ßÙßÇÂþ»­

Industrial Engineering

Devashish Das, Assistant Professor
Research Interests: Health Care Systems Engineering, Applied Probability and Statistics
Email | Phone: 813-974-8294 |

Dr. Devashish Das's group develops stochastic models and statistical learning algorithms for solving medical-decision-making problems. Emerging technologies such as wearable devices and internet-based technologies produce large amounts of complex data that require new mathematical models and data-analytic methods to develop new therapies and healthcare policies to improve health outcomes. In collaboration with the Mayo Clinic, Dr. Das's group focuses on solving such problems. Some on the ongoing projects in the group are:

Modeling the impact of lack of testing on policy making and the spread of COVID-19

Unlike well-known viral infections, when new viral diseases like COVID-19 emerge, it requires developing new methods of testing for the virus. Naturally, there is a delay in making the necessary number of tests available to measure the true prevalence of the disease. We are developing mathematical models that help us provide probable estimates of the real spread of the disease when only partial information about the spread of the disease, resulting from a lack of tests, is available. Also, we try to answer how much testing capability is needed to estimate the true prevalence of the disease with adequate accuracy.

Developing blood-transfusion control using wearable devices to treat trauma patients

Preventing cardiovascular collapse in trauma patients who are losing blood is critical to saving lives. New sensors can provide real-time information about the condition of trauma patients. However, using the signal from such sensors to develop therapies and make blood-transfusion decisions remains an unsolved problem. Our research group is working on algorithms to understand the physiology of trauma from the sensor data and use them to improve outcomes for patients.

Monitoring the flow of patients in hospital emergency departments

Hospital emergency departments are the front door to the healthcare system for a significant number of patients and overcrowding the emergency department is a problem in most emergency medical centers. Our group is using large amounts of sensors data for Mayo Clinic to develop computational methods to improve the flow of patients in the ED and improve access to emergency medical services.

 

Hadi Charkhgard, Assistant Professor
Research Interests: Multi-Objective Optimization, Operations Research, Integer Programming
Email | Phone: 813-974-2090 |

Dr. Hadi Charkhgard is the director of Multi-Objective Optimization Laboratory (MOOLab) in the department of Industrial and Management Systems Engineering at the ßÙßÇÂþ»­ (USF). The MOOLab was established by Dr. Charkhgard in August 2016 when he joined the USF and graduated its first PhD student in December 2019. The MOOLab highly supports diversity and groups underrepresented in academia. This is highlighted by the fact that the first graduate of the MOOLab was from Colombia[1] and he is now working as an Operations Research Scientist at Amazon Robotics. Moreover, the MOOLab currently has four PhD students from three different countries including China, Iran, and Pakistan.  Dr. Charkhgard has been highly active in prompting diversity. He has a track-record of research collaborations with three Hispanic female/male PhD students outside of his group. In a recent effort, he has started to learn Spanish to be able to better communicate with Hispanic students and encourage them to join the MOOLab. He is also currently co-advising a female African-American master’s student. Overall, the hope is that by continuing such efforts and the availability of exciting research opportunities, the MOOLab can continue to promote diversity. PhD students joining the MOOLab can  work on a variety of  research topics including but not limited to improving decision making process in natural resource management, improving transportation systems, query optimization in database systems, exploring the nexus of machine learning and optimization, and developing the next-generation of general-purpose mathematical optimization tools/algorithms. All research topics rely heavily on computing and mathematical optimization. These skills can be viewed as two wings that help students become more competitive in the job market after graduation and the MOOLab is a unique place to learn them.  By joining the MOOLab, PhD students can take advantage of the long-term collaboration of the laboratory with the United States Geological Survey for solving complex decision-making problems arising in natural resource management such as controlling invasive species and reserve design problems.  The laboratory also has a strong relationship with the department of Civil and Environmental Engineering as well as the department of Computer Science at the USF.

 [1]  The exit interview of the first PhD graduate of the MOOLab.  

 

Changhyun Kwon, Associate Professor
Research Interests: Transportation Systems Analysis, Service Operations, Risk Management
Email | Phone: 813-974-5588|

Research opportunities

Neural Combinatorial Optimization Methods for Vehicle Routing Problems 
This project intends to develop mathematical and computational optimization methods for solving various vehicle routing problem, rising in the fields of urban food delivery, same-day parcel delivery, ridesharing services, and other shared mobility services. Vehicle routing problems are to determine the route and schedule of service vehicles to visit customer locations, to deliver or pickup items or people. Service vehicles can be trucks, shuttle buses, passenger cars, aerial drones, or any other mediums that can provide mobility and logistics services. The computational methods will be based on large-scale heuristic methods such as large neighborhood search methods, combined with neural networks approaches.

 

Ankit Shah, Assistant Professor
Research Interests: Cybersecurity Analytics, Deep Reinforcement Learning, Adversarial Machine Learning, Combinatorial Optimization, Stochastic Dynamic Programming
Email | Phone: 813-974-5584 |

Ankit Shah is an Assistant Professor of Industrial and Management Systems Engineering at the ßÙßÇÂþ»­. He also holds a courtesy faculty appointment of Assistant Professor with the Computer Science and Engineering department. His research interests lie at the intersection of Computer Science, Operations Research, Information Technology, and Data Analytics with a focus on cybersecurity issues of societal concern. His research is in the field of Artificial Intelligence (AI) for cybersecurity, where he has developed reinforcement learning-based methodologies, in collaboration with the Army Research Laboratory, Adelphi, MD, to address many important cybersecurity issues. His current research projects include optimizing cyber vulnerability triage and mitigation, developing hybrid human-AI intrusion detection alert management system, and developing intelligence from wide-area motion imagery data.

 Dr. Shah’s research focuses on the development of AI-based decision support systems that assist decision-makers in the process of analyzing collected data, identifying critical information such as cyber vulnerabilities and threats, and prescribing optimal mitigation actions. He uses advanced machine learning techniques such as deep learning, transfer learning and adversarial machine learning to create robust models for anomaly detection, and uses approximate dynamic programming and deep reinforcement learning methodologies for dynamic decision-making under uncertainty.

 

 

Mingyang Li, Assistant Professor
Research Interests: Bayesian Data Analytics, Data Mining, System Informatics
Email|Phone: 813-974-5579 |

Data-enabled Aged Care Analytics

Due to the rapid aging of the baby boomer generation, the U.S. will soon experience a significant growth in elderly people who may suffer from multiple chronic diseases, disabilities and impairments. To meet with the excess demand and improve quality of care for the older adults, this line of research is to develop and apply advanced data-enable analytics techniques, such as statistical models, machine learning and computer simulation, integrated with domain expertise in geriatrics, long-term care services research and nursing care industrial practice, to address the complexity of diverse and rich healthcare data in the aged care systems, such as administrative claims data, complex survey and assessment data as well as in-situ monitoring data. The proposed analytics models and techniques will improve prediction of adverse events (e.g., fall) occurrence of older adults, allow better understanding of older adults’ service utilization of different aged care settings, ranging from hospital, assisted living facilities and nursing homes, and facilitate cost-effective and proactive resource (e.g., capacity, workforce) planning decisions in aged care systems.

Bayesian Modeling and Computation of Heterogeneous Data

Conventional statistical modeling often
assumes a homogenous population of units, such as patients, while in many real applications, the overall population can be quite heterogeneous. This line of research is to develop novel Bayesian data science and informatics methodologies, including Bayesian parametric/non-parametric models and effective and efficient Bayesian sampling algorithms (e.g., Markov Chain Monte Carlo methods), to analyze complex heterogeneous lifetime data and heterogeneous longitudinal data. The proposed non-parametric formulation relaxes the conventional modeling assumption of pre-specifying a fixed number of sub-pupations. The proposed Bayesian sampling algorithms also allow the joint model parameters estimation and sub-population number identification in a one-step fashion.

Multi-level/Multi-fidelity Data Fusion

Modern complex system often has complex data structures, e.g., multi-level (e.g., component level and system level), mixed-type (e.g., time-to-event data, trajectory data and counts data) and multi-fidelity data. This line of research is to develop generic, coherent, flexible and scalable data fusion framework to integrate multi-source, multi-level, mixed-type and/or multi-fidelity data to improve prediction accuracy and system performance. Some successful achievements include the improvement of lifetime data modeling accuracy and precision, the sample size reduction of experimental design and the improvement of in-situ data monitoring performance (e.g., crowd tracking and surveillance).

Predictive and Prescriptive Analytics for Performance Improvement of Complex Systems

With the advancement of sensing technology and information storing systems, a data-rich environment has been created in many complex systems. This line of research is to develop data-driven models using data mining, computational intelligence and multi-objective optimization methods to tackle complex problems (e.g., modeling, prediction, design, monitoring, diagnostics, prognostics, planning, scheduling, control, etc.) in a data-rich environment and improve prediction as well as real-time decision-making through advanced
modeling and analysis of high volume, high-dimensional and high throughput operational performance data.