Operational Research Commons^™

Bridging Models And Real-World Performance: Capturing Represented, Hidden, And Unrepresented Dynamics In Complex Systems, Dixizi Liu

All Dissertations

Complex systems often depend on adaptive performance that is difficult to capture using structured data or formal models alone. This dissertation asks how predictive modeling, coordination-cost measurement, and simulation can be integrated to better represent the visible, hidden, and unrepresented dynamics that shape real-world system performance.

Sterile Processing Departments (SPDs) provide the empirical context for this work because they support safe and timely surgical care while operating under high workload variability, time pressure, and tight coordination with operating rooms. Hospitals rely on SPDs to clean, assemble, sterilize, and deliver surgical instrument trays so surgeries can start on time and proceed …

Business Process Redesign For Reducing Undelivered Product Return Losses In E-Commerce – An Explainable Ai Approach, Venkataraghavan Krishnaswamy, Deepa R, Himanshu Sharma

Journal of International Technology and Information Management

Product returns in e-commerce affect the profitability of the e-tailer. We adopt a two-stage approach to reduce undelivered product returns in an e-commerce firm. First, we develop and compare machine learning techniques—logistic regression, decision trees, Naïve Bayes, random forest, adaptive boosting, gradient boosting, stochastic gradient boosting, and deep neural networks—on their ability to predict undelivered returns. Next, we use explainable methods, such as relative importance and Shapley values, to develop insights from the best-performing machine learning model. Finally, we use these insights and the predictive model to redesign the firm’s order fulfillment and return processes. A Post-implementation evaluation of the …

Data-Driven Dynamic Decision-Making Using Discrete Optimization And Supervised Machine Learning, Navid Rashedi

Dartmouth College Ph.D Dissertations

In recent years, the operations research community has developed data-driven optimization techniques to solve complex combinatorial problems with the aid of machine learning. This thesis contributes to these efforts by combining machine learning with optimization to expedite online decision-making, with applications in transportation and healthcare.

In the domain of airline operations recovery, the focus is on the aircraft recovery process—repairing disrupted schedules by minimizing overall disruption costs. Traditional exact methods are too time-consuming, while heuristic approaches often yield poor solution quality and lack generalizability across varying formulations. To address these challenges, this research employs supervised machine learning to identify near-optimal …

Learning To Dogfight: Proximal Policy Optimization Vs. Double Deep Q Network For 2v2 Air Combat With Directed Energy Weapons In Afsim, Caden W. Wilson

Theses and Dissertations

This research utilizes reinforcement learning (RL) to train two blue agents each imbued with a directed energy weapon (DEW) in a 2v2 within visual range air combat maneuvering problem. A phased solution approach is employed to repeatedly tune and train several RL algorithm implementations: Proximal Policy Optimization (PPO) and Double Deep Q Network (DDQN). Phase I of training includes reward shaping for basic flight elements such as altitude, airspeed, and target proximity. Phase II of training builds off policies developed in Phase I, but rewards emphasize winning the aerial engagement by any means necessary. DDQN significantly outperforms PPO in Phase …

Machine Learning With Flight Data Recorder Data For Flight Fuel Consumption Predictions, Adam C. Levandowski

Theses and Dissertations

This study applies advanced Machine Learning (ML) to Flight Data Recorder (FDR) data for fuel consumption predictions. It explores feature engineering, model selection, and Hyper-Parameter Optimization (HPO) across all flight phases. Baseline models like Ordinary Least Squares (OLS) regression, Multi- Layer Perceptrons (MLPs), and decision trees are compared to Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs) with Gated Recurrent Unit (GRU) layers, and XGBoost. Results analyze segmentation strategies, tailored features, and model performance. A counterfactual analysis compares ML models to operational fuel predictions, demonstrating their deployment potential. Findings establish a foundation for future ML-driven advancements in aviation fuel optimization.

Multiphysics Modeling Of Solid Oxide Fuel Cells For Gradient Minimization And Inductive Loop Analysis In Impedance Spectroscopy Using Machine Learning-Based Microstructural Property Estimation, Muhammad Usman Khan

College of Graduate Studies: Theses & Dissertations

Solid oxide fuel cells have significant advantages in renewable energy utilization due to their high efficiency, fuel flexibility, and low emissions. However, despite the numerous efforts of technology, thermal and current density gradients and impedance behavior fluctuations are still causing performance degradation. A combined computational framework that integrates machine learning and three-dimensional Multiphysics modeling is needed to investigate and optimize the performance of solid oxide fuel cells. A machine learning model, trained on synthetic microstructure data by percolation analysis, is used to predict important microstructural parameters like triple phase boundary density and geometric tortuosity. These are then employed in a …

A Random Forest-Based Q-Learning Algorithm: Toward Interpretable Artificial Intelligence, Victor R. Rae

Theses and Dissertations

A growing demand exists for interpretable artificial intelligence models, leading to extensive research efforts to enhance the explainability and transparency of policies generated by reinforcement learning (RL) methods. This research develops random forest-based RL algorithms as a logical progression in this academic pursuit. The algorithms are evaluated using three standard benchmark environments from OpenAI gym — CartPole, MountainCar, and LunarLander — and compared to implementations of the Deep Q-learning Network (DQN) and Double DQN (DDQN) algorithms for various metrics, including performance, robustness, efficiency, and interpretability. The random forest-based algorithms exhibit superior performance to both neural network-based algorithms in two out …

Bayesian Augmentation Of Object Detection Algorithms To Enhance Object Classification Stability, Taylor D. Markham

Theses and Dissertations

Neural networks, despite their prowess in computer vision, often exhibit "flickering". Flickering occurs when networks fail to maintain consistent object representation across frames, leading to inaccurate and inconsistent output. This problem is particularly critical in mission-surety applications where reliable object recognition is crucial. This research presents a novel approach that combines existing object detection and tracking algorithms like YOLO and SORT with a Bayesian backend model. This Bayesian backend incorporates probabilistic reasoning to analyze the network's confidence in its predictions and infer the most likely object identity across multiple frames, effectively reducing flickering and enhancing robustness.

U.S. Army Cadet Command Branch Prediction Model, Daniel M. Krizan

Theses and Dissertations

The current system for providing US Army ROTC cadets their branches leaves significant uncertainty until the final pronouncement of branch assigned. This uncertainty can be alleviated by providing a prediction model for cadets to input personal data and desired branch to identify likelihood of receiving the request. This thesis produces a machine learning model capable of producing branch prediction for cadets.

Federated Analysis Of Wearables Data For United States Air Force Mental And Physical Readiness, Sung Yong O

Theses and Dissertations

This thesis investigates the use of machine learning and deep learning models within a federated learning framework to predict physical and mental readiness in military personnel, using wearable technology data. The collaboration with the 711th Human Performance Wing’s STRONG Lab highlights the importance of readiness as emphasized by the National Defense and Security Strategies. The study evaluates various predictive models, incorporating federated learning to ensure data privacy and security in healthcare systems. By analyzing a comprehensive dataset, the research aims to contribute to military readiness enhancement through technological advancements, supporting health and wellness initiatives to bolster the effectiveness of military …

Application Of Learning Processes For Improving Last-Mile Logistics Optimization At Scale, Seyedeh Shaghayegh Rabbanian

LSU Doctoral Dissertations

The escalating demands of omnichannel retailing, rapid urbanization and shifting customer behaviors have propelled last-mile vehicle routing logistics to the forefront of research. This last-mile phase, recognized as a significant contributor to costs and pollution in the supply chain, necessitates efficient route optimization to minimize expenses and environmental impact. This research delves into machine learning based techniques for solving large-scale Vehicle Routing Problem (VRP), a fundamental concern in last-mile logistics, aiming to optimize delivery vehicle routing amidst diverse customer nodes and operational constraints. Three primary research subproblems are analyzed: utilizing machine learning for constructive solutions, Variable Neighborhood Search (VNS) metaheuristic, …

Federated Active Learning For Network Intrusion Detection, Matthew D. R. Sauer

Theses and Dissertations

This thesis addresses challenges with detecting attacks on computer networks within a Federated Learning (FL) framework, when labeled instances are few. We explore the integration of active learning (AL) and semi-supervised learning (SSL). AL efficiently uses data that would otherwise be wasted or require substantial time for labeling. SSL provides capacity to train models that have a limited amount of labeled data, by utilizing additional unlabeled data that is available. We show how FL combined with AL or SSL can realize a detection system that adapts and trains quickly to new networks, reducing the total amount of data labeling needed. …

Classification And Analysis Of Twitter Bot And Troll Accounts, Callan P. Mccormick

Theses and Dissertations

This research trains, tests, and analyzes bot and troll classification models using publicly available, open source datasets. Specifically, it applies decision tree, random forest, feed forward neural networks, and long-short term memory neural networks with hyperparameters tuned via designed experiment to five labeled bot datasets created between 2011 and 2020 and one dataset labeling state-sponsored disinformation accounts or trolls. The first three models utilize account profile features, while the last model applies natural language processing techniques, specifically GloVe embedding, to analyze a user’s Tweet history. Results indicate that the random forest model outperforms the other three models with an average …

Short-Term Prediction Of Icu Admission For Covid-19 Inpatients, Yoon Sang Lee, Riyaz T. Sikora

Journal of International Technology and Information Management

Since the COVID-19 outbreak, many hospitals suffered from a surge of some high-risk inpatients needing to be admitted to the ICU. In this study, we propose a method

predicting the likelihood of COVID-19 inpatients’ admission to the ICU within a time frame of 12 hours. Four steps, the Bayesian Ridge Regression-based missing value imputation, the synthesis of training samples by the combination of two rows (the first and another row) of each patient, customized oversampling, and XGBoost classifier, are used for the proposed method. In the experiment, the AUC-ROC and F-score of our method is compared with those of other …

Credit Card Fraud Detection Using Machine Learning Techniques, Nermin Samy Elhusseny, Shimaa Mohamed Ouf, Amira M. Idrees Ami

Future Computing and Informatics Journal

This is a systematic literature review to reflect the previous studies that dealt with credit card fraud detection and highlight the different machine learning techniques to deal with this problem. Credit cards are now widely utilized daily. The globe has just begun to shift toward financial inclusion, with marginalized people being introduced to the financial sector. As a result of the high volume of e-commerce, there has been a significant increase in credit card fraud. One of the most important parts of today's banking sector is fraud detection. Fraud is one of the most serious concerns in terms of monetary …

Predicting The Likelihood And Scale Of Wildfires In California Using Meteorological And Vegetation Data, Matthew Walters

Graduate Theses and Dissertations

Wildfires have devastating ecological, environmental, economical, and public health impacts through the deterioration of water and air quality, CO2 emissions, property damage, and lung illnesses. The early detection and prevention of wildfires allow for the minimization of these risks. The use of Artificial Intelligence (AI) in wildfire detection and prediction has been highly researched as a tool to assist firefighters in stopping wildfires in its early stages. The three common wildfire prediction categories include image and video detection, behavior prediction, and susceptibility prediction. Data such as climate, weather, vegetation, satellite images, and historical wildfire data is most commonly used. Many …

An Exploratory Analysis Of Time Series Econometric Data For Retention Forecasting Using Deep Learning, John C. O'Donnell

Theses and Dissertations

Officer retention in the Air Force has been researched many times in an attempt to better predict the personnel needs of the Air Force for the future. There has been previous work done in regards to specific AFSCs and how their retention compares to specific yet similar private sector jobs. This study considers different econometric time series statistics as a feature space and an average Air Force officer separation rate as the response variable for the multivariate time series analysis deep learning techniques. The econometric indicators used in this study are New Business Formations, New Durable Good Orders, and the …

Training Logic And Random Forest Models To Predict It Spending, Jacob P. Batt

Theses and Dissertations

The Air Force must modernize, but the distribution of funds for technology remains as tight as ever. To this end, the Air Force Audit Agency is looking to utilize machine learning techniques to enhance their capabilities. This research explores Logistic Regression and Random Forest modeling to streamline data collection and cost classification. The final Logistic Regression model identified 4 significant attributes out of the 36 given and was 85 accurate in predicting whether a purchase amount was over or under $10,000. To expand beyond binary classification, a six-category classification Random Forest model was developed. It identified 6 significant attributes and …

Using Custom Ner Models To Extract Dod Specific Entities From Contracts, Kayla P. Haberstich

Theses and Dissertations

The Air Force Sustainment Center collected 3.7 million contracts onto the Air Force Research Laboratory’s high power computers. They are in the format of a .pdf or scanned document, making them unstructured data. The Data Analytics Resource Team extracted the documents into a textual format for use in further analysis. This thesis looks to extract four DOD specific entities (NSN, Part Number, CAGE Code, and Supplier Name) from the contracts using custom NER models. This newly extracted information will allow the Air Force to identify what parts are supplied by which vendors. This information along with historical CLIN pricing for …

Extracting Patterns In Medical Claims Data For Predicting Opioid Overdose, Ryan Sanders

Graduate Theses and Dissertations

The goal of this project is to develop an efficient methodology for extracting features from time-dependent variables in transaction data. Transaction data is collected at varying time intervals making feature extraction more difficult. Unsupervised representational learning techniques are investigated, and the results compared with those from other feature engineering techniques. A successful methodology provides features that improve the accuracy of any machine learning technique. This methodology is then applied to insurance claims data in order to find features to predict whether a patient is at risk of overdosing on opioids. This data covers prescription, inpatient, and outpatient transactions. Features created …

Algorithms For Multi-Objective Mixed Integer Programming Problems, Alvaro Miguel Sierra Altamiranda

USF Tampa Graduate Theses and Dissertations

This thesis presents a total of 3 groups of contributions related to multi-objective optimization. The first group includes the development of a new algorithm and an open-source user-friendly package for optimization over the efficient set for bi-objective mixed integer linear programs. The second group includes an application of a special case of optimization over the efficient on conservation planning problems modeled with modern portfolio theory. Finally, the third group presents a machine learning framework to enhance criterion space search algorithms for multi-objective binary linear programming.

In the first group of contributions, this thesis presents the first (criterion space search) algorithm …

Intelligent Software Tools For Recruiting, Swatee B. Kulkarni, Xiangdong Che

Journal of International Technology and Information Management

In this paper, we outline how recruiting and talent acquisition gained importance within HRM field, then give a brief introduction to the newest tools used by the professionals for recruiting and lastly, describe the Artificial Intelligence-based tools that have started playing an increasingly important role. We also provide further research suggestions for using artificial intelligence-based tools to make recruiting more efficient and cost-effective.

Identification Of Reverse Engineering Candidates Utilizing Machine Learning And Aircraft Cannibalization Data, Marc Banghart

International Journal of Aviation, Aeronautics, and Aerospace

As military aircraft continue to remain in service and age, cannibalization of parts is increasing. Proactive identification of parts that are at high risk for cannibalization will inform engineering processes such as reverse engineering, thus allowing potentially reducing lead time to develop new parts. The research objective was to develop a causal structure that can be used for prediction of when cannibalization actions may occur. Bayesian networks allow encoding of causality between various descriptive features given a data set. The method utilized a tabu search algorithm, identified the underlying causal structure and the associated node probabilities. The method is then …

Methods To Address Extreme Class Imbalance In Machine Learning Based Network Intrusion Detection Systems, Russell W. Walter

Theses and Dissertations

Despite the considerable academic interest in using machine learning methods to detect cyber attacks and malicious network traffic, there is little evidence that modern organizations employ such systems. Due to the targeted nature of attacks and cybercriminals’ constantly changing behavior, valid observations of attack traffic suitable for training a classifier are extremely rare. Rare positive cases combined with the fact that the overwhelming majority of network traffic is benign create an extreme class imbalance problem. Using publically available datasets, this research examines the class imbalance problem by using small samples of the attack observations to create multiple training sets that …

A Fortran Based Learning System Using Multilayer Back-Propagation Neural Network Techniques, Gregory L. Reinhart

Theses and Dissertations

An interactive computer system which allows the researcher to build an optimal neural network structure quickly, is developed and validated. This system assumes a single hidden layer perceptron structure and uses the back- propagation training technique. The software enables the researcher to quickly define a neural network structure, train the neural network, interrupt training at any point to analyze the status of the current network, re-start training at the interrupted point if desired, and analyze the final network using two- dimensional graphs, three-dimensional graphs, confusion matrices and saliency metrics. A technique for training, testing, and validating various network structures and …