Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Alta de Waal, Ph.D., University of Pretoria
Presented at the 8th Annual BayesiaLab Conference on October 27, 2020.
While activity-based travel demand generation has improved over the last few decades, the behavioural richness and intuitive interpretation remain challenging. We argue that it is essential to understand why people travel the way they do and not only be able to predict the overall activity patterns accurately. If one cannot understand the ‘why?” then a model's ability to evaluate the impact of future interventions is severely diminished. Bayesian networks (BNs) provide the ability to investigate causality and is showing value in recent literature to generate synthetic populations. This research is novel in extending the application of BNs to daily activity tours. Results show that BNs can synthesise both activity and trip chain structures accurately. It outperforms a frequentist approach and can cater for infrequently observed activity patterns, and patterns unobserved in small sample data. It can also account for temporal variables like activity duration.
Alta de Waal, Ph.D. Centre for Artificial Intelligence Research Department of Statistics, Faculty of Natural and Agricultural Sciences University of Pretoria, South Africa
Alta currently holds a senior lecturer position in the Department of Statistics, University of Pretoria, South Africa. She has 20 years of experience in the design, development, and implementation of different components in the AI value chain. She develops Bayesian network models in application areas such as student throughput models, wildlife security, environmental risk management, and transportation. Alta also studies natural language processing (NLP) with a special interest in probabilistic distributional semantic methods.
Spatially Discrete Probability Maps for Anti-Poaching Efforts (Paris, 2017)
John Carriger, Ph.D., U.S. Environmental Protection Agency
Presented at the 8th Annual BayesiaLab Conference on October 27, 2020.
Coral reefs are highly valued ecosystems currently threatened by both local and global stressors. Given the importance of coral reef ecosystems, a Bayesian network approach can benefit an evaluation of threats to reef conditions by revealing details about the relationships between variables. To this end, we used available data to evaluate the overlap between local stressors (overfishing, watershed-based pollution, marine-based pollution, and coastal development threats), global stressors (acidification and thermal stress) and management effectiveness with indicators of coral reef health (live coral index, live coral cover, population bleaching, colony bleaching and recently killed corals). We constructed Bayesian networks using available data for each coral health indicator both globally and for specified regions (Pacific, Atlantic, Australia, Middle East, Indian Ocean, and Southeast Asia). Sensitivity analysis helped evaluate the strength of the relationships between different stressors and reef condition indicators. Management effectiveness was also examined for directionality and strength of relationships. The relationships between indicators and stressors were evaluated with conditional analyses of linear and nonlinear interactions. This process used standardized direct effects and target mean analyses to predict changes in the mean value of the reef indicator from individual changes to the distribution of the predictor variables. The standardized direct effects analysis identified higher potential risks between coral reef indicators and stressors in and across regions when relationships approximated linearity. Additional measures, including the minimums and maximums of the target mean analysis, were used to support the relationship analysis. The Bayesian network approach helped characterize relationships among indicators used for coral reef management by examining the sensitivity of reef condition indicators to indicators of threats and management effectiveness.
EPA Disclaimer: 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 (carriger.john@epa.gov), Susan H. Yee, William S. Fisher
U.S. Environmental Protection Agency, Office of Research and Development, Center for Environmental Solutions and Emergency Response, Land Remediation and Technology Division, Environmental Decision Analytics Branch
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.
Steven F. Wilson, Ph.D., EcoLogic Research
Presented at the 8th Annual BayesiaLab Conference on October 26, 2020.
As global conservation actions become more urgent, informed decision-making requires robust analyses of the costs and benefits of policy options, based on available evidence. Recovery planning for endangered species must assume a cause-and-effect relationship between proposed management interventions and population responses; however, most current ecological knowledge is derived from observational studies because experiments are largely infeasible or unethical. Weak and conflicting inferences about causal mechanisms have created debate and confusion among decision-makers, planners and stakeholders. While causal modelling techniques are well-developed and common in other policy domains that face similar challenges, the approach is nearly absent in conservation biology. I examine the challenge of woodland caribou recovery efforts in Canada through the lens of causal modelling, highlighting recent, high-profile debates and illustrating how a causal modelling approach can help to bring resolution while supporting robust forecasting and decision support.
Steven F. Wilson, Ph.D., EcoLogic Research, 302-99 Chapel Street, Nanaimo, BC V9R 5H3, Canada, steven.wilson@ecologicresearch.ca
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.
Using Bayesian Networks to Characterize Wildlife Habitat Use (Chicago, 2018)
The Social Graph—Using Bayesian Networks to Identify Spatial Population Structuring Among Caribou in Subarctic Canada (Paris, 2017)
Big data, small data: Bayesian networks in environmental policy analysis in Canada’s energy sector (Fairfax, 2015)
Use of causal modeling with Bayesian networks to inform policy options for sustainable resource management (Nashville, 2016)
Raphael Girod, MAHA
The immunization approach, Expanded Programme on Immunization (EPI), is a powerful public health strategy for improving child survival; the policy for any Ministry of Health is to ensure that as many children as possible receive the full series of vaccines on their national routine immunization schedule.
The overarching goal of that study is to demonstrate (i) the usefulness of the data mining and modeling approaches in the context of an immunization program and, for the communication aspect, (ii) how mapping and analytics capabilities could be used so as to brainstorm and communicate the proposed actionable insights among the decision-makers in mandated coordination bodies at national and regional levels.
The EPI Bayesian Network Model for Ghana displays the available variables (data elements extracted from the District Health Information & Management System) and the arcs, which have been “manually” laid down, thanks to the theory of change, to justify causal assumptions (or the linkages among the variables in the EPI model). Indeed, Public health knowledge is key.
Out of these data elements, the number of children vaccinated and the number for the three types of vaccination sessions (fixed, outreach, and at school) are considered specifically.
Thanks to the optimization algorithm, it is possible to lay down that the best solution gears towards increasing the number of fixed and school vaccinations sessions and, lowering the number of outreach vaccinations sessions, but because of the several contextual factors to be considered, any realistic and meaningful concrete decisions should be taken only at district or sub-district levels.
Whatever the diversity and the complexity of the local situation at the sub-district level, we take recourse of the “batch outlier” procedure so as to come up with priority actions.
A Bayesian network can serve as an inference engine, and thus simulate that public health program comprehensively. Through simulation, we can obtain all associations that exist in the EPI program, and, most importantly, we can compute causal effects directly. Overall, strategies to improve vaccination must be percolated top-down up to sub-districts and communities.
Raphael is a project manager by experience, he gained public health skills as (i) health expert in charge of many result-oriented monitoring missions in Asia and Africa, (ii) as health project coordinator based at the Ministry of Health in Guinea, and (iii) as Local Fund Agent Project leader for Global Fund in Burkina-Faso, all these complementary experiences enables him to fully appraise the strategic, financial, epidemiological and impact evaluation stakes in the process of providing reliable information for high-quality programming…The most updated capability relates to modeling and data-mining, thanks to BayesiaLab.
Passionate about data analysis and public health, his current work aims at collating data sources in order to structure complex datasets, to inform indicator measurements in order to support the strengthening of knowledge management related to global health, especially in regards to immunization programs.
Presented at the on October 26, 2020.
Raphael Girod is the founder of the MAHA organization. MAHA stands for .
Michael L. Thompson, Ph.D.,
Presented at the 8th Annual BayesiaLab Conference on October 27, 2020.
Recommender systems are some of the most useful business applications built using Machine Learning. In our talk, we demonstrate how to build a recommender system for movies using Bayesian Machine Learning. Moreover, the unique features of BayesiaLab, like “Most Relevant Explanation” and “Evidence Instantiation”, allow us to extend the recommender system so we can gain insights into the audiences of each movie. Yet, we ask for more! We suggest extensions to BayesiaLab’s already powerful feature set.
Dr. Michael L. Thompson is retired from the Procter & Gamble Company, where he led Bayesian Analysis R&D in consumer & market modeling. His degrees are in Chemical Engineering: B.S., Northwestern University, ’82; M.S., MIT, ’84; and Ph.D., MIT, ’96, with a minor in Statistics and Artificial Intelligence. Michael has extensive experience in the process industry, having worked for Dow, Alcoa, Amoco, and Mitsubishi Chemical (Japan). At P&G for 21 years, Michael applied his expertise in Bayesian Analysis, especially Bayesian belief networks (BBN), to deliver results in the consumer-packaged goods (CPG) industry. His contributions spanned business functions, including R&D, Engineering, Manufacturing, Marketing, and Business Analytics. He has authored journal articles ranging from fluidized bed reactors to hybrid probabilistic and first-principles biochemical models to optimal consumer product design. Currently, Michael is a Term Adjunct in the Lindner College of Business at the University of Cincinnati, where he teaches Bayesian Analysis to candidates for the Master of Science in Business Analytics. He also serves on the Advisory Board for the Retail AI Lab of the Northwestern University Retail Analytics Council.
Bayesian Sense-Making in Data Science (Chicago, 2018)
Bayesian, Bayesia, BayesiaLab … and P&G (Orlando, 2013)
Corey Neskey
Presented at the 8th Annual BayesiaLab Conference on October 29, 2020.
Risk assessment is challenging when data is unavailable, hard to obtain, or costly to process. Organizations often request estimates from experts instead. I present an implementation of the Modified Beta PERT distribution within a Bayesian network that facilitates such expert knowledge elicitation and can be used as part of more sophisticated Bayesian networks that also incorporate real and estimated datasets.
Corey Neskey has been providing analyses, architecting secure environments, and leading security program implementations in IT security and risk since 2011. His career started with informing executive decision-making using algebraic data analyses for explanation, simulation, and attribution (i.e., intelligence analysis, forensics, SOC, CIRT), and optimization. His toolset expanded to more descriptive and predictive methods (i.e., machine learning/AI for risk assessment, vulnerability prioritization, and event correlation). He is now developing skills for integrating these analytical areas and expanding beyond algebraic methods and static probability calculus to using Bayesian network models.
Sri Srikanth, CISCO
As a B2B company, Cisco likes to understand its overall Mindshare as a precursor to eventual market share in specific technology areas it sells into. The type of information that flows into Mindshare tends to be sparse and not frequently collected, and relying on any single measure leads to incorrect conclusions regarding Cisco’s overall Mindshare. Learn how the team of data scientists and analysts turned to BayesiaLab to unearth the right way to combine this sparse data, as well as glean suggestions for optimization of the underlying measures to increase mindshare and, eventually, market share.
Sri Srikanth, srisrik@cisco.com Advanced Data Analytics & Strategy Global Insights & Analytics Cisco
Viswanath (Sri) Srikanth is the Advanced Analytics Manager for Cisco’s Global Marketing Analytics team, Cisco Marketing. During his tenure at Cisco, he has led data initiatives in understanding and defining Customer Engagement, Marketing Attribution, Customer Journey, and more. His work received multiple industry recognition, including the ANNY Award in 2017, the ANA Award in 2018, and the Highly Commended DRUM Citation in 2018. Before joining Cisco, Sri worked at IBM and, among other things, chaired the creation of an industry standard for customer data collection at the W3C standards organization.
Driving Digital Customer Engagement Powered by Bayesian Networks (Nashville, 2016)
Geo-intelligence remote sensing platforms situated over spatially diverse areas are often tasked with geo-intelligence surveillance and adversarial monitoring for military organizations. Limited resources disallow continuous sampling of local areas at the same time, necessitating a need for smart sensing of diverse environments according to a rational evidence-based rule. Such algorithms should not only provide insight into which local region should be focused on, but should also facilitate decisions as to which environmental features should be measured over time once a local site has been selected. Multi-component optimal learning observational arrays are demonstrated using numerically simulated data of turbulent flow to show not only the feasibility of how individual observational platforms should be chosen in a Bayesian sense but also how goal state-directed sampling of complex systems or turbulent processes over local regions can be accomplished. A Bayesian amalgamation algorithm guides which observational arrays perform knowledge gradient policy based optimal learning to smartly sample observations in local regions. Machine learning and operations research algorithms function as data agnostic, Bayesian processors demonstrating how geo-intelligence information can be efficiently captured to help solve data-driven problems.
Dr. Nicholas Scott is a modeling scientist and physical oceanographer and has been a member of the professional staff at Riverside Research in Dayton, OH, since October 2012. He investigates the applicability of traditional and non-traditional signal and image processing techniques to the extraction of information from remotely sensed imagery. This includes hyperspectral and multispectral imagery. His present work includes statistical modeling of geo-intelligence information, sensor array time series analysis of environmental data, and the application of pattern recognition techniques to turbulent flow imagery and numerically simulated data. He is also involved in the application of probabilistic graphical modeling algorithms for information fusion and statistical inference.
Bayesian Structural Field Analysis (Durham, 2019)
Bayesian Network Modeling of Imagery Features From Direct Numerically Simulated Turbulent Sediment-Laden Oscillatory Flow (Chicago, 2018)
In this presentation, I briefly discuss the methodological approaches used in the teaching of modern data analysis for policy analysis and related majors at the intermediate/upper-intermediate level. Particular focus will be on the development of a conceptual bridge between traditional econometric methods, which remain the primary statistical tools in policy analysis, and modern methods of data analysis – Statistical Learning via Bayesian Networks. Finally, I will present several examples of using the BayesiaLab software package in empirical studies in different fields of policy analysis: international migration, cybersecurity, and environmental policy.
Alexander Alexeev, Ph.D. Lecturer in Data Analysis and Modeling Paul H. O’Neill School of Public and Environmental Affairs Indiana Unversity
Presented at the on October 30, 2020.
Nicholas Scott, Ph.D. Principal Machine Learning Scientist Open Innovation Center Riverside Research
Today, brands are out of sync with customers’ needs and preferences when it comes to personalized marketing. The focus of the talk is to highlight how Course5 is developing a solution leveraging causal discovery framework to develop ‘Customer Preference Sequencing’ - a customer genome which can generate hyper-personalized recommendations of Product, Price, Promotion & Messaging & communication to ensure brands engagement with their customers stay relevant with customer’s changing needs & preferences.
In my talk, I demonstrate how Course5 is disrupting the space of hyper-personalization with its new solution offering and why it is important for industry to consider and adopt such solution in challenging times we are in.
Zabi comes with 15+ years of experience in applied artificial intelligence and data sciences. He has worked for marque clients such as Lenovo, Intel, Microsoft, YouTube, Del Monte, Wrigley, T-Mobile, etc. enabling digital transformation using A.I.
Zabi in his current role leads applied artificial Intelligence and data science practice with Course5 Intelligence. In his previous stint with other companies, he has experience in designing and executing machine learning models and developed teams in niche solutions.
Zabi comes from applied statistics background. He has a master's in statistics and recently selected as Top 40 data scientists in India by Analytics India Magazine. Zabi carries acute interest in machine reasoning, causal inference, experimental designs along with machine learning and data science.
Knowledge Elicitation & Application (Durham, 2019)
Presented at the on October 30, 2020.
Zabi Ulla S, Sr. Director Advanced Analytics,
Presented at the 8th Annual BayesiaLab Conference on October 30, 2020.
This talk gives two examples of how data-driven Bayesian Networks were applied to fisheries datasets to untangle and represent very complex systems. For the first example, I used dynamic Bayesian networks to explore and interrogate how environmental covariates, from the past and present, may drive the abundance and recruitment of mud crab in regions of the Gulf of Carpentaria in Northern Australia. Data used included weather observations and commercial logbooks from mud-crabbers.
For the second example, I modelled the associations of species that were caught by commercial fishers in regions off the east coast of Queensland, Australia.
Challenges I encountered, and insights I made during the modelling process will be discussed.
Amanda Northrop, Senior Fisheries Scientist Department of Agriculture and Fisheries Queensland Government amanda.northrop@daf.qld.gov.au www.daf.qld.gov.au
I am a data scientist with over 15 years of experience in a variety of industries both in Australia and abroad. I have gained knowledge in industry, academia and government. My current role is in Queensland with the Australian Government, which involves building mathematical models of fish stocks population dynamics, and calibrating the models using observed data. Some of my previous employers include Procter and Gamble, Australian National University and the Transport Accident Commission. I am passionate about good science and translating statistical concepts for non-experts.
The natural resource management domain sits at the intersection of incomplete knowledge of the natural world and diverse and often competing societal expectations. While governments strive to develop policy that is ‘evidence-based’ achieving that goal can be challenging. Multiple sources of evidence (e.g. expert knowledge), and competing interpretations of that evidence, can lead to disagreements and ineffective policy. This presentation describes an example of using Bayesian belief networks to combine expert and empirical knowledge to develop forest management policy on public lands in Ontario, Canada. Expert knowledge based BBNs were developed through a series of workshops, validated and updated using literature and monitoring data, from which sensitivity and scenario analysis were used to extract policy options. While this approach to policy development required an initial time investment above the norm the result has paid dividends including detailed documentation of the knowledge base for future revisions, increased transparency and public acceptance of the policy development process, and an objective means to identify key uncertainties to apply limited research and monitoring resources. Current and future development focuses on machine learning from research and monitoring databases.
Mike Brienesse is a Policy Advisor with the Ministry of Natural Resources and Forestry in Sault Ste. Marie, Ontario, Canada. He has worked for both government and forest industry in diverse roles including forest operations, management planning, resource analysis, and policy development. He graduated with an HBScF from Lakehead University, completed an executive MBA with McGill University, and holds a certificate in Strategic Decisions and Risk Management from Stanford University. His professional interests include knowledge systems, decision analysis, and non-timber forest products.
John F. Carriger, Ph.D. & Randy A. Parker, Ph.D., U.S. Environmental Protection Agency
In contaminated site assessments, knowledge of the direct and indirect factors related to stressor impacts on individuals, populations, and communities of organisms is used for designing alternatives that manage or remediate ecological risks. The ecological risk assessment (ERA) framework (USEPA, 1998) provides a logical approach that can help generate this type of information. One key product of the problem formulation step in an ERA is the conceptual site model (CSM). The CSM is a graphical depiction of the risk environment that traces the fate and transport pathways of contaminants from sources of contamination (e.g., a leaking storage tank) to receptors (i.e., the ecological endpoints of concern in the risk problem). The CSM guides the development of methods for assessing ecological risk scenarios and for remediation alternative design. The qualitative and quantitative aspects of Bayesian networks may support CSM development and risk characterization. In particular, the diagrammatic representation from the qualitative aspects of causal Bayesian networks (i.e., the directed acyclic graphs) adds explanatory depth for developing the evidence base for risk characterization and remediation interventions. The components of conceptual Bayesian networks can be used to represent the characteristics and measures of the risk factors. The connections help to decompose, piece together, and explore the potential relationships that bring about high-risk scenarios. Causal pathway analysis of the conceptual Bayesian networks provides visualizations of potential exposure pathways from initial and intermediate sources to receptors. Remediation options that would break the transport of contaminants to ecological receptors can then be identified from the causal pathways. The quantitative aspects of Bayesian networks support the propagation of uncertainties in the exposure relationships, the effects of exposure, and the effectiveness of risk management interventions. It is necessary to adequately estimate the uncertainties for improved causal inferences and judgments about ecological risks and the effectiveness of remediation. Even if the conceptual network is not quantified, the structures support mechanistic and statistical designs for risk characterization. These and other largely unexplored benefits of conceptual Bayesian networks for assessing and managing contaminated sites will be discussed in this presentation.
EPA Disclaimer: The findings and conclusions in this abstract have not been formally disseminated by the U.S. EPA and should not be construed to represent any agency determination or policy.
Dr. Lionel Jouffe, CEO, Bayesia S.A.S.
It has become customary for BayesiaLab to first present major new releases at the annual BayesiaLab Conference.
At the 8th Annual BayesiaLab Conference at the Laval Virtual World, Dr. Lionel Jouffe, CEO of Bayesia, provided a first glimpse of the new and improved features of BayesiaLab 10.
Presented at the 8th Annual BayesiaLab Conference on October 26, 2020.
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.
Martin Paul Block, Ph.D., Northwestern University
The Master of Science program in Integrated Marketing Communications at Northwestern University consists of 15 courses taken over 15 months. One of the courses is Marketing Mix Modeling, which focuses on relating marketing spending variables to a criterion variable, usually sales, for the purpose of effectively planning future marketing activity spending. Preparing appropriate data is a major component of the course, as is considering different analytical approaches. The course has traditionally focused on regression-based techniques, but now it also adds Bayesian Belief Networks through the application of BayesiaLab. Students find the software easy to apply and solve some of the problems found in the regression-based approach. Students use the software to solve cases and in a term project. The use of BayesiaLab opens the door to discussions about causality and interactive managerial applications.
October 26–30, 2020 — A Virtual Reality Event
Our first all-virtual BayesiaLab Conference at the Laval Virtual World was a great success! With nearly 200 registrations, we had more participants than ever. Despite time differences, researchers from British Columbia to Japan joined us in real time for the event. A big "thank you" to all who made this conference a success!
If you missed some of the talks, we uploaded recordings of all presentations and the corresponding slides.
Presented at the 8th Annual BayesiaLab Conference on October 27, 2020.
The use of marijuana, tobacco, and opioid analgesics has long been associated with psychological distress and chronic pain. The aim of the study was to identify the effects of internalizing and externalizing psychological symptoms utilizing Bayesian Network models to gain a greater understanding of the interrelationships that may exist between pain, psychological and sociodemographic variables. This study included adults enrolled in Wave 4 (2016-2017) of the Population Assessment of Tobacco and Health (PATH) Study. The PATH Study is an ongoing, longitudinal cohort study of tobacco use and related health outcomes in the U.S. Participants with data on variables related to cigarette smoking, marijuana use, psychological distress from the Global Appraisal of Individual Needs Short Screener (GAIN-SS), opioid use, and pain levels (n=32,551) comprised the study sample. The GAIN-SS identifies individuals at risk for mental health or substance use disorders and has been used across the 4-point Likert scale asking about past year problems with internalized disorder, externalized disorder, substance use, and crime/violent behaviors (crime/violent behavior items were excluded from the PATH Wave 4 study). Established smoking status was defined as lifetime use of 100 cigarettes whereas non-established use was defined as not reaching the 100 cigarette lifetime use threshold. The analyses included population sample weights which accounted for missing data. We employed an augmented naïve Bayes (ABN; Bayesialab 9) supervised learning algorithm to identify the interrelationships between the use of pain medications, smoking status, pain intensity, alcohol and marijuana use, GAIN-SS factors, and select sociodemographic characteristics. The initial Augmented Naïve Bayes was cross-validated via the K-folds procedure, with K=10. The final ABN structure was learned and optimized via the minimum description length (MDL) scoring algorithm. The initial MDL score was 732,140.862, representing Entropy (H) = 23.0785 (Standard Deviation: 3.9705), and the final MDL score was 689,686.468, representing Entropy (H) = 21.7285 (Standard Deviation: 3.9812), with mean information compression of 5.7427%. Overall relationship analyses with pain indicated that the internalized behavior factor was the most important predictive variable among these participants suggesting that by knowing pain, we reduce our uncertainty regarding internalized behavior by 2.49% on average. Smoking status, opioid use, and marijuana use were also found to be associated with pain. The severity of internalized, externalized problems and substance use disorders along with current and former established cigarette smoking were associated with high levels of pain intensity, suggesting that self-reported pain is an important factor to consider in smoking cessation and substance abuse counseling programs.
Mahathi Vojjala, MPH Doctoral Candidate, Department of Epidemiology, School of Global Public Health, New York University
Mahathi Vojjala is a third-year doctoral candidate in the Epidemiology track working with Dr. Raymond Niaura. She is a 2017 MPH graduate from New York University School of Global Public Health (GPH) with a concentration in Epidemiology. Prior to her MPH, Mahathi received a B.A. in religion and public health from Rutgers University. Mahathi’s previous research focused on youth smoking initiation and media advertising, dual and poly use of substances specifically marijuana and cigarettes among young adults, media portrayal of alcohol and tobacco in movie trailers and youth smoking rates, and more recently, use of oral analgesics combined with marijuana, alcohol, and cigarettes among people with chronic pain. Mahathi is primarily interested in assessing the benefits and risks of e-cigarettes by examining metabolic biomarker profiles of tobacco user groups using the Population Assessment of Tobacco and Health (PATH) Study.
Marcel de Dios, Ph.D. Assistant Professor, Department of Psychological Health & Learning Sciences, College of Education, University of Houston
Dr. Marcel de Dios is a faculty member in the Department of Psychological, Health and Learning Sciences (PHLS) at the University of Houston. He received his Ph.D. in Counseling Psychology from the University of Miami in 2007. He completed his clinical psychology predoctoral internship at Denver Health and Medical Center and moved on to a two-year postdoctoral research fellowship in behavioral medicine at Brown University Medical School. During his post-doctoral fellowship, Marcel conducted research related to smoking cessation with HIV + Latino smokers. Upon the completion of his fellowship in 2009, Marcel became a faculty member in the Department of Psychiatry & Human Behavior at Brown University Medical School. As a faculty member, his work expanded to include other sub-populations of substance users including young adult marijuana users, Latino light smokers, methadone maintenance smokers, and emerging adults struggling with alcohol and marijuana use. In October of 2012, Marcel relocated to Houston Texas, and became a faculty member in the Department of Health Disparities Research at MD Anderson Cancer Center where he continued his work in the area of smoking cessation funded through an NIH K01 award. In 2017, Marcel joined the faculty of the Counseling Psychology Ph.D. Program at the University of Houston. He has continued his work in the areas of substance use including projects related to young adult marijuana and alcohol users, smokers, and opioid abusers.
Helen Sanchez Doctoral Student, Counseling Psychology, College of Education, University of Houston
Helen Sanchez is a Counseling Psychology doctoral student at the University of Houston, working under the direction of Dr. Marcel de Dios. Prior to her doctoral studies, Helen completed research assistantships in the Health Behavior Research Group at Texas A&M University and the Prinsloo Neuromodulation Lab at MD Anderson Cancer Center. Currently, Helen is a graduate research assistant for the Psychology of Addiction Collaborative at the University of Houston, and her research interests broadly include substance use, health disparities among racial/ethnic minority populations, and health risk perception. Her current work focuses on the use of tobacco and other substances by South Asian Americans. Clinically, Helen works as a psychology intern with the Houston Fire Department, providing psychotherapy to first responders and their family members.
Raymond Niaura, Ph.D. Interim Chair of the Department of Epidemiology and Professor of Social and Behavioral Sciences, School of Global Public Health, New York University
Dr. Raymond Niaura is a psychologist and an expert on tobacco dependence and treatment, as well as substance use. Dr. Niaura has a long history of extramural funding for research projects that have examined the biobehavioral mechanisms of tobacco dependence, including factors that influence adolescent and early adult tobacco and e-cigarette use trajectories. He has also conducted a number of clinical trials that have focused on pharmacological and behavioral interventions for tobacco cessation with an emphasis on disadvantaged and vulnerable subpopulations. Dr. Niaura’s work has been highly influential, and he has published over 400 peer-reviewed articles, commentaries, and book chapters, including the book The Tobacco Dependence Treatment Handbook: A Guide to Best Practices.
His work over several decades has been highly influential and cited and has significantly shaped public policy related to tobacco use and cessation
Steven Frazier, Georgia Pacific
Presented at the 8th Annual BayesiaLab Conference on October 27, 2020.
Contradictory conclusions have been made about whether unarmed blacks are more likely to be shot by police than unarmed whites using the same data. The problem is that by relying only on data from ‘police encounters,’ there is the possibility that genuine bias can be hidden. We provide a causal Bayesian network model to explain this bias – which is called collider bias or Berkson’s paradox – and show how different conclusions arise from the same model and data. We also show that causal Bayesian networks provide the ideal formalism for considering alternative hypotheses and explanations of bias.
Steven Frazier is an experienced engineer with over 40 years of experience in manufacturing management and operations. Graduating with a BSME from Washington University St. Louis in 1979, he has managed and worked in production operations, maintenance, environmental, and safety for Procter and Gamble, ITW, Coca-Cola, Georgia Pacific, and now with OnPoint, a part of Koch Engineered Solutions. As such, he has been involved in the manufacture of a variety of products, from Cascade dishwashing detergent, flexible plastic, and stretch films, metalized films, paper, and building products such as lumber, plywood, oriented strand board, and gypsum board. In 2000, he received 3 patents related to liquid packaging and bag-in-box technology while working for Coca-Cola. He became interested in Bayesian Networks in 2014 because of their ability to predict, discover, and define cause and effect displayed in simplified graphical models. With his extensive domain knowledge, he is currently working as an advanced analytics process engineer for OnPoint, applying machine learning and Bayesian networks to help customers optimize and solve problems related to their manufacturing operations.
Steven Frazier can be described as a “geek’s geek” – throughout his career, he has been the go-to guy in learning increasingly complex applications to enhance the tool kit for production managers, for operations excellence, problem-solving analytics, and innovations to increase safety, optimize production and bottom-line profitability. He has 3 United States patents; he is a peer-reviewed co-author, and his breadth of experience spans iconic brands of Anheuser Busch, Procter and Gamble, Coca-Cola, and Georgia Pacific. With a Washington University Bachelor of Science in Mechanical Engineering, he has managed and worked in production operations, maintenance, environmental, mergers and acquisitions. As a comparative advantage, Steve has leveraged Bayesian Networks -- to increase manufacturing quality performance while maximizing profit, to examine beneficial reuse analysis, and to assess competitive product quality as an element of capturing increased market share.
In his role with Georgia Pacific, he built a first-of-its-kind materials model that surpassed all other comers – both “foreign & domestic.” Today he is part of the Koch Industries OnPoint working as an advanced analytics process engineer applying machine learning to optimize customers manufacturing operations. He became a “thought leader” for Bayesialab when he gave his wife a crash course in big data Bayesian Networks to help her complete her Masters in Sustainability at Georgia Tech. Now that is real-world brand reach – Bayesian Networks applied.
Nicolas Clerc, Caterpillar, Inc.
Presented at the 8th Annual BayesiaLab Conference on October 29, 2020.
Nicolas Clerc, Caterpillar, Inc.
Governmental “lock-downs” in response to COVID-19 have resulted in huge losses to businesses of all sizes across the United States. As businesses turn to insurers to make them whole, insurers are pushing back, arguing that COVID-19 has not caused the “direct physical loss or damage” required by their insurance contracts. The result is litigation, with 1,249 federal court cases reported by PennLaw’s Covid Coverage Litigation Tracker through September 21. With billions of dollars and thousands of bankruptcies in the balance, this presentation uses Bayesian networks to model the probability and magnitude of settlements or verdicts in favor of business plaintiffs in such cases.
Kurt Schulzke, JD, CPA, CFE, teaches forensic accounting and audit analytics at the University of North Georgia. He has published on revenue recognition, materiality, expert witnessing, economic damages, and business valuation through a Bayesian networks lens in a variety of outlets, including the Columbia Journal of Transnational Law, Vanderbilt Journal of Transnational Law, Journal of Forensic Accounting Research, Tennessee Journal of Business Law, and The Value Examiner. With an M.S. in Applied Statistics from Kennesaw State University, he is equally adept as counsel, expert witness, or neutral in valuation-related matters.
Presented at the on October 30, 2020.
Kurt S. Schulzke, JD, CPA, CFE Associate Professor of Accounting & Law University of North Georgia Email:
The purpose of this study was to develop data-driven decision support models relating to higher education. This was done by applying Bayesian networks as an artificial intelligence (AI) method to student throughput data in order to discover relationships between modules in academic programmes. In this study, we developed a Bayesian network which describes the critical pathways to success in academic programmes. We furthermore show that it can be used to optimise existing curricula in academic programmes and understand the impact of interventions such as summer schools on student success. It also identifies weaknesses such as bottlenecks within the curriculum and deficiencies in prior exposure or schooling of students in order to improve student success.
We applied Bayesian networks on two academic programmes: Engineering and Veterinary Science. These two programmes are vastly different in structure as Engineering provides more curriculum options to students and for Veterinary Science, students need to adhere to a strict set of modules for accreditation purposes.
The overall impact of this study is on academic programme decision support such as curriculum optimization and high impact intervention strategies.
Alta de Waal, Ph.D. Centre for Artificial Intelligence Research Department of Statistics, Faculty of Natural and Agricultural Sciences University of Pretoria, South Africa
Alta currently holds a senior lecturer position in the Department of Statistics, University of Pretoria, South Africa. She has 20 years of experience in design, development and implementation of different components in the AI value chain. She develops Bayesian network models in application areas such as student throughput models, wildlife security, environmental risk management and transportation. Alta also studies natural language processing (NLP) with a special interest in probabilistic distributional semantic methods.
Spatially Discrete Probability Maps for Anti-Poaching Efforts (Paris, 2017)
Presented at the on October 29, 2020.
Presented at the 8th Annual BayesiaLab Conference on October 30, 2020.
While public opinion is a strong driver of policy change in democratic societies, the complex interactions of climate risk perceptions, knowledge and belief in climate science and their combined effects on support for policies aimed at mitigating climate change are not very well understood. This study presents a novel machine learning approach to learn a “probabilistic structural equation model (PSEM)” for understanding complex interactions among climate risk perceptions, beliefs, knowledge, political ideology, socio-demographics, geographic representation and their combined effects on support for mitigation policies. With foundations in Bayesian Network theory and information theory, PSEMs use the principle of Kulback-Leibler divergence to rank order the relative importance of factors that explain structural drivers measured as latent variables and dynamics of support for climate policies among different segments of populations. The PSEM is derived from publicly available mixed-pool “Climate Change in the American Mind” (CCAM) dataset collected between 2008 and 2018 (N=22,416). The estimated PSEM predicts that 27.38% of the US population strongly supported climate policy action, while 59.46% are lukewarm supporters and 13.15% strongly oppose climate policy interventions. Predicted posterior probabilities of opposers, lukewarm supporters and strong supporters of climate policy action conditional upon beliefs, concern, global warming risk perceptions, ideology and other predictors in the network can be estimated from the PSEM. Consistent with theoretical expectations, we find that the strong supporters of climate policy are more likely to be strong believers, highly concerned, alarmed and tend to be very liberal or somewhat liberal. In contrast, strong opposers of climate policy are more likely to be climate deniers, skeptics or doubtful, not concerned, risk deniers and very conservative or somewhat conservative. The conditional probability distributions of lukewarm policy supporters (the largest group among the US population at 59.46%) display probably the most novel and revealing findings of this PSEM: Lukewarm supporters are more likely to be ambivalent and moderate believers and less likely to be strong believers. Further, their likelihood of being not concerned about climate change is slightly higher compared with the population mean. The lukewarm policy supporters also contain fewer people who perceive high risk from climate change. Finally, from ideology standpoint, lukewarm supporters represent relatively larger segment of moderate/independents. Poor adoption of climate policy proposals in the US can be attributed to this silent majority of lukewarm supporters who perceive little to moderate risk from climate change and remain ambivalent about human-induced climate change.
Asim Zia, Ph.D. Professor of Public Policy & Computer Science: Department of Community Development and Applied Economics & Department of Computer Science at the University of Vermont Director: Institute for Environmental Diplomacy and Security Co-Director: Social Ecological Gaming and Simulation Lab Fellow: Gund Institute for Environment 146 University Place, Morrill 208E, Burlington VT 05405 USA Phone: +1 802-656-4695 Asim.Zia@uvm.edu https://www.uvm.edu/cals/cdae/profiles/asim_zia http://www.uvm.edu/~azia/
Asim Zia has made substantive scientific and policy contributions towards advancing the Sustainability and Resilience of Human Environmental Systems. He is an internationally known leader in developing computational models of Social Ecological Systems, Complex Adaptive Systems and Governance Networks. He has published 58 journal articles, 19 book chapters and 3 books, totaling 80 peer-reviewed publications. He has served as a Principal Investigator, Co-Principal Investigator, or Senior Personnel on 22 research grants worth more than $60 Million sponsored by NSF, USDA, US DoD, US DoT and MacArthur Foundation. He has a Ph.D. in Public Policy from the Georgia Institute of Technology.
Machine learning how human risk perceptions shape behavior (Nashville, 2016)