Tracks

The profession of operations research and advanced analytics is constantly developing, growing, and expanding. This track is intended to bring together practitioners and researchers who are working at the edges of the profession to share new areas, explain open problems, formalize new problem areas that are just coming to the fore, and define challenging questions for further development.

Monday, April 15, 9:10–10:00am

A Tutorial on Robust Optimization

Dick den Hertog
In this presentation we explain the core ideas in robust optimization and show how to successfully apply them in practice.

Real-life optimization problems often contain parameters that are uncertain, due to, e.g., estimation or implementation errors. The idea of robust optimization is to find a solution that is immune against these uncertainties. The last two decades efficient methods have been developed to find such robust solutions. The underlying idea is to formulate an uncertainty region for the uncertain parameters for which one would like to safeguard the solution. In the robust paradigm it is then required that the constraints should hold for all parameter values in this uncertainty region. It can be shown that, e.g., for linear programming, for the most important choices of the uncertainty region, the final problem can be reformulated as linear optimization or conic quadratic optimization problems, for which very efficient solvers are available nowadays. Robust Optimization is valuable for practice, since it can solve large-scale uncertain problems and it only requires crude information on the uncertain parameters. Some state-of-the-art modeling packages have already incorporated the robust optimization technology.

In this tutorial we restrict ourselves to linear optimization. We will treat the basics of robust linear optimization, and also show the huge value of robust optimization in (dynamic) multistage problems. Robust optimization has already shown its high practical value in many fields: logistics, engineering, finance, medicine, etc. In this tutorial we will discuss some of these applications. We will also highlight some of the most important (recent) papers on Robust Optimization.

Monday, April 15, 10:30–11:20am

Interpretable AI

Dimitris Bertsimas
We introduce a new generation of machine learning methods that provide state of the art performance and are very interpretable. We introduce optimal classification (OCT) and regression (ORT) trees for prediction and prescription with and without hyperplanes. We show that (a) Trees are very interpretable, (b) They can be calculated in large scale in practical times and (c) In a large collection of real world data sets they give comparable or better performance than random forests or boosted trees. Their prescriptive counterparts have a significant edge on interpretability and comparable or better performance than causal forests. Finally, we show that optimal trees with hyperplanes have at least as much modeling power as (feedforward, convolutional and recurrent) neural networks and comparable performance in a variety of real world data sets. These results suggest that optimal trees are interpretable, practical to compute in large scale and provide state of the art performance compared to black box methods. We apply these methods to a large collections of examples in personalized medicine, financial services, organ transplantation among others.

Monday, April 15, 11:30am–12:20pm

Driving Transparency and Eliminating Reconciliations Across Your Enterprise Value Chain Using Blockchain

Sharad Malhautra
Global Supply Chains thrive on innovations in their business model and technology stacks to help drive growth and reduce costs for the enterprise. The session will include the impact of tokenizing assets and inventory on driving transparency and automation across your value chain. The session will also include several demos of blockchain solutions driving transparency and eliminating reconciliations across enterprise networks.

Monday, April 15, 1:50–2:40pm

Analyzing Everyday Language to Understand People

James Pennebaker
The words people use in everyday language reveal parts of their social and psychological thoughts, feelings, and behaviors. An increasing number of studies demonstrate that the analysis of the most common and forgettable words in English — such as pronouns (I, she, he), articles (a, an, the), and prepositions (to, of, for) — can signal honesty and deception, engagement, threat. status, intelligence, and other aspects of personality and social behaviors. The social psychology of language goes beyond machine learning and, instead, identifies the underlying links between word use and thinking styles. Implications for using text analysis to understand and connect with customers, employees, managers, friends, and even yourself will be discussed.

Monday, April 15, 3:40–4:30pm

Robotic Process Automation

Russell Malz
The rate of change of new disruptive technologies continues to accelerate, and the impact of AI and ML will be bigger and much faster than the impact the Internet had on business. While startups can be built from the ground up based on new AI and ML capabilities, established companies face unique challenges when they look to leverage AI and ML offerings to innovate. Nowhere is this more true than in the analytics domain where both skills and task automation are critical for success. In this talk, learn how innovative companies have successfully used Robotic Process Automation (RPA) to improve efficiency, and how the most forward looking companies are accelerating innovation by establishing the right organizational mindset, and leveraging enabling technologies to gain competitive advantage. Working with big data analytics, and AI strategies at Acxiom, Ayasdi, and Blue Prism, Russ Malz has helped dozens of F500 clients establish and accelerate cognitive strategies.

Tuesday, April 16, 9:10–10:00am

Quantum Computing – Why, What, When, How

Yianni Gamvros
Why Quantum Computing has the potential to significantly disrupt Business Analytics. When will real-world Business Analytics tasks move from classical computers to quantum computers? What are the use cases that can be addressed by Quantum Computing, in the short and medium term? What are current industry and thought leaders working on today? How do Quantum Computers solve optimization problems?

Tuesday, April 16, 10:30–11:20am

Intuition is Unreliable, Analytics is Incomplete

Karl Kempf
In today’s environment, there are many cases where the difference between a good decision and a poor one can be hundreds of millions if not billions of dollars. Decision makers strive to apply their intuition, but intuition is unreliable. Sometimes it is useful, other times misleading. Analytics practitioners want to apply their computational tools, but given the complexity of these cases their models are inescapably incomplete.

Nobel Laurates have explored one side of this situation from the perspective of human psychology. Simon pointed out the bounded rationality available to decision makers while Kahneman described the plethora of biases afflicting that same population. But they failed to supply answers to two important questions crucial to analytics professions. 1) How bad are the decision makers left to rely only on their intuition? Stated a different way, when we start to apply analytics, how much benefit can we reliably expect? 2) Can we benefit from utilizing the intuition? Can analytics inform intuition AND intuition inform analytics to supply a solution superior to either technique applied alone?

We briefly supply an answer to the first question based on projects at Intel Corporation over the past 30+ years. Our answer to the second question occupies the bulk of our presentation. This includes evaluation of ideas from the literature including “pre-mortems” and “nudges”, but will focus on two related approaches we have found to be especially powerful. At one extreme, we will describe support systems for decision makers in operations with examples drawn from manufacturing and supply chain. At the other extreme we address systems that support senior management in deciding product development funding to maximizing profits.

Tuesday, April 16, 11:30am–12:20pm

Explainable AI

Jari Koister
Financial Services are increasingly deploying AI models and services for a wide range of application. These applications include credit life cycles such as credit onboarding, transaction fraud, and identity fraud. In order to confidently deploy such models, these organizations require models to be interpretable and explainable. They also need to be resilient to adversarial attacks. In some situations, regulatory requirements apply and prohibits application of black-box machine learning models.

This talk describes tools and infrastructure that FICO has developed as part of the platform to support these needs. The support is uniquely forward-looking and one of the first platforms to support these aspects of applying AI and ML for any customer.

What we will cover: (1) Examples of Financial Services applications of AI/ML; (2) Specific needs for Explainability and resiliency; (3) Approaches for solving explainability and resiliency.; (4) Regulatory requirements, and how to meet them.; (5) A platform that provide support for xAI and Mission Critical AI.; (6) Further research and product development directions.

Tuesday, April 16, 1:50–2:40pm

Detecting Tax Evasion: A Co-evolutionary Approach

Sanith Wijesinghe
We present an algorithm that can anticipate tax evasion by modeling the co-evolution of tax schemes with auditing policies. Malicious tax non-compliance, or evasion, accounts for billions of lost revenue each year. Unfortunately when tax administrators change the tax laws or auditing procedures to eliminate known fraudulent schemes another potentially more profitable scheme takes it place. Modeling both the tax schemes and auditing policies within a single framework can therefore provide major advantages. In particular we can explore the likely forms of tax schemes in response to changes in audit policies. This can serve as an early warning system to help focus enforcement efforts. In addition, the audit policies can be fine tuned to help improve tax scheme detection. We demonstrate our approach using the iBOB tax scheme and show it can capture the co-evolution between tax evasion and audit policy. Our experiments shows the expected oscillatory behavior of a biological co-evolving system.

Tuesday, April 16, 4:40–5:30pm

Optimal Pay Determination to Reach Diversity Goals

Margret Bjarnadottir
People Analytics are a fast growing field; quantitative methods are becoming main stream in HR departments. There is a great opportunity for the Operation Research Community to play a significant role in how HR decisions are made in the 21^st century. In this talk we will review the growing field of People Analytics and take a deep dive into how data driven decision making can support salary decisions, focusing on demographic pay gaps.

The gender pay gap (and other demographic pay gaps) are a topic of discussion in the boardroom, in the media and among policy makers, with multiple new legislation being passed in a number of states as well as across Europe: in Great Britain, France, and Iceland to name a few. While the methodology for determining pay discrimination is known and mostly agreed upon (a log-regression model), how to close a pay gap has remained an open question. Who should get raises and how much? We apply optimization and descriptive analytics to address this knowledge gap. We first describe a cost optimal approach based on statistics and optimization that can meet the “equal pay for equal work” standard for less than half the cost of the naive method of increasing all female workers’ wages equally. In order to balance cost efficiency with fairness we discuss other fairness driven algorithmic approaches that address and close the gender pay gap. These approaches while more expensive than the cost optimal approach can still save significant costs compared to the naïve approach. We further explore the impacts of closing the gap based solely on cost efficiency, which in some cases are surprising, for example we can show that there may exists men within a firm who if they receive salary increases will reduce the gender pay gap. These men strongly typify male employees in terms of traits. We demonstrate the above algorithmic approaches, savings and costs, using real data from our developing partners.