Operations and Information Management
M.Sc. Industrial Engineering and Management, Linkoping University, Linkoping, Sweden
B.Sc. Industrial Engineering, Iran University of Science and Technology, Tehran, Iran
Discrete Optimization, Machine Learning
Logistics and Supply Chain Management
Transportation, Scheduling, Inventory Management
Combinatorial Auctions, Procurement Auctions
My research explores applications of operations research, data analytics, and machine learning models to solve complex business problems in transportation, logistics, and supply-chain management. Through my PhD studies at the University of Connecticut (UConn) and my summer research internship at Fidelity Investments, I have had the opportunity to develop a diverse research portfolio centered on tackling large-scale decision-making problems through the development of novel solution frameworks that are broadly applicable beyond the applications studied.
Discrete Optimization: Applications of discrete optimization in logistics and supply-chain management problems have always been among my primary areas of interest. I worked on algorithms to efficiently solve variants of the lot-sizing problem with supplier selection and quantity discounts [5,6]. I’ve also developed exact optimization algorithms for consistent vehicle routing problems, a particularly complex setting for routing and scheduling, that utilize branch-and-price and lazy-cut generation. The problem of study is motivated by the operations of an instrument-calibration company in Europe, and the algorithms I developed lead to more consistent schedules, potential cost savings, and faster deliveries . The modeling framework and algorithms I’ve developed can be used more broadly for other routing and scheduling problems involving consistency requirements, including problems faced by large shipping companies like FedEx and UPS, where ensuring consistent vehicle arrivals is a critical operational consideration.
Data-Driven Optimization: Taking advantage of the abundant data in today’s industries requires new techniques that integrate data analytics with advanced optimization methods. In , I couple geo-spatial and temporal analytics with optimization to identify transportation lanes and trade opportunities among a pair of companies. Specifically, we design an end-to-end decision making framework that utilizes real GPS and sensor data collected from multiple carriers in the US together with spatial-clustering algorithms to identify frequent and repeatable asset movement patterns. The identified lanes are then matched using discrete optimization to evaluate the environmental and financial benefits of the proposed framework. The integration of big-data analytics with optimization-based decision-making is of growing interest in a variety of disciplines, and my PhD training in a department that integrates operations and information systems puts me in a good position to continue to contribute to this literature stream for years to come.
Procurement Auctions and Market Design: As one of the main components of my PhD dissertation research, I am studying efficient combinatorial auction/exchange mechanisms, proposing an electronic transportation market as a means of profitable collaboration among multiple trucking transportation companies . The paper also studies how to implement different payment rules to analyze the potential economic outcomes on a transportation market. This stream of research integrates economic theory and operations research to overcome complexities that arise in market design, where environmental and financial benefits can be achieved by balancing the trade-offs among communication complexity, computational complexity, and truth-telling incentives. Exploring different business domains where combinatorial auctions are applicable will continue to be one of my areas of ongoing research.
Machine Learning and Optimization: During my summer 2019 internship with Fidelity Investments research group in Boston, I worked on leveraging optimization techniques to improve binary classification machine-learning models. In predictive modeling, it is often observed that for some of the observations in the data-set, the resulting predictions contradict what practitioners consider reasonable. For instance, a binary classifier may predict a person with generally high quality financial metrics to default on a loan, or a soccer player may be predicted to suffer from a certain type of muscle injury even though a kinesiologist might recognize the athlete as low risk. In , I am working on investigating how expert opinion can be incorporated into predictive models (here, logistic regression) in order to boost their performance and practical interpretability, in addition to increasing the chance of use in the real-world. We hope that this will result in practitioners and experts having more trust in what they have typically viewed as black-box predictions.
Papers Under Review
 “Consistent routing and scheduling with simultaneous pickups and deliveries,” under 2nd round review at Production and Operations Management (joint work with David Bergman and Robert Day).
 “Designing a sustainable backhaul framework using telematics sensor data and analytics,” under first round review at MIS Quarterly (after a reject-and-resubmit) (joint work with Sudip Bhattacharjee, Robert Day and David Bergman).
 “Combinatorial exchanges for truckload transportation” (joint work with Robert Day)
 “Constrained logistic regression to avoid undesirable predictions” (joint work with Serdar Kadioglu and David Bergman)
 M. Mazdeh, M. Emadikhiav, and I. Parsa, “A heuristic to solve the dynamic lot sizing problem with supplier selection and quantity discounts,” Computers and Industrial Engineering, vol.85, pp. 33-43, 2015-05.
 I. Parsa, M. Emadikhiav, M. Mazdeh and S. Mehrani, “A multi supplier lot sizing strategy using dynamic programming,” International Journal of Industrial Engineering Computations, vol. 4, no. 1,
pp. 61-70, 2013-01.