Data Science Commons^™

Mapping Urban Tree Canopy Using Publicly Available Satellite Data, Rosemary Mcguinness

Theses and Dissertations

This project addresses the need for accessible, cost-effective tools for quantifying spatial and temporal changes in tree canopy cover in urban areas. Urban tree canopy provides a wide range of ecosystem services, including lowering air temperatures, reducing pollution, and mitigating stormwater runoff. Cities around the world have placed the expansion of their urban forests at the center of their sustainability goals. Consistent and timely data on urban tree canopy is essential for urban greening initiatives to succeed. Existing methods of accessing information about urban tree canopy are highly technical, costly, and labor-intensive, while the freely available source of tree canopy …

A Machine Learning Framework For Predicting Fabrication Hours For Industrial Steel Structure Projects, Dalia Ibrahim

Theses and Dissertations

Construction projects are considered high risk projects especially due to their required large capital making them require extreme attention in estimation as overestimating a project will lead to losing bids and underestimating them will lead to incurring more costs than budgeted resulting in losses. However, estimators are often faced with very tight timelines to finish their estimates leading them to primarily rely on their experience disregarding some crucial factors resulting in inaccurate estimates. In the steel structures industry, the steel fabrication phase accounts for 30 to 40% of the overall project cost; in addition, the steel industry is labor driven; …

Impact Of Operational Ladar Occlusions On Point Cloud Instance Segmentation, Andrew D. Gibson

Theses and Dissertations

Data exploitation techniques are the enabler for technological advancements in military ISR applications of ladar ISR. By identifying instances of military objects in observed scenes, point cloud deep learning models can unlock new standards of real-time information delivery to warfighters. Although current deep learning training datasets do not include real-world collection occlusions consistent with military applications, this research characterizes SPT model performance by adding occlusions to the DALESObjects dataset via artificial flyby simulations.We find that a baseline model trained on unoccluded data suffers performance degradation on both semantic and instance segmentation tasks when evaluated on occluded data, but that the …

A Spatial Decision Support System For Rent Estimation Of Retail Spaces In Manhattan Using Geographically Weighted Regression And Spatial Regression, Andie M. Migden Miller

Theses and Dissertations

This report outlines an automated, three-phase Spatial Decision Support System that creates models to estimate rent of retail spaces across Manhattan. First, enrich data with predictors. Second, optimize spatially aware neighborhood-level models by combining GWR, spatial regression, and non-spatial regression. Finally, visualize results in an Esri-based WebApp.

Representation Learning For Generative Models With Applications To Healthcare, Astronautics, And Aviation, Van Minh Nguyen

Theses and Dissertations

This dissertation explores applications of representation learning and generative models to challenges in healthcare, astronautics, and aviation.

The first part investigates the use of Generative Adversarial Networks (GANs) to synthesize realistic electronic health record (EHR) data. An initial attempt at training a GAN on the MIMIC-IV dataset encountered stability and convergence issues, motivating a deeper study of 1-Lipschitz regularization techniques for Auxiliary Classifier GANs (AC-GANs). An extensive ablation study on the CIFAR-10 dataset found that Spectral Normalization is key for AC-GAN stability and performance, while Weight Clipping fails to converge without Spectral Normalization. Analysis of the training dynamics provided further …

Denoising Diffusion Probabilistic Models Based Accelerated Mri, Alexander Francis Bugielski

Theses and Dissertations

Magnetic Resonance Imaging (MRI) is a cornerstone in obtaining intricate visualizations of anatomy and physiological processes within the human body. However, its extensive scan duration not only causes patient discomfort but also increases the likelihood of motion-induced artifacts in the images. To address such a challenge, this study investigates deep neural network models for reconstructing high-resolution MRI images from noisy and significantly undersampled data in a supervised learning manner. Specifically, it compares three models: a conventional U-Net, a self-attentive U-Net, and an innovative probabilistic diffusion model that builds upon the self-attentive U-Net architecture. These models are evaluated on their ability …

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 …

A Staged Framework For Llm-Powered Information Extraction In Government Contracts, Jung H. Yae

Theses and Dissertations

The manual extraction of meaningful insights and conversion of content into structured forms to enhance document processing require substantial resources and are susceptible to errors. Despite numerous applications of various Natural Language Processing (NLP) models to streamline the manual process, challenges persist due to domain-specific data constraints and the deficiency of annotated data. This study attempts to address these challenges by leveraging a Large Language Model (LLM) to analyze government contracts. Through rigorous evaluation, we demonstrate the LLM’s effectiveness in information extraction and mitigating hallucinations, achieving a 87.86% accuracy in metadata extraction.

On Intrinsic Dimensionality Of Data Sets And Neural Networks, Ori Chachmo

Theses and Dissertations

The concept of Intrinsic Dimensionality (ID) is of special interest in the field of Neural Networks (NNs) since it promotes both (a) a deeper understanding of the underlying mechanisms, and (b) embraces parsimonious modeling (that is, building the right-sized model for the task) with associated benefits to processing speed and storage requirements. This thesis explores the concept of ID via two separate, but related, questions. First, we study the potential of NN ID prediction by exploiting easily obtained quantities measured on the data. We then explore NN ID as an independent concept by comparing the results of different methods for …

Automated Image Registration For Titanium Aircraft Components Via Resolution-Robust Parallel Neural Networks, Paige T. Luebbering

Theses and Dissertations

Titanium alloys are vital to the structural integrity of military and commercial aircraft, comprising numerous critical components. These components are composed of microtexture regions (MTRs) that, at a specific size and orientation, can lead to aircraft failure. Existing MTR testing methods, such as Electron Backscatter Diffraction, often fall short in effectively detecting these MTRs without causing damage to the component. Addressing this gap, this thesis develops a Parallel Convolutional Neural Network (CNN) model tailored for multi-resolution image registration of Polarized Light Microscopy (PLM) images to enhance MTR identification in a non-invasive manner. The findings reveal a significant enhancement in the …

Analyzing The Effects Of Data Variability & Volume On Predicting Particulate Matter (Pm2.5): Insights From A Machine Learning Approach, Jada A. Macharie

Theses and Dissertations

Accurately predicting PM_2.5 concentrations are imperative to the future of public health and environmental policies. Machine learning models incorporating spatial and temporal datasets to predict PM_2.5 are often limited by data availability constraints and poor resolution satellite imagery.

Adaptive Multi-Label Classification On Drifting Data Streams, Martha Roseberry

Theses and Dissertations

Drifting data streams and multi-label data are both challenging problems. When multi-label data arrives as a stream, the challenges of both problems must be addressed along with additional challenges unique to the combined problem. Algorithms must be fast and flexible, able to match both the speed and evolving nature of the stream. We propose four methods for learning from multi-label drifting data streams. First, a multi-label k Nearest Neighbors with Self Adjusting Memory (ML-SAM-kNN) exploits short- and long-term memories to predict the current and evolving states of the data stream. Second, a punitive k nearest neighbors algorithm with a self-adjusting …

Developing Machine Learning And Time-Series Analysis Methods With Applications In Diverse Fields, Muhammed Aljifri

Theses and Dissertations

This dissertation introduces methodologies that combine machine learning models with time-series analysis to tackle data analysis challenges in varied fields. The first study enhances the traditional cumulative sum control charts with machine learning models to leverage their predictive power for better detection of process shifts, applying this advanced control chart to monitor hospital readmission rates. The second project develops multi-layer models for predicting chemical concentrations from ultraviolet-visible spectroscopy data, specifically addressing the challenge of analyzing chemicals with a wide range of concentrations. The third study presents a new method for detecting multiple changepoints in autocorrelated ordinal time series, using the …

Large Language Models, Prompting, And Synthetic Data Generation For Continual Named Entity Recognition, Charles I. Cutler

Theses and Dissertations

With the ever-growing amount of textual data, the task of Named Entity Recognition (NER) is vital to Natural Language Processing (NLP), a field which focuses on enabling computers to understand and manipulate human language. NER enables the extraction of information from unstructured text. Accurate information extraction is crucial for applications ranging from information retrieval to systems for question-answering. To ensure that NER models are robust to changes in data distributions and capable of recognizing new entity types, one may consider expanding the capabilities of an existing model. Continual learning is a paradigm within machine learning. It studies the objective of …

Penalized Interpolating B-Splines And Their Applications, Kylee L. Hartman-Caballero

Theses and Dissertations

One of the most studied data analysis techniques in Numerical Analysis is interpolation. Interpolation is used in a variety of fields, namely computer graphic design and biomedical research. Among interpolation techniques, cubic splines have been viewed as the standard since at least the 1960s, due to their ease of computation, numerical stability, and the relative smoothness of the interpolating curve. However, cubic splines have notable drawbacks, such as their lack of local control and necessary knowledge of boundary conditions. Arguably a more versatile interpolation technique is the use of B-splines. B-splines, a relative of Bézier curves, allow local control through …

Title: I: L1-Norm Matrix Completion For Recommender Systems Ii: Conjecturing-Based Classification, Fatemeh Valizadeh Gamchi

Theses and Dissertations

Recommendation systems are essential for providing personalized user experiences, but their performance can be affected by outliers especially in traditional collaborative filtering methods that use the L2-norm. To address this challenge, we developed two new algorithms, SharpEl1rs and SharpEl1rs-Impute, based on the L1-norm to improve resistance against extreme values and effectively handle missing data. Our experimental setting was designed to compare these proposed methods with existing techniques. Then our algorithms are applied to real datasets to assess their performance, with findings indicating that our proposed models offer improved accuracy in some cases and solid performance in others for industrial-scale recommendation …

Improving The Quality Of Micronutrient Food Composition Data In Sub-Saharan Africa Using Predictive Analytics, Junior Mbuyamba Muka

Theses and Dissertations

High-quality food composition data are indispensable for decision making in several health, agricultural, and nutrition-related activities. Analytical data are the most accurate type of food composition data (FCD). Yet, most food composition databases (FCDBs) compiled in Sub-Saharan Africa only contain a small amount of analytical data collected from foods consumed in this region. Most of the data are either borrowed / copied from FCDBs collected in other parts of the world, calculated, imputed, or presumed data. Even with these alternative types, FCDBs in Sub-Saharan Africa have not reached a satisfactory level of completeness, as they still have missing data for …

Scalable And Explainable Self-Supervised Motif Discovery In Temporal Data, Somayeh Bakhtiari Ramezani

Theses and Dissertations

The availability of a scalable and explainable rule extraction technique via motif discovery is crucial for identifying the health states of a system. Such a technique can enable the creation of a repository of normal and abnormal states of the system and identify the system’s state as we receive data. In complex systems such as ECG, each activity session can consist of a long sequence of motifs that form different global structures. As a result, applying machine learning algorithms without first identifying the local patterns is not feasible and would result in low performance. Thus, extracting unique local motifs and …

Study Of Augmentations On Historical Manuscripts Using Trocr, Erez Meoded

Theses and Dissertations

Historical manuscripts are an essential source of original content. For many reasons, it is hard to recognize these manuscripts as text. This thesis used a state-of-the-art Handwritten Text Recognizer, TrOCR, to recognize a 16^th-century manuscript. TrOCR uses a vision transformer to encode the input images and a language transformer to decode them back to text. We showed that carefully preprocessed images and designed augmentations can improve the performance of TrOCR. We suggest an ensemble of augmented models to achieve an even better performance.

An Investigation Into Applications Of Canonical Polyadic Decomposition & Ensemble Learning In Forecasting Thermal Data Streams In Direct Laser Deposition Processes, Jonathan Storey

Theses and Dissertations

Additive manufacturing (AM) is a process of creating objects from 3D model data by adding layers of material. AM technologies present several advantages compared to traditional manufacturing technologies, such as producing less material waste and being capable of producing parts with greater geometric complexity. However, deficiencies in the printing process due to high process uncertainty can affect the microstructural properties of a fabricated part leading to defects. In metal AM, previous studies have linked defects in parts with melt pool temperature fluctuations, with the size of the melt pool and the scan pattern being key factors associated with part defects. …

Machine Learning And Causality For Interpretable And Automated Decision Making, Maria Lentini

Theses and Dissertations

This abstract explores two key areas in decision science: automated and interpretable decision making. In the first part, we address challenges related to sparse user interaction data and high item turnover rates in recommender systems. We introduce a novel algorithm called Multi-View Interactive Collaborative Filtering (MV-ICTR) that integrates user-item ratings and contextual information, improving performance, particularly for cold-start scenarios. In the second part, we focus on Student Prescription Trees (SPTs), which are interpretable decision trees. These trees use a black box "teacher" model to predict counterfactuals based on observed covariates. We experiment with a Bayesian hierarchical binomial regression model as …

Ide-Based Learning Analytics For Assessing Introductory Programming Skill, Phyllis J. Beck

Theses and Dissertations

Providing a sufficient level of personalized feedback on students' current level of strategic knowledge within the context of the natural programming environment through IDE-based learning analytics would transform learning outcomes for introductory programming students. However, providing sufficient insight into the programming process was previously inaccessible due to the need for more complex and scalable data collection methods and metrics with a wider variety for understanding programming metacognition and the full programming process.

This research developed a custom-built web-based IDE and event compression system to investigate two of the five components of a five-dimensional model of cognition for programming skill estimation …

Integrating Data Science With Reliability Engineering: A Study Of Crucial Knowledge And Skills, Altricia Jordan

Theses and Dissertations

Data science, as a discipline can be used in any area. However, in order to utilize data science techniques, data scientist must be taught domain knowledge, referred to as a partner discipline, in the area with which the techniques are to be utilized. Using a quantitative analysis of publicly available information and survey methodology, this research developed recommendations for training data scientist in the partner discipline of reliability engineering. A text analysis was performed on reliability engineering job requisitions resulting in a set of knowledge and skills topics that organizations desire in reliability engineering hires. Using the results of the …

Pruning Ghsom To Create An Explainable Intrusion Detection System, Thomas Michael Kirby

Theses and Dissertations

Intrusion Detection Systems (IDS) that provide high detection rates but are black boxes lead
to models that make predictions a security analyst cannot understand. Self-Organizing Maps
(SOMs) have been used to predict intrusion to a network, while also explaining predictions through
visualization and identifying significant features. However, they have not been able to compete with
the detection rates of black box models. Growing Hierarchical Self-Organizing Maps (GHSOMs)
have been used to obtain high detection rates on the NSL-KDD and CIC-IDS-2017 network traffic
datasets, but they neglect creating explanations or visualizations, which results in another black
box model.
This paper offers …

An Analysis Of Text-Based Machine Learning Models For Vulnerability Detection, Kollin Ryne Napier

Theses and Dissertations

With an increase in complexity of software, developers rely more on reuse and dependencies in their source code via code snippets. As a result, it is becoming harder to identify and mitigate vulnerabilities. Although traditional analysis tools are still utilized, machine learning models are being adopted to expand efforts and combat such threats. Given the possibilities towards usage of such models, research in this area has introduced various approaches which vary in usability and prediction. In generalizing models to a more natural language approach, researchers have opted to train models on source code to identify existing and potential vulnerabilities. Exploratory …

Uncertainty Quantification In Federated Learning For Persistent Post-Traumatic Headache, Byungmoo Brian Kim

Theses and Dissertations

A post-traumatic headache (PTH), resulting from a mild traumatic brain injury (mTBI), potentially develops into persistent post-traumatic headache (PPTH). Although no known cure for PPTH exists, research has shown that receiving treatment at earlier stages of PTH lowers the risk of patients developing PPTH. Previous studies have shown machine learning (ML) models capable of predicting a patient’s PTH progression, but none have considered the issue of protecting patient privacy. Due to patient privacy, ML models only have access to data within the institution. Federated learning (FL) harnesses data from separate institutions without sacrificing patient privacy as institutions can run ML …

Named Entity Recognition From Biomedical Text, Maged Guirguis

Theses and Dissertations

As vast amounts of unstructured data are becoming available digitally, computer-based methods to extract relevant and meaningful information are needed. Named entity recognition (NER) is the task of identifying text spans that mention named entities, and to classify them into predefined categories. Despite the existence of numerous and well-versed NER methods, the bio-medical domain remains under-studied. The objective of this research is to identify an efficient technique for NER tasks from biomedical data. This is achieved by investigating using deep learning technologies namely pre-trained BERT [1] model and its variances SciBERT [2] and BioBERT [3]. Preprocessing the data before passing …

Skeleton-Based Hand Gesture Recognition Using Data-Level Fusion, Oluwaleke Yusuf

Theses and Dissertations

Hand Gesture Recognition (HGR) is a form of perceptual computing that allows artificial systems to capture and interpret human gestures. HGR has applications in human-machine interaction, virtual reality, augmented reality, and human behavior analysis. The human hand can assume a near-infinite number of poses and orientations to form myriad gestures, thus increasing the difficulty of the HGR task.

The hand skeleton of connected joints effectively describes the hand’s geometric shape and thus contains richer semantic gesture information while eliminating noise from individual differences in physical hand characteristics. The efficacy and computational efficiency of skeleton-based HGR frameworks can be significantly enhanced …

Face Anti-Spoofing And Deep Learning Based Unsupervised Image Recognition Systems, Enoch Solomon

Theses and Dissertations

One of the main problems of a supervised deep learning approach is that it requires large amounts of labeled training data, which are not always easily available. This PhD dissertation addresses the above-mentioned problem by using a novel unsupervised deep learning face verification system called UFace, that does not require labeled training data as it automatically, in an unsupervised way, generates training data from even a relatively small size of data. The method starts by selecting, in unsupervised way, k-most similar and k-most dissimilar images for a given face image. Moreover, this PhD dissertation proposes a new loss function to …

Modern Approaches And Theoretical Extensions To The Multivariate Kolmogorov Smirnov Test, Gonzalo Hernando

Theses and Dissertations

Most statistical tests are fully developed for univariate data, but when inference is required for multivariate data, univariate tests risk information loss and interpretability. This research 1) derives and extends the multivariate Komolgorov Smirnov test for 2 and into m-dimensions, 2) derives small sample critical values for the KS test that are not reliant on sample size simulations or correlation between variables, 3) extends large sample estimations and current KS implementations, and 4) provides sample size and power calculations in order to enable experimental design with respect to testing for differences in distributions. Through extensive simulation, we demonstrate that our …