Data Science Commons^™

Federated Learning For Early Medical Diagnosis: Enhanced Diabetic Retinopathy Detection In Smart Healthcare, Mohammad Nasajpour Esfahani

Master's Theses

This thesis investigates the role of federated learning as a privacy-preserving solution for modern healthcare challenges. In traditional machine learning, sensitive medical data must be centralized for model training, raising concerns about privacy, security, and regulatory compliance. Federated learning offers an alternative by allowing hospitals, clinics, and personal health devices to collaboratively train shared models without exchanging raw patient data. The study first explores how federated learning is being used across various healthcare domains, including cancer detection, medical imaging, and disease prediction— highlighting its potential to support secure collaboration across institutions. It addresses key benefits such as data privacy, scalability, …

Machine Learning For Predictive Energy And Emissions Modeling Of Vehicles And Power Grids In The United States, S M Tanvir Faysal Alam Chowdhoury

Dissertations

The environmental benefits of electric vehicle (EV) adoption depend on more than replacing internal combustion engine vehicles with electric powertrains. EV adoption reshapes electricity demand, interacts with regional generation mixes, and influences travel behavior and congestion, creating a coupled transportation-energy system in which vehicle and power-plant emissions must be evaluated together. This dissertation develops machine-learning frameworks for predicting energy consumption and emissions from vehicles and power grids under rising EV adoption. The first component forecasts grid emissions from EV charging. Using simulation data from NREL's Cambium database, a Prophet-based time-series framework predicts carbon dioxide, nitrous oxide, and methane emission rates …

Is Hockey Still Canada’S Game? How Usa Teams Have Won Every Stanley Cup Since 1994, Aaron Montgomery, Long Doan, Joe Demaio, Michael Frankel

Symposium of Student Scholars

The last Canadian team to win Lord Stanley’s cup in the National Hockey League was the Montreal Canadiens in 1993. Since then, each championship has been claimed by a team geographically located in the United States. Is this streak unusual? Perhaps it is particularly unusual in light of the fact that Hockey is known as Canada’s game. Is Hockey in its modern incarnation still Canada’s game? Given the long history of the NHL should we expect such a streak to occur at some point in time? Are we too fixated on the geographic location of the teams in question? Perhaps …

Prism (Proxy Recognition And Inclusion Scoring Method), Destiny Raburnel, Crystal Tubbs, Md Abdullah Al Hafiz Khan

Symposium of Student Scholars

AI-driven automated hiring tools are reshaping how companies find talent, but they often reproduce the hidden biases embedded in their training data. Our project, PRISM (Proxy Recognition and Inclusion Scoring Method), investigates how subtle demographic signals, specifically first names associated with gender and race, influence AI resume screening even when candidates have identical qualifications. We built a controlled dataset of resumes that are identical in every way except for the applicant's first name, with each resume using a racially neutral surname to isolate how first names alone affect scoring. We tested these resumes against job postings in technology, healthcare, and …

A Dual-Model Machine Learning Framework For Predictive Maintenance Of Industrial Digital Press Components, Richmond Darko, Emmanuel Adabor

African Conference on Information Systems and Technology

This study presents a novel dual-model predictive maintenance framework designed to improve maintenance scheduling for components in industrial digital presses. The framework integrates two complementary approaches: a Threshold-Based Maintenance Approach (TBMA) for components operating within acceptable usage limits, and an Overdue Severity-Based Maintenance Approach (OSBMA) for those that have exceeded their expected lifespans or show signs of critical degradation. This study uses real-world operational data from a Konica Minolta C6000 press. It applies advanced machine learning models, including Gradient Boosting Machines and Random Forest for classification, and Generalized Additive Models (GAM) for Remaining Useful Life (RUL) prediction. The goal is …

A Complete Transfer Learning-Based Pipeline For Discriminating Between Select Pathogenic Yeasts From Microscopy Photographs, Ryan A. Parker, Danielle S. Hannagan, Jan H. Strydom, Christopher J. Boon, Jessica Fussell, Chelbie A. Mitchell, Katie L. Moerschel, Aura G. Valter-Franco, Christopher Cornelison

Faculty Articles

Pathogenic yeasts are an increasing concern in healthcare, with species like Candida auris often displaying drug resistance and causing high mortality in immunocompromised patients. The need for rapid and accessible diagnostic methods for accurate yeast identification is critical, especially in resource-limited settings. This study presents a convolutional neural network (CNN)-based approach for classifying pathogenic yeast species from microscopy images. Using transfer learning, we trained the model to identify six yeast species from simple micrographs, achieving high classification accuracy (93.91% at the patch level, 99.09% at the whole image level) and low misclassification rates across species, with the best performing model. …

Dynamic Knowledge Elicitation: Leveraging Student Feedback For Improved Language Model Distillation, Reuven Muller

Master's Theses

Large Language Models (LLMs) have significantly advanced the field of natural language processing but remain resource-intensive and impractical for many organizations. Specialist models offer a viable alternative, often developed through Knowledge Distillation (KD) techniques. However, traditional KD methods rely on predefined static datasets to elicit knowledge from the teacher model, failing to dynamically address the weaknesses of the student model during training. This research introduces two novel methods for adaptive knowledge elicitation: Feedback-Driven Question Generation and Agent-Based Targeted Question Generation. These methods iteratively expand the training dataset based on the student model’s performance, leveraging a teacher model to generate targeted …

Trends From 20 Years Of Artificial Intelligence In Financial Services In Africa, Nthabiseng Moela, Lerato Matlala, Jackie Ma, Dipuo Maphutha, Hossana Twinomurinzi

African Conference on Information Systems and Technology

The need for financial inclusion in Africa, particularly for marginalised groups like women and small businesses, highlights the importance of leveraging Artificial Intelligence (AI). This study provides a bibliometric analysis of AI's integration into African financial services from 2003 to 2023. The key results show a significant increase in AI use, particularly in fraud detection, credit risk prediction, and stock market volatility forecasting, with 49% of the research coming from South Africa, Nigeria, and Tunisia. However, areas like financial development management, inflation control, and gender disparities in loan access remain underexplored. The emphasis has been on the technical implementation of …

20 Years Of Repo Interest Rate Determination Using Ai: Global Trends And Africa, Takalani Rasalanavho, Henry Hondo, Kevin Julius, Marius Alembong, Sikelela Madonsela, Hossana Twinomurinzi

African Conference on Information Systems and Technology

This study investigated the application of artificial intelligence (AI) in determining repo interest rates, which play a vital role in guiding monetary policy, controlling inflation, and ensuring economic stability. Through a bibliometric review of research from 2004 to 2024, the findings highlight AI's transformative impact, particularly in forecasting, optimising repo rate decisions, and improving risk assessment for more effective monetary policy. However, the study also identifies a significant gap in AI usage for repo rate determination in African countries, with contributions largely limited to Ghana, Nigeria, Egypt, and South Africa. This underrepresentation poses a risk of Africa falling behind in …

The Classification Of Internet Memes Through Supervised And Unsupervised Machine Learning Algorithms, William H. Little

Symposium of Student Scholars

Memes, those captivating internet phenomena, effortlessly deliver online entertainment. By leveraging time-series data from Google Trends, we can vividly illustrate and dissect the dynamic trends in meme popularity. Previous studies have discerned four distinct post-peak popularity patterns— "smoothly decaying," "spikey decaying," "leveling off," and "long-term growth"—and elegantly modeled these using ordinary differential equations.

This research introduces a programmatic approach that harnesses both supervised and unsupervised machine learning algorithms. The dataset, now expanded to over 2000 elements, becomes the canvas for exploration. The K-means algorithm identifies clusters, which then serve as labels for the supervised SVC algorithm. The overarching goal is …

Accessing Advanced National Supercomputing And Storage Resources For Computational Research, Ramazan Aygun

All Things Open

This presentation will cover ACCESS (Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support), and Kennesaw State University's involvement in Open Science Data Federation program as a data origin to help researchers and educators with or without supporting grants to utilize the nation’s advanced computing systems and services. ACCESS, a program established and funded by the National Science Foundation, is an ecosystem with capabilities for new modes of research and further democratizing participation. The presentation covers how to apply for allocations on ACCESS. The last part of the presentation will briefly explain Open Science Data Federation and Kennesaw State University's involvement as …

A Holistic And Collaborative Behavioral Health Detection Framework Using Sensitive Police Narratives, Martin Keagan Wynne Brown

Dissertations

Identifying behavioral health is paramount for law enforcement officers to provide appropriate follow-up community care. In the current practice, law enforcement offices manually identify these behavioral health cases to allow the designation of the relevant follow-up resources. Police reports generated by officers' response to 911 calls remain an untapped resource for identifying such incidents. Therefore, we advocate for the incorporation of manual annotations from experts, natural language processing (NLP), active learning, advanced machine learning, and ensemble techniques to detect behavioral health cases within police reports. In this dissertation, we develop tools and frameworks to automatically detect behavioral health cases from …

Multi-Modality Transformer For E-Commerce: Inferring User Purchase Intention To Bridge The Query-Product Gap, Srivatsa Mallapragada

Dissertations

The rapid growth of e-commerce has necessitated the development of sophisticated product retrieval systems that can effectively match user queries with relevant products. However, the semantic gap between queries and products remains a significant challenge, as traditional retrieval methods often fail to capture the nuances of user purchase intentions. E-commerce click-stream data and product catalogs offer critical user behavior insights and product knowledge that are untapped in the current product search algorithms. This dissertation presents learning strategies that leverage the query-product transaction logs to enrich the pipeline of our proposed multi-modal transformer model, which transforms initial user queries into pseudo …

Cm-Ii Meditation As An Intervention To Reduce Stress And Improve Attention: A Study Of Ml Detection, Spectral Analysis, And Hrv Metrics, Sreekanth Gopi

Master of Science in Computer Science Theses

Students frequently face heightened stress due to academic and social pressures, particularly in de- manding fields like computer science and engineering. These challenges are often associated with serious mental health issues, including ADHD (Attention Deficit Hyperactivity Disorder), depression, and an increased risk of suicide. The average student attention span has notably decreased from 21⁄2 minutes to just 47 seconds, and now it typically takes about 25 minutes to switch attention to a new task (Mark, 2023). Research findings suggest that over 95% of individuals who die by suicide have been diagnosed with depression (Shahtahmasebi, 2013), and almost 20% of students …

Data Quality Checks: Implementation With Popular Data Collection Crowdsourcing Platforms, James Down, Gregory Balkcom, Kristine Duncan, Ngan (An) Truong, Andrew Lewis

Symposium of Student Scholars

The utilization of online crowdsourcing platforms for data collection has increased over the past two decades in the field of public health due to the ease of use, the cost-saving benefits, the speed of the data collection process, and the accessibility of a potentially true representative population. Although these platforms offer many advantages to researchers, significant drawbacks exist, such as poor data quality, that threaten the reliability and validity of the study. Previous studies have examined data quality concerns, but differences in results arise due to variations in study designs, disciplinary contexts, and the platforms being investigated. Therefore, this study …

Machine Learning In Minecraft: Proof Of Concept For Object Detection Oriented Autonomous Bots In Minecraft, John Merkin

Symposium of Student Scholars

Machine learning provides new methods of problem solving through applied pattern recognition. An interesting challenge is to utilize machine learning in the automation of tasks and behaviors in virtual environments. Minecraft is an open-world, sandbox style game giving players nearly limitless freedom to alter a procedurally generated world. In the survival game mode, the player must collect resources to craft tools and build structures. The collection of resources can be tedious, so this project seeks to automate the standard initial task of collecting wood. By combining a convolutional neural network with API, a bot can collect resources while remaining scalable …

Codesigning A Big Data Analytic Tool For Girl Child Learner Drop Out From Eastern Cape Province -South Africa, Nobert Rangarirai Jere, Nosipho Carol Mavuso, Nelly Sharpley

African Conference on Information Systems and Technology

Developing sustainable solutions is critical for adoption of digital solutions. As the high number of learners dropping out of school continues to increase, it is critical to find innovative ways of predicting and preventing high drop out. Current literature has documented a number of factors that influence learner drop out. Innovative ideas, techniques and activities have been undertaken to motivate learners to stay at school. It is unfortunate that most of the initiatives have not helped to avoid drop out of learners. The study is based on a mixed approached that was used targeting female learns from Oliver Tambo District …

Quantification Of Various Types Of Biases In Large Language Models, Sudhashree Sayenju

Doctor of Data Science and Analytics Dissertations

Natural Language Processing (NLP) systems are included everywhere on the internet from search engines, language translations to more advanced systems like voice assistant and customer service. Since humans are always on the receiving end of NLP technologies, it is very important to analyze whether or not the Large Language Models (LLMs) in use have bias and are therefore unfair. The majority of the research in NLP bias has focused on societal stereotype biases embedded in LLMs. However, our research focuses on all types of biases, namely model class level bias, stereotype bias and domain bias present in LLMs. Model class …

Operation Enduring Freedom: Improving Mission Effectiveness By Identifying Trends In Successful Terrorism, Dalton Shaver

Symposium of Student Scholars

This research examines how the characteristics of terrorist attacks predict the chance of an attack succeeding, where an attack is defined as successful if the intended attack type is carried out. Data from The Global Terrorism Database (https://www.start.umd.edu/gtd) was analyzed across three geographical missions within Operation Enduring Freedom: Trans-Sahara, Horn of Africa, and the Philippines. The three models were able to distinguish between successful and unsuccessful attacks at 78.74%, 82.11%, 74.25%, respectively. Using predicted probabilities of success obtained from each logistic regression models, the medians were plotted to compare the characteristics of terrorist attacks across missions. The coefficients for each …

Employee Attrition: Analyzing Factors Influencing Job Satisfaction Of Ibm Data Scientists, Graham Nash

Symposium of Student Scholars

Employee attrition is a relevant issue that every business employer must consider when gauging the effectiveness of their employees. Whether or not an employee chooses to leave their job can come from a multitude of factors. As a result, employers need to develop methods in which they can measure attrition by calculating the several qualities of their employees. Factors like their age, years with the company, which department they work in, their level of education, their job role, and even their marital status are all considered by employers to assist in predicting employee attrition. This project will be analyzing a …

Reducing Restaurant Inventory Costs Through Sales Forecasting, Tyler Mason, Chris Schoen, Trevor Gilbert, Jonathan Enriquez

Senior Design Project For Engineers

Family Restaurant is a local restaurant in the greater Atlanta area that serves a variety of dishes that include an assortment of 19 different proteins. Currently, Family Restaurant places protein orders based on business intuition, and tends to over-stock and sometimes under-stock. To minimize inventory costs by reducing over-stocking and preventing under-stocking of proteins, we applied Facebook Prophet (FB Prophet), ARIMA, and XG Boost machine learning models to predict protein demand and then fed these results into a Fixed Time Period inventory model to make an overall order suggestion based on the specified time period. We trained our models on …

A Multistage Framework For Detection Of Very Small Objects, Duleep Rathgamage Don, Ramazan Aygun, Mahmut Karakaya

Published and Grey Literature from PhD Candidates

Small object detection is one of the most challenging problems in computer vision. Algorithms based on state-of-the-art object detection methods such as R-CNN, SSD, FPN, and YOLO fail to detect objects of very small sizes. In this study, we propose a novel method to detect very small objects, smaller than 8×8 pixels, that appear in a complex background. The proposed method is a multistage framework consisting of an unsupervised algorithm and three separately trained supervised algorithms. The unsupervised algorithm extracts ROIs from a high-resolution image. Then the ROIs are upsampled using SRGAN, and the enhanced ROIs are detected by our …

Fairness And Privacy In Machine Learning Algorithms, Neha Bhargava

Master of Science in Computer Science Theses

Roughly 2.5 quintillion bytes of data is generated daily in this digital era. Manual processing of such huge amounts of data to extract useful information is nearly impossible but with the widespread use of machine learning algorithms and their ability to process enormous data in a fast, cost-effective, and scalable way has proven to be a preferred choice to glean useful insights and solve business problems in many domains. With this widespread use of machine learning algorithms there has always been concerns about the ethical issues that may arise from the use of this modern technology. While achieving high accuracies, …

A Maturity Model Of Data Modeling In Self-Service Business Intelligence Software, Anna Kurenkov

Master of Science in Information Technology Theses

Although Self-Service Business Intelligence (SSBI) is continually being adopted in various industries, there is a lack of research focused on data modeling in SSBI. This research aims to fill that research gap and propose a maturity model for SSBI data modeling which is generalizeable between different software and applicable for users of all technical backgrounds. Through extensive literature review, a five-tier maturity model was proposed, explained, and instantiated in PowerBI and Tableau. The testing of the model was found to be simple and intuitive, and the research concludes that the model is applicable to enterprise SSBI environments. This research is …

Appley: Approximate Shapley Values For Model Explainability In Linear Time, Md Shafiul Alam

Doctor of Data Science and Analytics Dissertations

We have seen complex deep learning models outperforming human benchmarks in many areas (e.g. computer vision, natural language processing). Clever architectures and higher model complexity are two of the major drivers of such outstanding performances. Higher model complexity generally makes the decision-making process of a model opaque to human perception. But understanding the decision-making process is very important for many reasons including enhancing trust in the model's prediction, improving model robustness, gaining actionable insight from why a model made a particular prediction, and discovering new knowledge about a problem. Model explainability has been an active area of research for some …

A New Kind Of Data Science: The Need For Ethical Analytics, Jonathan Boardman

Published and Grey Literature from PhD Candidates

Ethics can no longer be regarded as an add-on in data science and analytics. This paper argues for the necessity of formalizing a new, practically-oriented sub-discipline of AI ethics by outlining the needs, highlighting shortcomings in current approaches, and providing a framework for ethical analytics, which is concerned with the study of the ethical issues surrounding the development, deployment, and/or dissemination of ML/AI systems and data science research, as well as the development of tools and procedures to mitigate ethical harms. While data science and machine learning are primarily concerned with data from start to finish, ethical analytics is concerned …

Ethical Analytics: A Framework For A Practically-Oriented Sub-Discipline Of Ai Ethics, Jonathan Boardman

Doctor of Data Science and Analytics Dissertations

Ethics can no longer be regarded as an add-on in data science and analytics. This dissertation argues for the necessity of formalizing a new, practically-oriented sub-discipline of AI Ethics by outlining needs, highlighting shortcomings in current approaches, and providing a framework for Ethical Analytics, a field concerned with the study of the ethical issues surrounding the development, deployment, and/or dissemination of ML/AI systems and data science research, as well as the development of tools and procedures to mitigate ethical harms. While data science and machine learning are primarily concerned with data from start to finish, ethical analytics is concerned primarily …

Multidisciplinarity In Data Science Curricula, Hossana Twinomurinzi, Siyabonga Mhlongo, Kelvin J. Bwalya, Tebogo Bokaba, Steven Mbeya

African Conference on Information Systems and Technology

This paper sought to identify and compare disciplinary emphases in data science curricula across South Africa’s 26 public universities using a website scoping review method. The key findings reveal that only 12 of the 26 universities offer data science programmes that are publicly accessible on their websites. Of those 12, only 5 offer data science at the undergraduate level, and these undergraduate programmes are objectified (entirely leaning) to the science, technology, engineering, and mathematics (STEM) disciplines. Only seven of the universities offer a few non-STEM subjects with only one offering more non-STEM subjects compared to STEM subjects. The implications are …

Debiasing Cyber Incidents – Correcting For Reporting Delays And Under-Reporting, Seema Sangari

Doctor of Data Science and Analytics Dissertations

This research addresses two key problems in the cyber insurance industry – reporting delays and under-reporting of cyber incidents. Both problems are important to understand the true picture of cyber incident rates. While reporting delays addresses the problem of delays in reporting due to delays in timely detection, under-reporting addresses the problem of cyber incidents frequently under-reported due to brand damage, reputation risk and eventual financial impacts.

The problem of reporting delays in cyber incidents is resolved by generating the distribution of reporting delays and fitting modeled parametric distributions on the given domain. The reporting delay distribution was found to …

Novel Instance-Level Weighted Loss Function For Imbalanced Learning, Trent Geisler

Doctor of Data Science and Analytics Dissertations

Binary classification using imbalanced datasets remains a challenge. Typically, supervised learning algorithms minimize the binary cross-entropy objective function to determine the final parameter estimates. This objective function assumes an equal class distribution between the minority (i.e. events) and majority (i.e. non-events) classes, which almost never exists in real-world modeling. In the imbalanced data setting, the equal class distribution is grossly violated, and the resulting parameter estimates are biased toward the majority class. To overcome the bias and improve model generalization, we focus on modifying the original binary cross-entropy objective function by uniquely weighting each minority class observation. We base our …