Projects

Food Bank Research

Project 1: Optimal Pop-up Market Locations for Food Bank of Eastern and Central North Carolina

• Developed a mixed integer programming and stochastic model for optimizing the locations for pop-up markets under NSF Award#2125600

• Developed Bender’s decomposition with feasibility cut algorithm to reduce computation time for large-scale problems

Project 2: Estimation of food Insecurity in Census Tract Level using machine learning

• Leveraging Food Atlas Research data from USDA, modeled tree-based (random forest, decision-tree) and adaptive

boosting regression to estimate food insecurity rate at census tracts level[Model]

Project 3: Food Rescue Landscape Analysis for Feeding America

• Developed interactive dashboards using Tableau to highlight current food rescue scenario and a surrogate framework using generalized linear models, and tree-based models (decision tree, random Forest) to estimate the ESG rating and performance of food rescue organizations. Compared the result with direct ESG grouping using classification models (logistics regression, support vector machine, naive Bayes, and tree classifiers)

Project 4: Stochastic Modelling of Distribution Under Extreme Events in Hunger Relief Supply Chain

• Developed a two-stage stochastic model for optimal distribution of donated foods to food insecure neighborhoods under disaster events and proposed L-shaped Bender’s cut formulation with feasibility cut to make the model computationally efficient. Additionally, solved the problem with a dynamic programming approach.

Project 5: Agent Behaviors in Hunger Relief Network

• Used step-wise regression and regularized models (Lasso, Ridge, elastic net) learning approach for predicting

needs for food insecure persons in the US [Project]

• Used Schelling’s model of segregation in agent-based simulation optimization for understanding norms in

different agents. [Project]

Data Standardization and Pop-up Market Location Analysis

• Developed a smart and data-driven approach for data records of food bank recipients for the Food Bank of Eastern and Central North Carolina.

• Developed an integrated approach for connecting datasets with the website for pantry locations and powerBI dashboards

• Developing mixed-integer linear programming (MILP) model for strategically placing pop-up markets to increase accessibility to foods for neighbors

Sustainable Vaccine Supply Chain from a Developing Country  Perspective

This research focuses on current issues and develops a mathematical formulation for establishing a vaccine supply chain (VSC) that provides an integrated solution for the location, inventory, and routing of facilities, vaccines, and vehicles involved in distribution. The VSC's economic performance is ensured by a multi-objective formulation of mixed- integer programming (MIP) that minimizes overall cost while also ensuring economic sustainability by minimizing greenhouse gas (GHG) emissions and social sustainability by maximizing employment throughout the network. The vaccines' shelf-life limit, demand uncertainty, and related supply chain (SC) parameters are also addressed in the study.

See publication here

Collaborative Optimization of a Supply Chain

A typical supply chain for a product can be decomposed into two important disciplines to maximize the overall supply chain surplus. Inventory is the first discipline that determines the amount of raw materials and number of finished products to be held in warehouses, plants, and markets that can maximize the inventory surplus of a supply chain. Whereas, the transportation discipline tries to determine the amount of raw materials and products to be transported to plants and markets, respectively, to maximize transportation surplus. Recent research has already focused on optimizing an overall supply chain using a traditional approach where the supply chain of a business has been considered as a single discipline. However, transportation and inventory disciplines in a supply chain may come into conflict with each other. Optimizing both disciplines as a single disciplinary problem may not result in a practically efficient solution. In this project, a multidisciplinary optimization approach has been used to solve the supply chain for an industrial product to maximize the overall supply chain surplus. There are different approaches to formulating a multidisciplinary optimization problem. In most of the approaches, only one designer controls the overall system design. In this project, collaborative optimization (CO) has been used to provide design autonomy to each discipline. The result obtained from the CO approach gives a supply chain surplus that is considerably practical by addressing the interrelationship of the disciplines.

Faculty supervisor: Dr. Kais Bin Zaman, Professor, Department of Industrial and Production Engineering, BUET.

See full technical report