Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Loading...
Ahsana Parammal Vatteri, Chair on Disaster Risk Reduction and Resilient Engineering
Presented at the 2023 BayesiaLab Spring Conference.
Natural hazards such as floods and earthquakes pose a significant risk to the lives and livelihood of people living in exposed regions and affect the continuation of education of students, which is a fundamental right. Yearly flood events with low inundation depths may not cause structural damage, however, resulting material degradation contributes to higher vulnerability to subsequent seismic events. This combined effect of flood and seismic hazards, along with other functional losses, ultimately results in disruption to education delivery. The socio-economic condition of the users-community also plays a role in the extent of such disruption. This problem demands consideration of a large number of dimensions to estimate the impact on the school system infrastructure in a locality. A Bayesian network (BN) model is proposed to handle the qualitative and quantitative nature of these variables, representing multiple schools in a locality as a system. Three factors are considered to contribute to the system disruption, namely, schools’ physical functionality loss from damage to infrastructure, accessibility and change of use, and social vulnerability. The impact is quantified through the probability of the system being in various states of disruption, which can support decision-making and strategic planning in the face of multiple hazards.
Ahsana Parammal Vatteri is a research fellow on the UNESCO Chair on Disaster Risk Reduction and Resilient Engineering, chaired by Prof. Dina D’Ayala at UCL. She works in the EPICentre research group in the field of earthquake and multi-hazard engineering and the application of Bayesian Networks for system resilience problems. She focuses on the confined masonry typology of buildings, especially school buildings, and the resilience of education infrastructure. Her other engagements include consultancy projects on seismic-safety assessment of schools for DFID UK, the 2020 EEFIT Aegean Earthquake and Tsunami reconnaissance mission, modified taxonomy for confined masonry school buildings for GPSS project for the World Bank, SECED Young Members Subcommittee, etc.
Dr. Lionel Jouffe, Bayesia S.A.S.
Presented at the 2023 BayesiaLab Spring Conference.
Knowledge Elicitation has been a central research topic for the BayesiaLab team for many years, so the arrival of ChatGPT last year has prompted us to leverage its innovative technology immediately with BayesiaLab.
Learn More About Knowledge Elicitation
In this presentation, we will show new functions that directly integrate ChatGPT into BayesiaLab, including:
Chat Completion
Image Generation
Embedding Generation
BayesiaLab's forthcoming Subject Matter Assistant, to be released in Version 11, can improve research workflows in several ways:
Accelerate the qualitative part of knowledge elicitation.
Generate practical natural language descriptions for latent factors created through BayesiaLab's clustering functions.
Automatically create images to illustrate nodes in a network.
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.
With participants from over 30 countries, the 2023 BayesiaLab Spring Conference was a truly international event. If you missed some of the talks, we have now posted recordings of all presentations to our .
Hamza Zerrouki, Department of Process Engineering, University Amar Telidji of Laghouat
In the last few decades, chemical and process industries have become more prone to accidents due to their complexity and hazardous installations. These accidents lead to significant economic and, most importantly, human losses. Risk management is one of the appropriate tools to guarantee the safe operations of these plants. Risk analysis is an important part of risk management, and it consists of different methods such as Fault tree, Bow-tie, and Bayesian network (BN). The latter has been widely applied for risk analysis purposes due to its flexible and dynamic structure. In the current presentation, we will expose the different applications of BN in chemical and process industries with examples to explain how BN can be used to conduct a risk assessment, safety, and risk analysis of these industries.
Not available.
Hamza Zerrouki, Department of Process Engineering, University Amar Telidji of Laghouat, Laghouat, Algeria
I joined the Department of process engineering in September 2018. Before that, I was with the Institute of Occupational Health and Safety at Batna 2 University as a Ph.D. student from 2013 until 2018.
I have a bachelor's degree in industrial and environmental safety from the University of Laghouat (2011) and a master’s degree in control of industrial risks from the University of Batna 2 (2013). My research is mainly focused on the safety assessment and management of petrochemical facilities regarding technological accidents and Bayesian network applications.
Serdar Semih Coşkun, İstanbul University Faculty of Economics
Presented at the 2023 BayesiaLab Spring Conference.
Football (soccer) stocks are substantially subject to investor sentiment stemming from football fields. Evaluating sentiment functions helps us understand how investors interpret field signals and attach value to those signals in stock markets. This study develops the Gaussian investor sentiment process exploration programming (GISPEP) framework for exploring investor sentiment as a function of probabilistic field signals. The GISPEP provides an alternative event-study approach based on prospect theory and Bayesian analysis. We use the GISPEP to set the causality between match results and stock returns of the Fenerbahçe (FB), Galatasaray (GS), and Beşiktaş (BJK) football clubs in Turkey. A natural experiment also enables us to test the effect of competitive emotion that varies across two seasons. Our results indicate that competitive emotions regulate the asymmetric rise of availability and loss aversion heuristics under ambiguous field signals. In addition, loss signals increase the heterogeneity of market expectations.
Serdar Semih Coşkun is an assistant professor at İstanbul University Faculty of Economics, Turkey. Serdar holds a Ph.D. in Business Administration from the same university. Serdar’s research interests include Bayesian statistics, behavioral economics, and supply chain management.
Xingang Zhao, Ph.D., Oak Ridge National Laboratory
Presented at the 2023 BayesiaLab Spring Conference.
Future advances in nuclear power technologies call for enhanced operator advice and autonomous control capabilities. One of the first tasks in developing such capabilities is the formulation of symptom-based conditional failure probabilities for power plant structures, systems, and components (SSCs). The primary goal is to aid plant personnel in (1) deducing the probabilistic performance status of the monitored SSCs and (2) detecting impending faults/failures. The task of estimating conditional failure probability is a bidirectional inference problem, and a logical approach is to use the Bayesian network (BN) method. As a knowledge-based artificial intelligence tool and a probabilistic graphical model, BN offers reasoning capability under uncertainty, graphical representation emulating the physical behavior of the target SSC, and explainability of the model structure and results. This presentation will provide an overview of the BN technique and the software tools for implementing BN models in this task, along with the associated knowledge representation and reasoning paradigm. The challenges with data availability and the general approach to target SSC identification will be highlighted. Two example case studies on the failure of (1) a centrifugal pump and (2) an electric motor will be presented to demonstrate the usefulness and technical feasibility of the proposed BN-based fault diagnostic artificial reasoning system using expert system shells.
Dr. Xingang Zhao is an R&D Scientist at Oak Ridge National Laboratory. He received his Ph.D. in nuclear science and engineering from the Massachusetts Institute of Technology. His research interests span multiple disciplines of clean energy systems and their intersections with artificial intelligence and decision science. He has been a major contributor to a diverse portfolio of research projects that advance the state of the art of modeling & simulation and digital engineering for nuclear and renewable energy applications.
Dr. Lal Hussain, Department of Computer Science & IT, University of Azad Jammu and Kashmir, Muzaffarabad, Pakistan
Presented at the 2023 BayesiaLab Spring Conference.
Accurate classification of brain tumor subtypes is important for prognosis and treatment. Researchers are developing tools based on static and dynamic feature extraction and applying machine learning and deep learning. However, static feature requires further analysis to compute the relevance, strength, and types of association. Recently Bayesian inference approach gains attraction for deeper analysis of static features to analyze a comprehensive analysis among extracted nodes (features). We computed the gray level co-occurrence (GLCM) features from brain tumor meningioma and pituitary MRIs and then ranked based on entropy methods. This study is focused to utilize the important highly ranked energy feature as the target node for empirical analysis of dynamic profiling and optimization to unfold the nonlinear dynamics of GLCM features extracted from Brain MRIs to distinguish the pituitary and meningioma glands.The highest strength of relationship was obtained between the nodes (Correlation→correlation2) yielding strength of relationship using KL and MI (1.4640), Pearson’s correlation (1.0000), with relative width 1.0000 and overall contribution of 16.67%. The segment profile analysis of top ranked target node with other extracted GLCM features using the Radar chart was computed which reflect the distributions based on 1 to 12 clock hours. We used the NHST t-test and Bayesian test to find the significance to distinguish with other different states such as <=0.273, c) <=0.368, d) <=0.471, e)>0.471. The clusters <=0.273, <=0.471 and >0.471 using both the test yielded the highly significant results with all the extracted GLCM features. The state <=0.368 yielded high significant results using both test with homogenity1, dissimilarity, correlation, correlation2, and autocorrelation, whereas significant results using NHST t-test with contrast and energy. Using the tornado graph, we visualize the maximum deltas in the posterior probabilities of the target states, and hard evidence is set on the selected variables. The highest association was yielded with entropy, homogeneity, dissimilarity, contrast, correlation, correlation2 cluster state <=0.273 followed by cluster state <=0.368, <=0.471, and >0.471. This indicates that high top ranked Energy feature prevails high associations with entropy, homogeneity, dissimilarity, contrast, correlation, and correlation which can be used as a better predictor for improved diagnosis and prognosis of brain tumor types. Previous studies rely on classification methods. However, this novel technique is proposed to further investigate the dynamics, associations, posterior probabilities, prior probabilities, marginal likelihood, prior means, and posterior means to further unfold the relevance and relationships among the extracted features. The proposed approach will be very helpful for improved diagnosis and prognosis of brain tumor types. The proposed method further unfolds the dynamics and detailed analysis of computed features based on GLCM features for a better understanding of the hidden dynamics for proper diagnosis and prognosis of tumor types leading to brain stroke.
Dr. Lal Hussain is an Assistant Professor at the Department of Computer Science & IT, University of Azad Jammu and Kashmir, Muzaffarabad, Pakistan. He obtained his MS in Communication and Networks from Iqra University, Islamabad, Pakistan, in 2012 with a gold medal. He received Ph.D. from the Department of Computer Science & Information Technology, University of Azad Jammu and Kashmir, Muzaffarabad, Pakistan in February 2016. He worked as visiting Ph.D. researcher at Lancaster University UK for six months under HEC International Research Initiative Program and worked under the supervision of Dr. Aneta Stefanovska, Professor of Biomedical Physics, Physics Department, Lancaster University, during 2014-2015 in the UK. Dr. Hussain recently completed a one-year Postdoctoral fellowship from Montefiore Medical Center and Albert Einstein College of Medicine, New York, USA, under the supervision of Dr. Tim Q Duong, Professor and Vice Chair of MRI research. He also worked at Duong Lab at Stony Brook University USA on different ongoing projects with Dr. Duong from January-March 2020. Recently, Elsevier ranked Dr. Hussain in the top-1% of the world scientists list of 2021 based on their research record. He is the author of more than 50 publications of highly reputed peer-reviewed and Impact Fact Journals as Principal author. He completed various funded projects as PI and Co-PI from Ignite, ICT Pakistan, and the University of Jeddah, Saudi Electronic University, Kingdom of Saudi Arabia. He presented various talks in Pakistan, the UK, Peru, and the USA. His research interest includes developing and optimizing AI tools, including machine learning, deep learning, and neural network algorithms, feature extraction and selection methods, information-theoretic methods, time-frequency representation methods, and cross-frequency coupling to predict disease severity, progression, survival, and recurrence. His area of interest includes biomedical signal and image processing problems, including prostate cancer, breast cancer, lung cancer, brain tumor, covid-19 lung infection with different modalities (i.e., MRI, CT, X-Ray, etc.), brain dynamics and diseases (i.e., autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD), Alzheimer's Disease).
Minghui Wang, Ph.D., Icahn School of Medicine at Mount Sinai
Presented at the 2023 BayesiaLab Spring Conference.
Understanding the molecular regulatory networks underlying insulin resistance (IR) is crucial to preventing type 2 diabetes and related metabolic disorders. However, our knowledge of the landscape of IR-related transcriptomic regulation in glucose-responsive tissues and its cell-type specificity regulatory mechanisms remains incomplete. To provide a comprehensive population-level understanding of the organizations and cell-type-specific regulations of gene expressions underlying IR,
we employed an integrative network biology approach to integrate the multi-omics and phenotypic data from well-powered African American (AA) and European ancestry (EA) cohorts. By integrating the state-of-art single-cell sequencing data analyses with bulk-tissue expression quantitative trait loci (eats), and coexpression and Bayesian causal networks, we presented trans-ethnic and cross-tissue results of IR in adipose and muscle tissues. We identified ethnically conserved cell-type signatures and gene modules associated with insulin sensitivity responses. We further prioritized modules enriched for cis-eQTL genes and predicted network driver genes for experimental validations. Together, this study revealed the cell-type-specific transcriptomic networks and signaling maps underlying insulin resistance in major glucose-responsive tissues.
Dr. Minghui Wang is an Associate Professor in the Genetics and Genomic Sciences at the Icahn School of Medicine at Mount Sinai. Dr. Wang was trained in statistical and quantitative genetics during his graduate studies at Fudan University and postdoctoral research at the University of Birmingham. At Icahn School of Medicine at Mount Sinai, his research focused on developing novel integrative network models that combine multi-layers of bulk-tissue and/or single-cell functional genomics data to uncover the cellular changes and hidden regulatory relationships among a large network of genes/proteins in disease-relevant tissues for human aging and human complex disorders like Alzheimer’s disease (AD), cancers, and diabetes, etc.
Andrew Holle & Becky Gibbs-Murray, American Innovations
Presented at the 2023 BayesiaLab Spring Conference.
In this presentation, we explore how BayesiaLab can work to identify trends in the DOT’s Pipeline and Hazardous Materials Safety Administration’s public incident and annual report data to identify strategies that pipeline operators can leverage to reduce incidents and predict failures more effectively.
Andrew Holle has been with American Innovations for over nine years and currently serves as the Senior Principal Systems Architect. Andrew has a BS in Chemical Engineering and an MS in Computer Science. He is currently focusing on innovative ways to bring the benefits of machine learning to customers through our product offerings.
Becky Gibbs-Murray has focused her entire career on building innovative, user-friendly tools that advance the oil and gas industry. She joined American Innovations in October 2021 as Senior Product Manager and holds a degree in Chemical and Petroleum Engineering from the Colorado School of Mines.
Mohammad Reza Valaei & Vahid Khodakarami, Ph.D., Bu-Ali-Sina University, Iran
Assessing risk for a start-up is always a complex and challenging task, as historical data is typically unavailable. Traditional methods are inadequate in capturing the full complexity, so more sophisticated tools are required. This paper presents a method for estimating the default rate at various stages of a start-up's life cycle using an expert-elicited Bayesian Networks (BNs) methodology. A prototype BN model is proposed to combine diverse sources of information, including historical data and expert knowledge. The model has a hierarchical structure to capture start-ups' known risk factors. It also uses the Noisy-OR operator to capture the unknown risks in each of the main categories.
The prototype model can be adjusted to capture the unique characteristics of each start-up and investor. 3 case studies were used to demonstrate the applicability of the model. The proposed method reduces the cognitive error of experts, takes advantage of the learning feature of BNs when updating default estimations, and takes into account the impact of investors' risk appetite. It also allows for ranking the most effective risk factors at various stages of the start-up life cycle.
Mohammad Reza Valaei, Bu-Ali-Sina University, Hamedan, Iran
Mohammad Reza Valaei is a Ph.D. candidate in Industrial Engineering at Bu-Ali-Sina University. His research interests include practical studies in financial risk management, start-up valuation, venture capital, and portfolio optimization.
Vahid Khodakarami, Ph.D., Bu-Ali-Sina University, Hamedan, Iran
Vahid Khodakarami is an associate professor at Industrial Engineering Department at Bu-Ali-Sina University. During his Ph.D. study at Queen Mary University of London in 2004, he was introduced to the Bayesian Networks technology. Since then, applying BN in real-life projects has become his main research interest. He has published several papers. Risk Assessment, Reliability, Sustainability Engineering, and Project Management are among Vahid's other research interests.
Presented at the .
Jenny Betsabé Vázquez-Aguirre, University of Veracruz
Presented at the 2023 BayesiaLab Spring Conference.
The study of causality can be traced back 300 years; its origins are attributed to two thinkers of the epoch (Emmanuel Kant and David Hume). Both philosophical currents are opposed, though both have the same objective, to explain the origins of causal processes in the human brain.
Several disciplines have been studying causality from different prospects; two of these disciplines are Cognitive Psychology (CP) and Artificial Intelligence (AI). Despite these areas having their origins together, currently, they work separately.
There are multiple efforts to replicate and explain causality based on prior knowledge; from CP, there have been different proposals; its aim is to describe how to learn the natural cause-effect relationships, and from AI, there are multiple algorithms focused on estimating or predicting causality.
From IA, a method for learning causality is through the intervention of variables in Bayesian Networks (BN); however, this requires prior knowledge that indicates the cause-effect direction within the connections in the network. For the above, our proposal is a joint implementation to create Causal Bayesian Networks (CBN) from the fusion of two areas, the CP and IA. The principal goal is an integrated two algorithms for the construction of the graphical models (CBN) from the dataset that can learn the causal relationships, giving direction to the arcs within the network, such that indicate who is the cause and who is the effect. For the construction, we used the Rescorla-Wagner model in the first algorithm and the Power-PC theory for the second; both methods belong to the CP.
The results obtained with this approach have been encouraging, and the CBNs acquired can be used to intervene in variables and estimate causal probabilities. For comparison, we used traditional Bayesian Networks proposed by AI.
We have tested with real and repository datasets, comparing our networks with those drawn manually by the experts who have provided us with the data. In the beginning, only 14% of the BNs constructed with traditional AI algorithms could be used for intervention purposes. Most of the CBNs obtained with this proposed algorithm can be used for this purpose and reflect more cause-effect connections than those created with other AI algorithms.
Jenny Betsabé Vázquez-Aguirre was born in Veracruz, Mexico. She has a bachelor's degree in statistics and two master's degrees: one in Quality Management and the other in Artificial Intelligence.
She has worked as a data analyst, software tester, process manager, as well as statistical methodology teacher.
She is currently studying the last semester of a Ph.D. in Artificial Intelligence, in which she is working on causal analysis through Causal Bayesian Networks principally, but also, she is continually working on data analysis with machine learning.
Mahmudur Fatmi, Ph.D., University of British Columbia – Okanagan
Presented at the 2023 BayesiaLab Spring Conference.
Agent-based microsimulation modeling techniques are adopted for urban system modeling mainly because of their capacity to address the complex interactions among individuals, households, and other urban elements. The performance of urban simulation models largely depends on the quality of the input data, which is generated through a population synthesis procedure. This study proposes a Bayesian network and generalized raking techniques for population synthesis. The Bayesian network is used to generate the synthetic population pool from the microsample, and generalized raking is used to fit the synthetic population with the control total. Some of the key features of the proposed population synthesis are as follows: accommodating heterogeneity based on both household and individual attributes, tackling missing/incomplete observations in the microsample, and generating a true synthesis of the population from the microsamples. A data-driven structure learning technique is adopted to generate effective and optimal structures among heterogeneous households and individuals. This Bayesian network + generalized raking procedure is implemented to generate a 100% synthetic population at the smallest zonal level of dissemination area for the Central Okanagan region of British Columbia. The results suggest that capturing heterogeneity within the Bayesian network has tremendously benefitted the reconstruction process in efficiently and accurately generating a synthetic population from the available microsample. Finally, this population synthesis is developed as a component of the agent-based integrated urban model, currently under development at The University of British Columbia’s Okanagan campus.
Dr. Mahmudur Fatmi is an assistant professor in Civil Engineering at UBC-Okanagan. His expertise revolves around travel demand modeling, particularly focusing on the interactions among population socio-demographics and their transportation and land use-related decisions. He contributes to developing econometric models and agent-based microsimulation techniques. He has partnerships with cities and transit agencies, and his research findings assist them in making effective transportation and land use policies and infrastructure investment decisions. He is a member of several transportation research and professional communities and societies. Dr. Fatmi’s contributions have been recognized through several local, provincial, and national awards, such as Transport Canada Scholarship and Nova Scotia Research and Innovation Scholarship.
María Pazo Rodríguez, School of Mining and Energy Engineering, University of Vigo
Presented at the 2023 BayesiaLab Spring Conference.
Heavy economic activities such are industry and agriculture strongly limit soil quality. Potentially Toxic Elements (PTEs) are a threat to public health and the environment. In this respect, it is critical to identify areas that require remediation. In the herein research, a geochemical dataset (230 samples) comprising 14 elements (Cu, Pb, Zn, Ag, Ni, Mn, Fe, As, Cd, V, Cr, Ti, Al, and S) was gathered throughout eight different zones distinguished by their main activity, namely, recreational, agriculture/livestock and heavy industry in the Avilés Estuary (North of Spain). Next, a stratified systematic sampling method was used at short, medium, and long distances from each area to obtain representative visualization of the total variability of the chosen attributes. The information was then combined into four risk classes (low, average, high, and in need of remediation) based on multiple sediment quality guideline (SRM) baseline values. Bayesian analysis, inferred for each area, was used to characterize PET correlations, with the unsupervised learning network technique being the best solution. According to the Bayesian network structure obtained, Pb, As, and Mn were chosen as key contamination parameters. For these three elements, the conditional probability obtained was allocated to each observed point, and a simple, direct index (Bayesian Risk Index-BRI) was constructed as a linear rating of the pre-defined risk classes weighted by the previously obtained probability. Lastly, BRI underwent geostatistical modeling. One hundred Sequential Gaussian Simulations (SGS) were computed. The mean image and standard deviation maps were obtained, resulting in high/low-risk clusters (local G clusters) and spatial uncertainty calculation. High-risk clusters are mostly located in the area with the highest elevation (agriculture/livestock) associated with low spatial uncertainty, which indicates the need for remediation. Air emissions, primarily from the metal industry, contribute to soil contamination by ETPs.
María Pazo Rodríguez is a doctoral student at the School of Mining and Energy Engineering affiliated with the University of Vigo.
My research is focused on developing Bayesian models that can guide the implementation of digital transformation in the mining and energy sector. To this end, the main objective is to equip the mining industry with risk assessment tools and decentralized decision models that enable its rapid decarbonization, thus contributing to meeting the growing demand for strategic minerals and addressing the strict environmental policies imposed by national and international organizations.
Anand Wilson & Buvana Iyer, Course5 Intelligence
Presented at the 2023 BayesiaLab Spring Conference.
Identified factors that drive customer lifetime value and components that drive higher CLV to drive revenue and improve customer retention. Using data from this study, we determine the strength of the relationship between Opportunity, Lifetime Value, and Customer Ratings. Graphs, causal insights, and early warning signals are generated through network analysis. As a result, strategies are better planned, and gaps are identified and addressed more efficiently. Surveys are conducted regularly to understand how to serve our most valuable customers better. By improving customer experience (CX), we are able to increase their lifetime value (LTV). This method also enables us to make precise recommendations about how to improve Client Economics. The Bayesian implementation offers some advantages, which help in more precision and enhanced flexibility in bringing different data sets together with actionable insights.
Anand has 9+ years of experience in applied artificial intelligence and data sciences. He has worked for marque clients such as Lenovo, Intel, Microsoft, Novartis, Novo Nordisk, GE, Mars Wrigley, PepsiCo, etc., enabling digital transformation using A.I.
In his current role, Anand focuses on developing and marketing solutions based on the Bayesian Network model theory, which enables us to quantify causality in an observational study. His major areas of work/research include Knowledge Modelling, Machine Learning with BayesiaLab, and Inference.
Anand comes from an applied statistics background. He has a master's degree in statistics. Anand carries an acute interest in machine reasoning, causal inference, and experimental designs, along with machine learning and data science.
Over 15 years of international and domestic market experience with a proven track record of leading high-profile strategic projects on a fast-moving set of priorities and business initiatives to translate strategic organizational goals into clear operational plans and derive measurable results.
Buvana consults C-level clients and has led projects achieving org-level implementation of enterprise analytics, including Software Development, BI & Analytics, and ML & AI solutions (at scale), adopting DevOps philosophy with agile delivery. Expertise in leveraging both traditional statistics and machine learning techniques to create solutions and deliver business value.
Buvana comes from applied Mathematics background. She has a master's degree in Mathematics. Buvana carries an acute interest in predictive analytics, statistical modeling, machine reasoning, and experimental designs, along with machine learning and data science.