Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Gabriel Andraos, Voya Financial
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 11, 2024.
This presentation picks up where my previous one ended (From Explanations to Interpretations at the 2022 BayesiaLab Conference). Back then, I described our approach to optimize the strengths of machines and humans through interpretable AI. This time, we will focus on the key elements that can make this a winning vs. losing combination (hint: not trivial and not obvious). We will explore the benefits of a neuro-symbolic approach and conclude with a proposed framework for optimal collaborative intelligence.
Gabriel Andraos jointly leads Voya’s Machine Intelligence group (VMI). As the co-head of VMI, he focuses on research and development in the application of AI and machine learning models for fundamental investing. Prior to this role, Gabriel was a managing partner and co-founder of G Squared Capital LLP, which was acquired by Voya in 2020. For more than 12 years - at G Squared and at Voya – the team has been running virtual employees – analysts, traders, and portfolio managers with transparent, explainable computer models anchored in fundamentals. Before that, Gabriel held senior investment roles in Europe, the U.S., and Asia, combining knowledge and experience in fundamental analysis with the latest tools in computing and data science. Gabriel received an MBA from Harvard Business School and a BA in Economics from Georgetown University. He also has a Certificate in Quantitative Finance and several artificial intelligence, data science, and machine learning accreditations.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Dr. Lionel Jouffe, Bayesia S.A.S.
Recorded at the 2024 BayesiaLab Spring Conference in Cincinnati on April 11, 2024.
This presentation explores the revolutionary potential of Hellixia, BayesiaLab's specialized assistant, for knowledge discovery and decision support using large-scale language models (LLMs) such as GPT-4.
Hellixia presents itself as a significant innovation, exploiting the wealth of knowledge embedded in LLMs to identify and suggest the most relevant dimensions of a field of study (a process akin to crowdsourcing).
Hellixia uses user-selected keywords (such as causes, levers, effects, concepts, forces, and ideas) to generate mathematical representations capturing the semantics (embeddings), enabling automatic learning of semantic Bayesian networks with BayesiaLab.
Hellixia doesn't limit itself to semantics; it can go further by testing causal hypotheses and automatically generating causal networks, thus providing invaluable help in analyzing and understanding complex relationships within the domains under study.
This ability to automatically translate latent knowledge into semantic and causal networks opens up new avenues for knowledge extraction and decision support, marking a significant advance in knowledge extraction and decision support, marking a significant advance in exploiting the capabilities of LLMs for a wide range of fields, including marketing, industry, medicine, economics, politics, literature, and even philosophy.
Dr. Lionel Jouffe is co-founder and CEO of France-based Bayesia S.A.S. Lionel holds a Ph.D. in Computer Science from the University of Rennes and has worked in Artificial Intelligence since the early 1990s. While working as a Professor/Researcher at ESIEA, Lionel started exploring the potential of Bayesian networks.
After co-founding Bayesia in 2001, he and his team have been working full-time on the development of BayesiaLab. Since then, BayesiaLab has emerged as the leading software package for knowledge discovery, data mining, and knowledge modeling using Bayesian networks. It enjoys broad acceptance in academic communities, business, and industry.
The BayesiaLab community is excited to meet again in person after four years of virtual conferences.
Graduate Hotel Cincinnati, 151 Goodman Drive, Cincinnati, Ohio 45219 April 11–12, 2024
The biannual BayesiaLab conference stands out as the premier event dedicated to applied research with Bayesian networks. Our conference aims to promote the practical deployment of Bayesian networks in research, analytics, and problem-solving.
What started in 2013 in Orlando, Florida, as a modest gathering has now blossomed into an international event, drawing attendees from various corners of the world. Over the years, it has fostered a vibrant community comprising researchers and professionals from many domains.
For questions about the conference, please reach out to us anytime:
Toll-Free: (888) 386-8383 #3
International: +1 615 988-7738 #3
Email: conference@bayesia.us
In conjunction with our conference, we will be offering introductory and advanced BayesiaLab courses. Join scientists from around the world to learn how to apply Bayesian networks to your research.
To get an idea of what a typical BayesiaLab Conference looks like, you can watch the recorded presentations from previous years.
Martin Block, Ph.D., Northwestern University
A Bayesian Media Influence Network solves several problems with a traditional regression-based media marketing mix model. First is the problem of consistent measures across different media types. This is solved by using syndicated survey media and marketing influence measures. Second is the problem of simultaneous consumption and the assumption of independence among predictor variables. Third is the problem of non-linear relationships that may exist between media types and a criterion variable such as sales. A Bayesian Belief Network solves these last two problems and provides an easy-to-understand tool to aid in what has been a traditionally difficult marketing asset allocation decision. Using women’s apparel as an example, the efficacy of the Bayesian Network for monthly spending is shown, identifying a pattern of media types. Factoring past brand purchase behavior into segments shows how the Bayesian Network can be used to target High-End brands, Accessories, and Active Shoes as examples. Other product categories and target definitions can certainly be used.
Martin Block is a Professor Emeritus in Integrated Marketing Communications at Northwestern University and a Director of the Retail Analytics Council. Prior to 1985, he was Professor and Chairperson of the Department of Advertising at Michigan State University. Prior to that, he worked as a Senior Market Analyst in Corporate Planning at the Goodyear Tire and Rubber Company. Co-author of Understanding China’s Digital Generation, Media Generations: Media Allocation in a Consumer-Controlled Marketplace, Retail Communities: Customer Driven Retailing, Analyzing Sales Promotion, and Business-to-Business Market Research. He has published in many academic research journals and trade publications and has several book chapters. Paul has a Ph.D. from Michigan State.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Kurt S. Schulzke, JD, CPA, CFE, University of North Georgia
In response to a lawsuit brought in 2013 by Maximilian Schrems, on May 12, 2023, the Irish Data Protection Commission (DPC) fined Meta Ireland (formerly Facebook) €1.2 billion for violating the EU's General Data Protection Regulation (GDPR) through its transfers to the United States of Facebook users' personal data. These transfers were found by EU authorities to violate Chapter V of the GDPR and the EU Charter of Fundamental Rights because, upon transfer, the data becomes accessible to the U.S. Government through the FISA 702 PRISM program and its successors. In addition to the fine, the DPC also ordered Meta Ireland to suspend further transfers of data to the United States. In the wake of this DPC decision, Meta Ireland must choose how to respond. Options include complying with the decision (i.e., suspending further data transfers), closing Meta's EU business, or violating the order. It is also possible that the DPC's decision might be overturned on appeal, in which case Meta could continue operating as it does now, transferring customer data to the United States with impunity. Meta's choice must be made in the face of considerable uncertainty and will impact future EU enforcement actions (e.g., more fines), as well as Meta's future EU-related revenues and expenses. This presentation models and optimizes Meta's choice using Bayesian networks and influence diagrams and illustrates how to deal with "functional asymmetry" in designing influence diagrams.
Kurt Schulzke, JD, CPA, CFE, is a Professor of Accounting & Law at the University of North Georgia. His teaching, research, and consulting thrive at the intersection of data science, accounting, law, and risk management. He has published in the Columbia Journal of Transnational Law, Vanderbilt Journal of Transnational Law, Tennessee Journal of Business Law, Journal of Forensic Accounting Research, and The Value Examiner. MAcc (Brigham Young University), J.D. (Georgia State University), M.S. Applied Statistics (Kennesaw State University).
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
To be presented at the in Cincinnati on April 11, 2024.
To be presented at the in Cincinnati on April 11, 2024.
Hana C. Long, Ph.D., NC State University
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 11, 2024.
To better manage emerging public health challenges, communities, water utilities, and regulators must improve methods to assess and prioritize environmental health risks. PFAS exposure in North Carolina is a prime example where methodologies for predicting and mitigating risks are developing concurrently with contaminant regulations and the allocation of mitigation funding. This project aims to predict the risk of private wells exceeding the provisional health goal for the PFAS GenX. PFAS compounds are notoriously difficult to model with mechanistic groundwater flow and fate and transport models. The sheer number of different PFAS chemicals and uncertainty in their individual and interacting characteristics all make them complex to model in the environment. Mechanistic models are also resource-intensive to develop and calibrate. This project builds upon previous work that developed a Machine-Learned Bayesian Network (MLBN) classification model to predict at-risk wells; current work integrates outputs from a mechanistic groundwater fate and transport model as input variables to new MLBNs, classified as low-, medium-, and high-effort models in terms of mechanistic modeling resources required. The performance of each model is compared to the mechanistic model predictions of at-risk wells using several performance metrics, including accuracy, area under the receiver operating characteristic curve (AU-ROC), and F-score curves, and the importance of each metric and model performance is discussed in the context of environmental health risks. Results show that MLBNs perform as well as the mechanistic models in accuracy and AU-ROC performance metrics while being more robust in terms of the range of decision thresholds selected for risk classification. High-effort models make slight improvements in AU-ROC metrics while more easily incorporating insights from mechanistic model performance without the need to recalibrate the mechanistic model. The project aims to assist regulators in advancing public health and methodologies to integrate traditional engineering models with machine-learning approaches.
Hana C. Long is a postdoctoral researcher in the Department of Civil, Construction, and Environmental Engineering at NC State University (NCSU). Her research uses mathematical optimization and statistical modeling to help communities make sustainable and resilient infrastructure decisions. Hana holds a PhD in Operations Research from NCSU. She previously worked as a project engineer in the Wastewater Research Group at the Los Angeles County Sanitation Districts and with the Community Resilience Group at the National Institute of Standards and Technology. She holds a Master's in Civil Engineering (NCSU) and a Bachelor's degree in Mathematics and Russian Language (Vanderbilt University).
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Kurt S. Schulzke, JD, CPA, CFE, University of North Georgia
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
On May 7, 2021, Darkside hackers exploited a leaked Colonial Pipeline Corporation (CPC) password, breaching a dormant VPN to infiltrate CPC’s IT system. Lacking a contingency plan, CPC entirely shuttered its pipelines, which at the time carried 45 percent of all jet fuel and gasoline consumed on the East Coast of the United States. This ransomware hack showcased stereotypical weaknesses in cybersecurity modeling, controls, and compliance monitoring and revealed the company's failure to create a response playbook or contingency plan, as required by U.S. Department of Transportation regulations. This presentation illustrates the use of Bayesian networks and influence diagrams for cybersecurity risk modeling, assessment, ranking, and management and suggests how their use might have prevented the Colonial Pipeline hack and/or mitigated its consequences to the company and other stakeholders.
Kurt Schulzke, JD, CPA, CFE, is a Professor of Accounting & Law at the University of North Georgia. His teaching, research, and consulting thrive at the intersection of data science, accounting, law, and risk management. He has published in the Columbia Journal of Transnational Law, Vanderbilt Journal of Transnational Law, Tennessee Journal of Business Law, Journal of Forensic Accounting Research, and The Value Examiner. MAcc (Brigham Young University), J.D. (Georgia State University), M.S. Applied Statistics (Kennesaw State University).
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Vikram Suresh, Ph.D., University of Cincinnati
Using Bayesian Belief Networks (BBN), variables from three graphs based on Relative Preference Theory (RPT) were integrated. The graphs in RPT have analogies to known graphs in other frameworks for human judgment but are derived from the only framework tested and found to meet physics criteria for lawfulness. The graphs are analogous to those found in Prospect Theory, Markowitz portfolio theory, and the Payne, Bettman, and Johnson adaptive strategy selection. RPT produces nonlinear graphs with R2 fits per person, on average > 0.9. RPT graphs calibrate value based on patterns in prior judgments (value function), limits in risk-reward and risk-aversion relations (limit function), and balance between reward and aversion judgments (tradeoff function). Using the same picture rating task administered to three distinct internet cohorts, each with more than 3,000 subjects, graphs were observed calibrating reward/aversion value (value function), associating risk to reward/aversion (limit function), and balancing reward against aversion (tradeoff function). Fifteen mathematical features of these graphs based on engineering approaches or behavioral finance constructs, such as loss aversion and risk aversion, were computed and then used as input for Bayesian Belief Network analysis and correlational analysis to identify consistent relationships between these fifteen mathematical features. When consistent relationships were observed, we then fit mathematical functions to the combined dataset of 10,000+ subjects. This analysis showed that the calibration of value (prospect theory-like graph) anchors the relationship of risk-reward variables in a distinct manner from how it anchors the relationship of risk to aversion, and four distinct clusterings of these graph features can be observed based on their graph of origin with highly interpretable relationships. The fifteen graph features can be combined and applied to predict the outputs of human judgment using machine learning. Thus, it is possible to create an individual profile or “fingerprint,” which can be used to predict behavior or other psychological conditions such as anxiety, depression, and suicidality. Marketers can use the features as a segmentation variable to identify the best prospects and design the most effective messages.
Vikram Suresh, Ph.D., has been a Postdoctoral Fellow in the Department of Computer Science at the University of Cincinnati since August 2023. He earned his Ph.D. in Business Administration (Quantitative Economics) from the University of Cincinnati in 2023. His research focuses on combining Bayesian hierarchical modeling, AI, and econometric methods to analyze various topics, including treatment outcomes in adolescents with depression, socioeconomic predictors of treatment outcomes in adults with major depressive disorder, and the impact of age on antidepressant response. He has several publications in peer-reviewed journals and is working on manuscripts related to statistical approaches in randomized controlled trials, income inequality estimation, and the decline in high school student performance using AI experiments.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Krishna Ganta, Rohit Warrier, Riley Mulhern, Ted Lillys, Jennifer Hoponick Redmon, Jacqueline MacDonald Gibson
Department of Civil, Construction, and Environmental Engineering, North Carolina State University, Raleigh, NC, USA
Research Triangle Institute International, Research Triangle Park, NC, USA
Brown & Caldwell, Englewood, CO, US
To be presented at the in Cincinnati on April 11, 2024.
Hans Breiter
and Martin Block, Shamal Lalvani, Sumra Bari, Nicole L. Vike, Leandros Stefanopoulos, Byoung-Woo Kim, Aggelos K. Katsaggelos
Medill Integrated Marketing Communications, Northwestern University, Evanston, IL, USA
Department of Electrical and Computer Engineering, Northwestern University, Evanston, IL, USA
Department of Computer Science, University of Cincinnati, Cincinnati, OH, USA
School of Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece
Department of Radiology, Northwestern University, Chicago, IL, USA
Department of Computer Science, Northwestern University, Evanston, IL, USA
Laboratory of Neuroimaging and Genetics, Department of Psychiatry, Massachusetts General Hospital and Harvard School of Medicine, Boston, MA, USA
Abraham Rojas Zuniga, Curtin University
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
Duplex stainless steel (DSS) alloys, recognised for their high mechanical strength and corrosion resistance, are increasingly utilised in the oil and gas industry to mitigate critical degradation risks in downhole environments. Elevated pressures, temperatures, and corrosive agents such as chlorides, carbon dioxide, and hydrogen sulphide characterise these settings. Collectively, these factors contribute to various environmentally assisted cracking mechanisms, predominantly stress corrosion cracking (SCC). This corrosion phenomenon significantly threatens production systems, potentially causing premature failures of metallic materials due to the synergistic effects of tensile stresses and corrosive media. Despite the growing adoption of DSS alloys, their performance in oil and gas applications remains inadequately understood within existing standards, rendering the operational boundaries of DSS perceived as overly conservative. While comprehensive research has explored DSS's resistance to SCC, the reliable assessment of SCC risks in the field remains a significant challenge. Consequently, there is a need for a framework to evaluate, with reasonable certainty, the viability of DSS applications in production systems. We address these limitations by introducing a data-centric approach through Bayesian networks (BNs) for assessing the SCC risks of DSS in downhole environments. We developed this BN model by combining various information sources, including industry standards, technical guidelines, and scientific papers. We used advanced pre-processing techniques, such as data imputation and synthetic minority oversampling, to prepare the dataset adequately. Furthermore, the BN model's structure and predictive accuracy were also compared with other modelling methods, such as XGBoost and SHAP analysis, which provide additional insights into the causality of SCC. More importantly, our BN model demonstrates that the SCC resistance of DSS alloys can comfortably exceed the operational threshold established in standards, currently within 0.02 – 0.2 bar of the partial pressure of hydrogen sulphide.
Abraham Rojas Zuniga abraham.rojaszuniga@postgrad.curtin.edu.au
Sam Bakhtiari sam.bakhtiari@curtin.edu.au
Ke Wang ke.wang2@curtin.edu.au
Chirs Aldrich chris.aldrich@curtin.edu.au
Victor M. Calo victor.calo@curtin.edu.au
Mariano Iannuzzi mariano.iannuzzi@alcoa.com
Curtin Corrosion Centre, Faculty of Science and Engineering, Curtin University.
Western Australian School of Mines: Minerals, Energy and Chemical Engineering, Faculty of Science and Engineering, Curtin University.
Computing and Mathematical Sciences, Faculty of Science and Engineering, Curtin University.
Abraham Rojas Zuniga
As a petroleum engineer with five years of experience, I have advanced my academic career with a Master's degree (M.Phil.) in Oil and Gas Engineering from the University of Western Australia. As a Ph.D. candidate at Curtin University, my research focuses on chemical engineering and artificial intelligence to improve our understanding of corrosion phenomena in hydrocarbon industry alloys. I am keenly interested in applying simulation techniques, ranging from deterministic models to data-driven methods, to investigate material science phenomena and enhance risk assessment strategies.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Rahul Pandey & Anand Wilson, Course5 Intelligence
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 11, 2024.
This study delves into the effectiveness of an integrated, weighted approach in influencing strategic brand decisions, highlighting the constraints of traditional machine learning and data science techniques that rely heavily on large, clean datasets and struggle with complex variable interrelations. By adopting Bayesian structural equation modeling, our research leverages prior knowledge, expert insights, and beliefs, focusing particularly on how various metrics impact market share. Through primary search surveys across different market segments, we assess brand perception, aggregating data from numerous sources over a set timeframe for comprehensive analysis.
The methodology begins with exploratory data analysis, leading to the testing of probabilistic associations to validate hypotheses and develop the model. An unsupervised network graph forms the core of our model, illustrating the interactions between variables from disparate data sources at a specific time. This base model undergoes scrutiny for statistical robustness and output validation, considering both total and direct effect values to determine weighted coefficients. Further refinement of the network model incorporates business and market inputs, ensuring the model's assumptions align with real-world conditions without undermining its stability. This strategic methodology offers nuanced, actionable insights into a brand's market position, enhancing adaptability and strategic decision-making as new information emerges.
Rahul Pandey, AVP Digital and Advanced Analytics at Course5 Intelligence
Rahul is an Applied AI and Data Science leader, experienced in setting up and scaling multi-disciplinary global high-performance applied data science teams in India and the US. Rahul is recognized as LinkedIn's top voice for – Data Science, Artificial Intelligence, GenAI, and Leadership. Rahul has also been awarded twice as “40 under 40 data scientists” in India consecutively in 2023 and 2024. In his current role, Rahul heads the ‘Global Data Science Practice’ at Course5i and has been instrumental in scaling data science practice and solutions at course5i. He has developed multiple solutions, written papers on applications of Generative AI, and presented them at conferences across the globe. He specializes in creating data science strategies to solve problems through thought leadership and the application of advanced algorithms for executive leaders in Fortune 100 and Fortune 500 companies across verticals. He has also developed unanimous trust among industry stalwarts for solving problems that reflect business value.
Anand Wilson, Senior Data Science Consultant, Advanced Analytics and Applied AI, Course5 Intelligence
Anand has over 11 years of experience in applied artificial intelligence and data sciences. He creates market solutions based on Bayesian Network theory, which can quantify causality in observational studies. His work and research areas include Knowledge Modelling and Machine Learning with BayesiaLab. Anand has a background in applied statistics and a keen interest in machine reasoning, causal inference, and experimental design.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Fathien Azuien Yusriza, Universiti Kuala Lumpur, Malaysian Institute of Aviation Technology
This study addresses the crucial purpose of enhancing supply chain (SC) efficiency and effectiveness in the face of unforeseen disruptions, particularly examining the impact of the COVID-19 pandemic on the airline catering sector. With a focus on SC performance assessment, the researchers propose and implement Bayesian network (BN) modeling as a strategic tool to measure and quantify the consequences of pandemic disruptions. The study employs forward and backward propagation analysis within the BN model to simulate and measure the impact of different triggers on SC performance and business continuity. The findings provide a valuable theoretical perspective on the use of BNs in pandemic SC disruption modeling, offering insights that can serve as a decision-making tool for predicting and comprehending the effects of pandemics on SC performance. This research contributes to a better understanding of SC dynamics in the context of unforeseen disruptions and provides a foundation for informed decision-making in the airline catering sector amid uncertainties.
Fathien Azuien Yusriza is a highly skilled professional with a Bachelor's Degree in Aviation Management from Universiti Kuala Lumpur, Malaysian Institute of Aviation Technology (UniKL MIAT), and a Master's degree in Engineering Technology (Aerospace). With a rich background, she has contributed her expertise to global logistics companies, including DHL APSSC and Airbus Malaysia, as a transport analyst. Currently engaged in research focusing on the effectiveness of aviation's supply chain management, she actively contributes to scholarly publications, including journal articles and book chapters, showcasing her commitment to advancing knowledge in the field.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
To be presented at the in Cincinnati on April 12, 2024.
Yong Zhang, Ph.D., Procter & Gamble
Recorded at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
We have been exploring different ways to leverage GenAI and causal inference to disrupt the speed, scope, and economics of product innovation at P&G. This includes LLM like ChatGPT and Mixtral_8X7B, open source packages like Langchain and LlamaIndex and third-party software/module Hellixia. We use these existing tools or build new tools on top of them to understand consumers and drive product innovation. We leverage these tools to generate structure summaries and reports from large volumes of data and to generate qualitative or semiquantitative structural causal models, etc.
Dr. Yong Zhang leverages Bayesian data and modeling science to develop a strategy for product design, manufacturing, storage, and transportation across P&G to improve consumers’ quality of life and drive positive influence on the environment and society under different climate change scenarios. He develops modeling and simulation methods and tools through Front End Innovation projects to enable and promote the capability across P&G for breakthrough consumer understanding and product innovation. The methods and tools can be used to extract and integrate information from a variety of data sources to find a “Body of Evidence” for consumer and product research based on Nonparametric Bayesian statistics and deep learning algorithms.
Emmanuel Keita, Sundiata
I propose a 3,000-year journey through time and space, from the earliest traces of oracular practices and the I Ching to the current use of so-called “generative (probabilistic) artificial intelligence (modeling)” with Bayesialab as a fabulous pedagogical tool.
Not forgetting God and a game of dice, this talk will question our perceptions of the world, our relationship to chance, and, by the way, our attitude to the eternal question of how to make (good...) decisions as humans in an ever-changing world 易!
Emmanuel KEITA is a consultant, keynote speaker, and trainer.
An advocate of "being more human in a woLRd of machines” he offers a unique perspective on decision-making at the intersection of technology (AI, BI, MDM, Process Mining, etc.), cognitive science, and traditional knowledge.
Coaching anyone facing stressful performance, fatigue, or motivational situations (human potential optimization techniques - TOP 2024), Emmanuel is also a qi gong instructor interested in the "study and reasoned use" of the I Ching as a reflexive and holistic strategic tool.
Independent expert for Collège Numérique FRANCE 2030, Emmanuel is also a National Defense Auditor (France).
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Steven F. Wilson, Ph.D., EcoLogic Research
Environmental policymaking is challenging because systems are complex, and rarely can we conduct experiments to test the relative costs and benefits of different policy options. Causal analysis methods allow us to estimate causal effects from observational data, and such methods are being applied increasingly often to predict the relative benefits of alternative policies. However, predictions based on only average causal effects provide an incomplete assessment of the value of potential interventions. Decision-makers also need to know how likely an outcome is to occur without the intervention (i.e., an assessment of causal attribution) or what outcomes could be expected if the intervention was only selectively applied (i.e., estimating context-specific causal effects). Answering these questions requires applying the counterfactual reasoning of “rung 3” of Pearl’s causal hierarchy. In fact, Pearl argued explicitly in his book Causality that “policy analysis is an exercise in counterfactual reasoning.” I used Bayesian Networks to model counterfactual outcomes on caribou populations of different land use policy interventions. While there are theoretical limitations to using Bayesian Networks for this purpose, the resulting counterfactual insights still provide additional value to decision-makers compared to observational or interventional analyses.
Steve Wilson has 30 years of experience working at technical and professional levels in strategic and operational planning for wildlife and other ecological values. He specializes in quantitative approaches to decision support and policy analysis. Steve holds a Ph.D. in wildlife ecology from the University of British Columbia in Vancouver.
The Social Graph—Using Bayesian Networks to Identify Spatial Population Structuring Among Caribou in Subarctic Canada (Paris, 2017)
Use of causal modeling with Bayesian networks to inform policy options for sustainable resource management (Nashville, 2016)
Big data, small data: Bayesian networks in environmental policy analysis in Canada’s energy sector (Fairfax, 2015)
To be presented at the in Cincinnati on April 12, 2024.
Presented at the in Cincinnati on April 12, 2024.
Steven F. Wilson, Ph.D., EcoLogic Research, 302-99 Chapel Street, Nanaimo, BC V9R 5H3, Canada,
John Carriger, Ph.D., U.S. Environmental Protection Agency
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
Causal structural models are used to capture knowledge of a problem domain through framing potential events as random variables and connections as causal arcs. Moreover, the properties of causal structural models foster additional insights on causal and inferential interactions among variables from interventions and observations. Conceptual site models are commonly used in environmental assessments for capturing the knowledge of the fate, transport, and risks at contaminated sites and form the basis for simulation models. The usage of causal structural models with conceptual site modeling may provide additional value for site remediation and assessments. We call this combination conceptual Bayesian networks (CBNs) and explore their application potential in contaminated site management for assessing the subsurface movement of contaminated plumes. Once constructed, the CBN can capture the hypothesized locations and movements of a plume as well as critical zones of offsite flux. Causal pathway identification can examine offsite transport pathways and the potential effects of remediation decisions that intervene on those pathways. Interventions for containing or removing subsurface contamination and breaking the transport pathways are graphically represented as decision nodes. Finally, measurement node types can explicitly include lines of evidence for subsurface processes in the CBN. Acausal pathways from influence paths provide additional information on statistical inferences when lines of evidence are observed individually or in conjunction. The CBN concept may provide additional insights beyond traditional conceptual site models and could be a valuable component in a site manager’s toolbox.
The views expressed in this presentation are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.
John F. Carriger, Michael C. Brooks, Carolyn Acheson, Ronald Herrmann, Lee Rhea
US Environmental Protection Agency, Office of Research and Development, Center for Environmental Solutions and Emergency Response, Land Remediation and Technology Division, Cincinnati, OH
US Environmental Protection Agency, Office of Research and Development, Center for Environmental Solutions and Emergency Response, Groundwater Characterization and Remediation Division, Ada, OK
US Environmental Protection Agency, Office of Research and Development, Center for Environmental Solutions and Emergency Response, Land Remediation and Technology Division, Cincinnati, OH (retired)
John Carriger is a research scientist at the U.S. Environmental Protection Agency’s Office of Research and Development in Cincinnati, Ohio. John has a marine science Ph.D. from the College of William and Mary. John’s research interests include applying risk assessment, decision analysis, and weight of evidence tools to environmental problems.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Steven Frazier, Georgia Pacific
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
This presentation explores the application of machine learning and Bayesian networks to enhance tissue softness, a crucial factor in the paper product industry. It begins with the presenter's extensive background in engineering and analytics, setting the stage for a deep dive into utilizing Bayesian networks to improve tissue manufacturing processes.
The focus is on refining the balance between paper strength and softness, crucial for producing high-quality tissue. The presentation covers the technicalities of tissue production, from fiber refining to the mechanics of papermaking, illustrating the complex interplay of factors affecting product quality.
A significant portion is dedicated to the evolution of model development, from initial challenges to advanced iterations that accurately predict tissue softness. Techniques like Jackknife and K-Fold cross-validation are discussed for model evaluation, highlighting the learning curve and adjustments made to enhance model performance.
Operational insights form the core of the latter part, where data analysis reveals optimal manufacturing conditions. The presentation touches on the importance of data integrity, model adaptability, and the role of human operators in implementing AI-driven recommendations.
Concluding, the presentation reflects on the project's broader impacts, emphasizing continuous improvement, user readiness, and aligning project goals with customer expectations. This summary encapsulates the journey and lessons learned in applying advanced analytics to improve tissue softness, underscoring the potential of machine learning in industrial applications.
With over 30 years of operations experience and 3 U.S. patents, Steven Frazier is an expert in creating high-performance solutions for complex business challenges. His multi-disciplinary approach has been pivotal in enhancing processes for leading Fortune 500 companies, including Coca-Cola, Procter & Gamble, and Georgia Pacific.
Steven specializes in Bayesian Networks and machine learning, providing critical insights that guide manufacturing efficiency and strategic value creation. His innovative modeling techniques have informed smarter procurement strategies, yielding substantial cost savings and process enhancements.
During his recent role at OnPoint (Koch Industries), Steven's models for tissue softness and strength revealed significant opportunities for value creation, contributing to both product and process improvements.
As he prepares to share his expertise at the Bayesialab conference on April 11, 2024, Steven's pragmatic and transformative approach is expected to resonate with a wide audience. With a solid educational foundation in Mechanical Engineering from Washington University and advanced certification in Lean Six Sigma from Villanova University, he's geared to propel organizations towards operational excellence and analytical innovation. A presentation you don’t want to miss.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Join us in Cincinnati and bring to life your work with Bayesian networks on our stage. The 2024 BayesiaLab Spring Conference promises rich insights into the latest in applied research using probabilistic graphic models and Bayesian networks. Besides, through enlightening courses and talks, immerse yourself in best practices with the BayesiaLab software platform.
Your platform choice, whether BayesiaLab or another, doesn't restrict your participation. We welcome entries from academia, government, industry, and solution providers, focusing on new applications, ongoing work, and capabilities within Bayesian networks.
Date: April 11–12, 2024 Slot Duration: 30–45 minutes (including 5–15 minutes of Q&A)
Title
Presenter's name & affiliation
Abstract (max. 300 words)
High-resolution presenter photos (min. 500x500 pixels)
Presenter biography (max. 100 words per presenter)
Proposed format (e.g., PowerPoint, Keynote, Prezi)
Accepted abstracts will be showcased on the Bayesia website before the conference. All presentations will be recorded. By submitting, you consent to your presentation and its recording being shared on the Bayesia website and related social media platforms.
All accepted presenters will enjoy a waived conference registration fee.
Direct your presentation proposals to conference@bayesia.us.
If you are attending a BayesiaLab Conference for the first time, please check out the archives from previous conferences.
Edwin Hui, University of St Andrews
Presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
Understanding the dynamics that regulate ecological resilience is becoming increasingly important in today’s world, as ecosystems face multiple global, regional, and local pressures. If pressures exceed a threshold, this may trigger a regime shift, where a system undergoes a step change to another state that can last for substantial periods of time. However, modeling such change is not simple as ecological data is scarce, and models often assume that relationships within ecosystems remain homogenous over time. In this talk, we document the application of non-homogenous Dynamic Bayesian Networks to various complex systems known to have undergone major structural changes.
Edwin Hui is a Ph.D. student from the University of St Andrews, where his research focuses on developing computational models to study resilience and regime shifts across complex systems. He is interested in applying a variety of statistical and computational tools to address ecological questions and study complex systems theory. Throughout his Ph.D., he aims to develop novel computational approaches to study complex systems across different disciplines, ranging from ecological to macroeconomic systems.
Alexander Alexeev, Ph.D. & Rafael Reuveny, Ph.D., Indiana University - Bloomington
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
Although in multiple episodes, Weather Disasters (WDs) are found to be potentially important factors promoting migration between countries and regions, it is unclear if their role is systematic or idiosyncratic. This issue is of great public attention and importance since the severity, frequency, and coverage of WDs are expected to grow as climate change progresses under business as usual. Recent studies suggest that large migration flows may be associated with violence, economic and demographic factors in the countries of origin while paying less attention to the effect of socio-economic conditions in the destination countries. Our research has twofold objectives. First, this study develops a Bayesian Network Analytic (BNA) framework that anticipates the potential for varied migration responses to WDs across countries and over time, and examines policy levers that might alter these responses, the complex interaction between different factors, and overcomes some limitations of econometric models routinely used in modeling of international migration. The network structure is learned from data for migration flows between 190 origins and 190 destinations from 1980 to 2009. Secondly, we compare and discuss the advantages, disadvantages, and results of both BNA and conventional econometric modeling, which were previously done on the same data.
Alexander Alexeev earned a Ph.D. in Public Affairs from Indiana University in 2010, with specializations in policy analysis and business economics. He also holds a Ph.D. in Physics from Odessa National University (Ukraine, 1996). Starting in 1997, Alexander taught physics, modeling, and radioecology for the Department of Physics at Odessa Hydrometeorological Institute. In 2001, he came to the United States to study environmental management and stayed for doctoral study. Since 2017, Alexeev has been a lecturer at Indiana University, teaching data analysis and statistical modeling courses. His interdisciplinary research interests include quantitative policy analysis, risk and security modeling, and decision-making.
Ibon Galparsoro Iza, Ph.D., AZTI · Marine Research Division
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
Economic activities are dependent upon natural capital (NC), which is responsible for 'Ecosystem Services' (ES). Understanding dependencies on NC provides insight into the ecosystem's capacity to maintain and develop activities in the future. To determine 'NC dependencies', we present a framework linking maritime activities (bottom trawling, artisanal fisheries, aquaculture, and tourism) to their demand for ES and, further, to the NC components responsible for their production. The framework was operationalized using a spatially explicit Bayesian Belief Network (BBN), using the Basque coast (SE Bay of Biscay) to illustrate our approach to identifying trends in the strength and spatial distribution of NC dependencies. For example, benthic trawling was dependent on sedimentary habitats, with a 'moderate' to 'high' dependency of 52% of the study area. The model can also extrapolate NC dependencies to a larger area where the activity currently does not operate, where benthic trawling was estimated to have higher utilization of ES in deeper waters. When NC dependencies are combined with economic and legislative factors, the current spatial distribution of the activity can be explained, and the potential socioeconomic impacts of management decisions could be predicted. The integrative approach contributes towards ecosystem-based spatial planning.
Ibon Galparsoro Iza, Ph.D., Principal Researcher. Marine and Coastal Environmental Management AZTI · Marine Research Division
PhD in Marine Sciences from the University of Vigo. Principal Researcher at AZTI’s Marine Research Unit. He has more than 15 years of professional experience in different lines of marine research applied to Integrated Coastal Zone Management. His main research interests include Marine Spatial Planning, assessment and mapping of marine and coastal ecosystem services, implementation of the European Marine Strategy Directive, seabed mapping and characterization, and modeling of benthic habitats.
Register here for the 2024 BayesiaLab Spring Conference, April 11-12, 2024:
Ray Niaura, Ph.D. & Shu Xu, Ph.D., New York University
To be presented at the 2024 BayesiaLab Spring Conference in Cincinnati on April 12, 2024.
There are approximately 1.1 billion tobacco smokers in the world who either suffer from or are at risk of smoking-related diseases such as cancers and cardiorespiratory conditions. Smokers who quit, especially at earlier ages, will gain back several years of life from smoking-related premature mortality. Quitting smoking, however, is difficult despite the availability of evidence-based quit methods, including medications and behavioral counseling. Electronic cigarettes (e-cigs) are being used by an increasing number of smokers, and the question arises whether e-cigs can help tobacco smokers quit. Data from randomized controlled trials (RCTs) suggests that e-cigs can serve as a substitute for cigarettes and are more effective than nicotine replacement therapies (e.g., nicotine patches, gum, lozenges) for smoking cessation. Data from observational studies are less clear about the association between the use of e-cigs and quitting smoking. We present results from two nationally representative US survey studies in which we examine factors that are associated with cigarette smoking and the use of e-cigs and whether the use of e-cigs is associated with quitting smoking (National Health Interview Survey – NHIS; Population Assessment of Tobacco and Health – PATH). Unsupervised BN learning of the 2022 NHIS survey showed a clustering of factors normally associated with cigarette smoking, including poverty and educational levels, depression, anxiety, disability, alcohol consumption, sexual orientation, nativity, and veteran status. There were only two direct paths from educational attainment and nativity to smoking status. E-cig use was directly associated with smoking status and age, suggesting that the determinants of use differ somewhat between cigarettes and e-cigs. Longitudinal data from the PATH study set up as a cross-lagged BN model, showed negative associations between cigarette and e-cig use at each successive year across 6 years, indicating a cumulative impact of e-cig use on smoking cessation. We will discuss challenges and options regarding longitudinal data analysis via Bayesian networks.
Dr. Niaura is a Professor of Social and Behavioral Sciences and Epidemiology and Chair of the Department of Epidemiology at the School of Global Public Health, New York University. From 2009-2017, he was Director of Research at the Schroeder Institute, Truth Initiative (formerly the Legacy Foundation) in Washington, DC. He has extensive expertise in tobacco dependence and treatment, and he has published over 400 peer-reviewed articles and several book chapters in this area. His interests include studying the biobehavioral substrates of tobacco dependence, evaluating behavioral and pharmacological treatments for cessation, and understanding and addressing public health disparities in tobacco-related burdens of illness and disability. He has been the Principal Investigator (PI) or co-investigator of over 70 NIH-funded grants, and he is the former President of the Society of Nicotine and Tobacco Research. He is currently a co-I of a large, multicenter initiative: the Population Assessment of Tobacco and Health Study (PATH, funded by the National Institute on Drug Abuse/Center for Tobacco Products, FDA), a national, longitudinal cohort study of more than 40,000 users and non-users of tobacco products ages 12+, including adolescents and young adults.
Shu (Violet) Xu, Ph.D. sx5@nyu.edu Clinical Assistant Professor Department of Biostatistics School of Global Public Health New York University
My work represents a balance of both statistical and applied aspects of quantitative methodology. My primary quantitative interests include evaluating and developing statistical methods for longitudinal data analysis. Specifically, My research focuses on various aspects of latent growth models, missing data methods, and causal inference models.
I have served as an Investigator/Biostatistician on more than 10 federally or locally funded research projects. I was PI of an NIH/NCI supplement award through the University of Michigan/Georgetown Center for the Assessment of the Public Health Impact of Tobacco Regulations (3U54CA229974), and the project aimed to examine the longitudinal effect of e-cigarette exposure on subsequent tobacco use patterns using conventional and causal mediation methods. I was also a co-investigator of an NIH/NCI R21 award (1R21CA260423-01). This project aims to assess the longitudinal impact of e-cigarette flavor, device, and marketing exposure on tobacco use and health outcomes using propensity score weighting and causal mediation methods. I am PI of an on-going NIH NIDA/FDA K01 award (1K01DA058408). This project aims to develop and implement causal machine-learning methods to inform tobacco regulatory sciences.
I have collaborated with substance use, family, and health researchers to advance and share my knowledge of quantitative methodology and pursue a better understanding of the social sciences and public health. I have conducted research with the Family Translational Research Group at New York University and the Methodology Center at Pennsylvania State University.
Bayesian Network Analysis of Cigarette Smoking and E-cigarette Use in U.S. Population Samples
Transforming Paper Product Quality and Machine Performance with Machine Learning & Bayesian Networks
Ray Niaura New York University
Lionel Jouffe Bayesia S.A.S.
Hellixia & Generative AI: Creating Semantic and Causal Bayesian Networks for Decision Support Video Available
Kurt Schulzke University of North Georgia
Martin Block Northwestern University
Kurt Schulzke University of North Georgia
Emmanuel Keita Sundiata
Vikram Suresh University of Cincinnati
Hana C. Long NC State University
Gabriel Andraos Voya Financial
Steven F. Wilson EcoLogic Research
Fathien Azuien Yusriza Airbus Malaysia
John Carriger U.S. EPA
Anand Wilson Course5 Intelligence
Rahul Pandey Course5 Intelligence
Yong Zhang Procter & Gamble
Leverage GenAI and Causal Inference to Disrupt Innovation Video Available
Abraham Rojas Zuniga Curtin University
Edwin Hui University of St. Andrews
An Application of Dynamic Bayesian Networks to Model Regime Shifts and Changepoint Processes Video Available
Alexander Alexeev Indiana University
Steven Frazier Georgia Pacific
Shu Xu New York University
.
.