Engineering Analytics

Advances in Research and Applications

Luis Rabelo, Edgar Gutierrez-Franco, Alfonso Sarmiento, Christopher Mejía-Argueta

• Publisher: CRC Press
• Year: 2021
• Print ISBN: 9780367685348
• eBook ISBN: 9781000453768
• Edition: 1
• Topic: Business
• Subtopic: Operazioni
• Index: Business

Engineering analytics is becoming a necessary skill for every engineer. Areas such as Operations Research, Simulation, and Machine Learning can be totally transformed through massive volumes of data. This book is intended to be an introduction to Engineering Analytics that can be used to improve performance tracking, customer segmentation for resource optimization, patterns and classification strategies, and logistics control towers.

Basic methods in the areas of visual, descriptive, predictive, and prescriptive analytics and Big Data are introduced. Industrial case studies and example problem demonstrations are used throughout the book to reinforce the concepts and applications. The book goes on to cover visual analytics and its relationships, simulation from the respective dimensions and Machine Learning and Artificial Intelligence from different paradigms viewpoints.

The book is intended for professionals wanting to work on analytical problems, for Engineering students, Researchers, Chief-Technology Officers, and Directors that work within the areas and fields of Industrial Engineering, Computer Science, Statistics, Electrical Engineering Operations Research, and Big Data.

1Interactive Visualization to Support Data and Analytics-driven Supply Chain Design Decisions

Milena Janjevic and Matthias Winkenbach

1.1 INTRODUCTION

Supply Chain Design problems typically involve defining the configuration of logistics facilities (i.e., their location, size, service zone) in a manner that maximizes some measure of performance. The Supply Chain Design field is rooted in conventions developed in the 1990s where the focus was on minimizing the total cost of operations, including facility cost, warehousing cost, and transportation cost, among other things. The underlying models used were discrete facility location models solved through various optimization methods (Melo, Nickel, and Saldanha-Da-Gama 2009). The fidelity of these models was constrained by limited data availability and limited computational power.

Today, however, the operating environment of most companies has changed dramatically. Over the last few decades, due to rapid globalization, the proliferation of digital technology and electronic commerce, the growing importance of brands and direct-to-consumer business, and a number of drastic geo-political changes, the conditions under which global supply chains have to efficiently and reliably match supply and demand have become increasingly volatile, uncertain, complex, and ambiguous.

Contemporary Supply Chain Design approaches need to appropriately address these changes in the competitive operating environment of companies. Specifically, they need to make effective use of the abundance of data that is available to companies today, the tremendous advances in predictive and prescriptive analytics, as well as the exponential growth in computational capabilities that recent decades have brought about. Moreover, as service speed, reliability, and flexibility have become critical determinants of competitive advantage, contemporary Supply Chain Design needs to incorporate the non-trivial trade-offs between revenue and cost.

In this chapter, we explore the roles of interactive visualization and human-in-the-loop optimization as a way to improve decision making around Supply Chain Design. These new approaches combine the power of analytical models and the implicit knowledge of expert human decision makers, to improve decision-making transparency and elaborate design solutions that are more capable of responding to the real-life challenges of companies.

1.2 DECISION MAKING IN SUPPLY CHAIN DESIGN

The design of supply chain networks typically entails the selection of a specific network configuration among multiple alternatives based on one or multiple objectives and can therefore be seen as a typical decision-making problem. In the following section, we discuss the main characteristics of this problem and the approaches to decision making in this field.

1.2.1 Characteristics of the Supply
Chain Design Decision-Making Problem

Supply Chain Design is a decision-making problem with some notable characteristics that render it particularly challenging. First, Supply Chain Design is a strategic decision-making problem. It involves strategic decisions on the number, location, capacity, and functions of the production and distribution facilities (Govindan, Fattahi, and Keyvanshokooh 2017). These typically involve substantial investments in infrastructure and are made on a long-term planning horizon. These strategic decisions are interrelated with medium-term tactical decisions (e.g., transportation modes, outsourcing) and short-term operational decisions (e.g., flow of goods in the network). As noted by Klibi, Martel, and Guitouni (2016), Supply Chain Design is a hierarchical decision-making problem, because to make the strategic decisions, decisions at lower levels must be anticipated.

Second, Supply Chain Design is a complex decision-making problem. Here, we employ this term as it is defined in the field of complexity theory (see, e.g., Choi, Dooley, and Rungtusanatham 2001). As noted by Pathak et al. (2007), supply networks typically involve complex interconnections between multiple suppliers, manufacturers, assemblers, distributors, and retailers. In addition, planners are faced with the inability to fully anticipate operational and tactical decisions (Klibi, Martel, and Guitouni 2016). The inability to completely describe the system in terms of its individual constituents and to entirely predict system behavior based on the individual parts results in its complexity (Cilliers and Spurrett 1999; Choi, Dooley, and Rungtusanatham 2001).

Third, Supply Chain Design is a group decision-making problem. Given the strategic role of Supply Chain Design for corporate strategy, choices in this area are relevant to multiple objectives and stakeholders. For example, as demonstrated by Lim and Winkenbach (2019), the configuration of last-mile supply chain networks is directly linked to service levels proposed by the company and has a major impact on the width of its product assortments. These decisions cannot therefore be taken in isolation by the supply chain department but require input and cooperation with other corporate functions, such as sales, marketing, and finance.

1.2.2 Decision Making in the Context of Supply Chain Design

Decision making has been explored from a number of perspectives. While there is no definite theoretical model on different steps and processes involved in decision making, we can differentiate between two main categories of decision making: analytical decision making and intuitive decision making (Cohn et al. 2013). Analytical decision making is often associated with “rational” models of decision making and is characterized by systematic and deliberate information analysis in order to reach a decision (Cohn et al. 2013; Dane and Pratt 2007). Conversely, intuitive decision making is characterized by rapid, non-conscious, and holistic associations, with decisions resulting from the interactions between prior knowledge and experience and the new incoming information (Dane and Pratt 2007; Cohn et al. 2013; Evans 2008).

The use of optimization models to support decision making in Supply Chain Design supports the analytical approach. The increase in the computational power and advances in optimization methods over time have significantly increased the ability of these models to accurately represent real-life business context (e.g., integrating uncertainty through stochastic optimization). However, no matter how elaborate, models are inherently limited and unable to comprehensively represent all aspects of a given problem (e.g., accounting for traffic congestion when estimating a delivery lead time). While increasing model fidelity improves their accuracy, it also entails higher complexity and can result in misunderstanding and lack of trust in what is perceived to be a “black-box” process (Meignan et al. 2015).

Intuitive decision making often translates to choices based on previous experience and implicit knowledge. Research on intuition effectiveness finds that this approach can be superior to analytical approaches, especially in complex decision situations where the amount of information is too huge to be processed deliberately Julmi (2019). However, it can also lead to a number of well-documented cognitive biases, such as the anchoring effect, the shared information bias, or the selection of the “satisficing” rather than the optimal solution (see Carter, Kaufmann, and Michel (2007) for a comprehensive review of judgement and decision-making biases in supply chain management). Literature shows that optimal strategic decision making may require both rationality and intuition (Calabretta, Gemser, and Wijnberg 2017). In line with this finding, Supply Chain Design can benefit from both approaches.

In addition, the multi-stakeholder nature of Supply Chain Design decision making requires approaches that are applicable in group settings where different team members, with potentially different knowledge and objectives, must collaborate to reach a decision. Here, the rational model of decision making generally prescribes decision-support tools integrating the objectives of different stakeholders, and where optimal trade-offs are found through techniques such as multi-objective optimization models or multi-criteria multi-actor analysis. However, in complex decision making characterizing with incomplete information and ambiguous objectives, limitations of pure modeling approaches have been acknowledged (see, e.g., Lebeau et al. 2018; Le Pira et al. 2017). Increasingly, the decision making and operational research literature suggest approaches promoting shared situational awareness and joint knowledge production with the aim of reaching consensus (see, e.g., Janjevic, Knoppen, and Winkenbach 2019; Hegger et al. 2012).

1.2.3 New Perspectives on Decision Making in Supply Chain Design

We identify two potential avenues for increasing the effectiveness of decision making in Supply Chain Design problems. The first avenue relates to methods that aim to combine analytical and intuitive approaches to decision making. Here, we focus on the approaches: human-in-the-loop optimization methods and visual analytic. The second avenue relates to the facilitation of group decision-making processes. These various approaches are now detailed.

Augmenting model-based decision making through human-in-the-loop optimization. Supply Chain Design is a semi-structured problem: it entails structured components (e.g., ability to model the network and employ data to quantify its performance) and unstructured components (e.g., ambiguous objectives or incomplete knowledge). Solving semi-structured problems involves a combination of both standard optimized solution procedures and human intuition or judgments (Niu, Lu, and Zhang 2009).

In the field of Supply Chain Design, human-in-the-loop optimization aims to leverage the combined strengths of machines and humans. A number of techniques can be employed to this end, such as interactive (multi-objective) re-optimization, interactive evolutionary algorithms, and human-guided search (see Meignan et al. (2015) for a comprehensive review). Different techniques allow for refinement of model specifications in an iterative fashion and/or fine-tuning of solutions which in turn enables users to gain better intuition about model results.

The interaction between the human and the optimization algorithm is typically established through a visual interface. However, to the best of our knowledge, literature still lacks a discussion on how to integrate human-in-the-loop optimization and visual analytic to enhance decision making. Liu et al. (2017) explore relationships between interactive optimization and visualization and remark that most extant literature in the field is almost completely silent about visualization and interaction techniques and the user experience. Extant human-in-the-loop optimization research is mainly focused on the development of efficient algorithms and is almost completely silent on the relationships between their use and behavioral processes involved in decision making.

Decision support through infor...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Preface
• Editor Biographies
• Introduction
• 1 Interactive Visualization to Support Data and Analytics-driven Supply Chain Design Decisions
• 2 Resilience-based Analysis of Road Closures in Colombia: An Unsupervised Learning Approach
• 3 Characterization of Freight Transportation in Colombia Using the National Registry of Cargo Dispatches (RNDC)
• 4 Data and Its Implications in Engineering Analytics
• 5 Assessing the Potential of Implementing Blockchain in Supply Chains Using Agent-based Simulation and Deep Learning
• 6 Market Behavior Analysis and Product Demand Prediction Using Hybrid Simulation Modeling
• 7 Beyond the Seaport: Assessing the Impact of Policies and Investments on the Transport Chain
• 8 Challenges and Approaches of Data Transformation: Big Data in Pandemic Times, an Example from Colombia
• 9 An Agent-based Methodology for Seaport Decision Making
• 10 Simulation and Reinforcement Learning Framework to Find Scheduling Policies in Manufacturing
• 11 An Advanced Analytical Proposal for Sales and Operations Planning
• 12 Deep Neural Networks Applied in Autonomous Vehicle Software Architecture
• 13 Optimizing Supply Chain Networks for Specialty Coffee
• 14 Spatial analysis of fresh food retailers in sabana centro, colombia
• 15 Analysis of Internet of Things Implementations Using Agent-based Modeling: Two Case Studies
• Index