Practice and promise. Management Science. (1988) ; 34: (6): 679-695. |
[27] | Mora M, Forgionne G, Cervantes F, Garrido L, Gupta JN, Gelman O. Toward a comprehensive framework for the design and evaluation of intelligent decision-making support systems (i-DMSS). Journal of Decision Systems. (2005) ; 14: (3): 321-344. |
[28] | Mora M, Phillips-Wren G, Marx-Gomez J, Wang F, Gelman O. The role of decision-making support systems in IT service management processes. Intelligent Decision Technologies. (2014) ; 8: (2): 147-163. |
[29] | Shim JP, Warkentin M, Shim Courtney JF, Power DJ, Sharda R, Carlsson C. Past, present, and future of decision support technology. Decision Support Systems. (2002) ; 33: (2): 111-126. |
[30] | Delen D, Demirkan H. Data, information and analytics as services. Decision Support Systems. (2013) ; 1: (55): 359-363. |
[31] | INFORMS. [homepage on the Internet]. What are Operations Research and Analytics? 2021 [cited 2021 May 7]. Available from: https://www.informs.org/Explore/Operations-Research-Analytics. |
[32] | Power DJ, Heavin C, McDermott J, Daly M. Defining business analytics: An empirical approach. Journal of Business Analytics. (2018) ; 1: (1): 40-53. |
[33] | Phillips-Wren G. AI tools in decision making support systems: A review. International Journal on Artificial Intelligence Tools. (2012) ; 21: (2): 1-11. |
[34] | Sprague RH, Jr. A framework for the development of decision support systems. MIS Quarterly. (1980) ; 1: : 1-26. |
[35] | Watson HJ. Revisiting Ralph Sprague’s framework for developing decision support systems. Communications of the Association for Information Systems. (2018) ; 42: (1): 13. |
[36] | Elam JJ, Konsynski B. Using artificial intelligence techniques to enhance the capabilities of model management systems. Decision Sciences. (1987) ; 18: (3): 487-502. |
[37] | Gray P, Watson HJ. The new DSS: Data warehouses, OLAP, MDD, and KDD. Proceedings of the AMCIS Conference. Phoenix, AZ; (1996) . |
[38] | Brereton P, Kitchenham BA, Budgen D, Turner M, Khalil M. Lessons from applying the systematic literature review process within the software engineering domain. Journal of Systems and Software. (2007) ; 80: (4): 571-583. |
[39] | Petersen K, Vakkalanka S, Kuzniarz L. Guidelines for conducting systematic mapping studies in software engineering: An update. Information and Software Technology. (2015) ; 64: : 1-8. |
[40] | The Institute for Health Metrics and Evaluation (IHME) [homepage on the Internet]. Seattle, WA: USA; 2021 [cited 2021 Apr 7]. Available from: http://www.healthdata.org/covid. |
[41] | Friedman J, Liu P, Troeger CE, Carter A, Reiner RC, Barber RM, Collins J, Lim SS, Pigott DM, Vos T, Hay SI, Murray CJL, Gakidou E. Predictive performance of international COVID-19 mortality forecasting models. Nature Communications. (2021) ; 12: (1): 1-3. |
[42] | Thomas L, Huang P, Yin F, Luo X, Almquist Z, Hip J, Butts C. Spatial heterogeneity can lead to substantial local variations in COVID-19 timing and severity. Proceedings of the National Academy of Sciences. (2020) ; 117: (39): 24180-24187. |
[43] | John Hopkins University & Medicine [homepage on the Internet]. Baltimore, MA: USA; 2021 [cited 2021 Apr 7]. Available from: https://coronavirus.jhu.edu/. |
[44] | Kim JS, Kavak H, Rouly CO, Jin H, Crooks A, Pfoser D, Wenk C, Züfle A. Location-based social simulation for prescriptive analytics of disease spread. SIGSPATIAL Special. (2020) ; 12: (1): 53-61. |
[45] | Mora M, Wang F, Phillips-Wren G, Marx-Gomez J. The role of DMSS analytics tools in software project risk management. in: Volume II Project Risk Management, Walter de Gruyter GmbH: Berlin, Engemann K and O’Connor R, eds, (2021) ; 49-74. doi: 10.1515/9783110652321-004. |
[46] | Mora M, Phillips-Wren G, Cervantes-Pérez F, Garrido L, Gelman O. Improving IT service management with decision-making support systems. in: Engineering and Management of IT-based Service Systems, Springer-Verlag: Berlin, Mora M, Gómez JM, Garrido L and Pérez F, eds, (2014) ; 215-232. doi: 10.1007/978-3-642-39928-2. |
Appendices
Appendix
Table A1
Journal in the analytics domain | ISSN journal | Total of articles located | Total of valid articles – screening 1 | Total of valid articles – Screening 2 |
---|---|---|---|---|
International Journal of Information Management | 02684012 | 45 | 20 | 0 |
Knowledge-Based Systems | 09507051 | 44 | 3 | 3 |
Information & Management | 03787206 | 8 | 0 | 0 |
MIS Quarterly | 02767783 | 0 | 0 | 0 |
Omega | 03050483 | 14 | 1 | 0 |
Journal of Strategic Information Systems | 09638687 | 5 | 1 | 0 |
Decision Support Systems | 01679236 | 18 | 2 | 0 |
Information Processing & Management | 18735371 | 24 | 8 | 5 |
European Journal of Operational Research | 03772217 | 75 | 6 | 4 |
European Journal of Information Systems | 0960085X | 17 | 15 | 3 |
Journal of Management Information Systems | 1557928X | 0 | 0 | 0 |
Management Science | 15265501 | 0 | 0 | 0 |
Information Systems Research | 15265536 | 1 | 1 | 0 |
IEEE Transactions on Computational Social Systems | 2329924X | 8 | 3 | 1 |
Journal of Decision Systems | 21167052 | 7 | 1 | 0 |
Journal of Intelligent & Fuzzy Systems | 18758967 | 59 | 34 | 17 |
Totals | 325 | 95 | 33 | |
Journal in the public health domain | ISSN Journal | Total of articles located | Total of valid articles – screening 1 | Total of valid articles – screening 2 |
The Lancet Public Health | 24682667 | 45 | 29 | 12 |
Annual Review of Public Health | 15452093 | 2 | 1 | 0 |
Epidemiologic Reviews | 14786729 | 2 | 1 | 1 |
European Journal of Epidemiology | 15737284 | 15 | 13 | 5 |
Eurosurveillance | 15607917 | 71 | 47 | 33 |
American Journal of Public Health | 15410048 | 41 | 31 | 2 |
Journal of Epidemiology and Community Health | 14702738 | 3 | 0 | 0 |
Journal of Health Economics | 18791646 | 1 | 1 | 1 |
Population Health Metrics | 14787954 | 1 | 1 | 0 |
Health Policy and Planning | 14602237 | 28 | 2 | 0 |
Public Health Reviews | 21076952 | 1 | 1 | 0 |
Medical Decision Making | 1552681X | 5 | 3 | 2 |
Health Care Management Review | 15505030 | 0 | 0 | 0 |
European Journal of Health Economics | 16187598 | 1 | 1 | 1 |
Int. Journal of Health Policy and Management | 23225939 | 66 | 21 | 6 |
BMC Medical Informatics and Decision Making | 14726947 | 13 | 7 | 6 |
Totals | 299 | 161 | 71 |