Data Science Commons^™

Interpretable Machine Learning For Self-Service High-Risk Decision Making, Charles Recaido

All Master's Theses

This research contributes to interpretable machine learning via visual knowledge discovery in General Line Coordinates (GLC). The concepts of hyperblocks as interpretable dataset units and GLC are combined to create a visual self-service machine learning model. Two variants of GLC known as Dynamic Scaffold Coordinates (DSC) are proposed. DSC1 and DSC2 can map in a lossless manner multiple dataset attributes to a single two-dimensional (X, Y) Cartesian plane using a dynamic scaffolding graph construction algorithm.

Hyperblock analysis is used to determine visually appealing dataset attribute orders and to reduce line occlusion. It is shown that hyperblocks can generalize decision tree …

A Monte Carlo Simulation Of Rat Choice Behavior With Interdependent Outcomes, Michelle A. Frankot

Graduate Theses, Dissertations, and Problem Reports (ETD)

Preclinical behavioral neuroscience often uses choice paradigms to capture psychiatric symptoms. In particular, the subfield of operant research produces nested datasets with many discrete choices in a session. The standard analytic practice is to aggregate choice into a continuous variable and analyze using ANOVA or linear regression. However, choice data often have multiple interdependent outcomes of interest, violating an assumption of general linear models. The aim of the current study was to quantify the accuracy of linear mixed-effects regression (LMER) for analyzing data from a 4-choice operant task called the Rodent Gambling Task (RGT), which measures decision-making in the context …

A Workflow For Unconventional Reservoirs Optimization Using Supervised Machine Learning In Conjunction With Orthorhombic Elasticity Modeling, Aymen Ab Ali Alhemdi

Graduate Theses, Dissertations, and Problem Reports (ETD)

Due to the anisotropy and heterogeneous nature of unconventional reservoirs like shale, a comprehensive parametric study to optimize hydraulic fracture treatment for such reservoirs is a tough challenge, especially when natural fractures are present. Most of the current frac simulators do not consider the anisotropy of rock elasticity in the shales. Besides, using the fracture simulation linked with reservoir simulation for the parametric study to understand the impact of multiple different design parameters on fracture propagation and production is time expensive and low efficient. The study proposes a workflow including a new orthorhombic (OB) rock algorithm to interpret geomechanical properties …

A Nomogram For Predicting Upper Urinary Tract Damage Risk In Children With Neurogenic Bladder, Qi Li, Miao Cai, Qingsong Pu, Shengde Wu, Xing Liu, Tao Lin, Dawei He, Jianguo Wen, Guanghui Wei

Faculty, Staff and Student Publications

PURPOSE: To establish a predictive model for upper urinary tract damage (UUTD) in children with neurogenic bladder (NB) and verify its efficacy.

METHODS: A retrospective study was conducted that consisted of a training cohort with 167 NB patients and a validation cohort with 100 NB children. The clinical data of the two groups were compared first, and then univariate and multivariate logistic regression analyses were performed on the training cohort to identify predictors and develop the nomogram. The accuracy and clinical usefulness of the nomogram were verified by receiver operating characteristic (ROC) curve, calibration curve and decision curve analyses.

RESULTS: …

Bayesian Spatiotemporal Modeling With Gaussian Processes, Qing He

Electronic Theses and Dissertations, 2020-2023

Bayesian spatiotemporal models have been successfully applied to various fields of science, such as ecology and epidemiology. The complicated nature of spatiotemporal patterns can be well represented through priors such as Gaussian processes. This dissertation is focused on two applications of Bayesian spatiotemporal models: a) anomaly detection for spatiotemporal data with missingness and b) zero-inflated spatiotemporal count data analysis. Missingness in spatiotemporal data prohibits anomaly detection algorithms from learning characteristic rules and patterns due to the lack of most data. This project is motivated by a challenge provided by the National Science Foundation (NSF) and the National Geospatial-Intelligence Agency (NGA). …

Phenotype-Genotype Analysis Of Caucasian Patients With High Risk Of Osteoarthritis, Yanfei Wang, Jacqueline Chyr, Pora Kim, Weiling Zhao, Xiaobo Zhou

Faculty, Staff and Student Publications

Background: Osteoarthritis (OA) is a common cause of disability and pain around the world. Epidemiologic studies of family history have revealed evidence of genetic influence on OA. Although many efforts have been devoted to exploring genetic biomarkers, the mechanism behind this complex disease remains unclear. The identified genetic risk variants only explain a small proportion of the disease phenotype. Traditional genome-wide association study (GWAS) focuses on radiographic evidence of OA and excludes sex chromosome information in the analysis. However, gender differences in OA are multifactorial, with a higher frequency in women, indicating that the chromosome X plays an essential role …

Re-Modeling The Interior: Spatial Methods And Policy Revisions To Improve Inventory And Designation Of Blm’S Areas Of Critical Environmental Concern, Amy H. Katz

Graduate Student Theses, Dissertations, & Professional Papers

The Bureau of Land Management (BLM) manages a vast amount of public land in the western United States, most of which they currently manage for multiple uses. Specific conservation and management of these lands could mitigate climate change impacts and contribute to the global initiative to conserve 30 percent of lands and waters by 2030. Particularly, the agency can achieve this through more effective administration of Areas of Critical Environmental Concern (ACEC), a designation that is prioritized under the Federal Land Policy and Management Act (FLPMA). To do so requires updated regulations that set clear parameters around inventory and designation, …

Trading Financial Instruments Like A Video Game: Searching For Profit Using Deep Reinforcement Learning., Sebastian Coombs

Graduate Student Theses, Dissertations, & Professional Papers

Buying and selling Stocks, Foreign Currencies (FOREX), Commodities, and Cryptocurrencies have been a source of wealth generation, and more often, wealth loss for many brave enough to enter the financial markets. In this paper, the author builds on the work of Williams, J. 2022 and develops an agent-based method to solve this wealth generation problem with the use of neural networks. The author points out some assumptions made by Williams, J. 2022 that were sound in theory, but made the implementation of the algorithm presented in their paper diverge from the theory. The author proposes a fundamentally different algorithmic method, …

Fashion Compatibility Prediction Using Ensemble Learning, Nathan Utzman

Graduate Theses, Dissertations, and Problem Reports (ETD)

Fashion is important both financially and for self-expression. There are many tasks in the fashion domain which can be addressed with artificial intelligence. The task of fashion compatibility prediction is to determine how well a set of items work together to form an outfit. Two main tasks are typically used to evaluate the performance of a fashion compatibility prediction model – Outfit Compatibility Prediction and Fill in the Blank.

In this work, a compatibility prediction model, which is based on the graph autoencoder, is evaluated. This same model is then used in a homogeneous ensemble learning approach, proposed to improve …

Hydrocarbon Pay Zone Prediction Using Ai Neural Network Modeling., Darren D. Guedon

Graduate Theses, Dissertations, and Problem Reports (ETD)

This paper captures the ability of AI neural network technology to analyze petrophysical datasets for pattern recognition and accurate prediction of the pay zone of a vertical well from the Santa Fe field in Kansas.

During this project, data from 10 completed wells in the Santa Fe field were gathered, resulting in a dataset with 25,580 records, ten predictors (logs data), and a single binary output (Yes or No) to identify the availability of Hydrocarbon over a half feet depth segment in the well. Several models composed of different predictors combinations were also tested to determine how impactful some logs …

Classifying Blood Glucose Levels Through Noninvasive Features, Rishi Reddy

Graduate Theses, Dissertations, and Problem Reports (ETD)

Blood glucose monitoring is a key process in the prevention and management of certain chronic diseases, such as diabetes. Currently, glucose monitoring for those interested in their blood glucose levels are confronted with options that are primarily invasive and relatively costly. A growing topic of note is the development of non-invasive monitoring methods for blood glucose. This development holds a significant promise for improvement to the quality of life of a significant portion of the population and is overall met with great enthusiasm from the scientific community as well as commercial interest. This work aims to develop a potential pipeline …

The Application Of Deep Learning And Cloud Technologies To Data Science, Ian A. Trawinski

College of Graduate Studies: Theses & Dissertations

Machine Learning and Cloud Computing have become a staple to businesses and educational institutions over the recent years. The two forefronts of big data solutions have garnered technology giants to race for the superior implementation of both Machine Learning and Cloud Computing. The objective of this thesis is to test and utilize AWS SageMaker in three different applications: time-series forecasting with sentiment analysis, automated Machine Learning (AutoML), and finally anomaly detection. The first study covered is a sentiment-based LSTM for stock price prediction. The LSTM was created with two methods, the first being SQL Server Data Tools, and the second …

Exo-Sir: An Epidemiological Model To Analyze The Impact Of Exogenous Spread Of Infection, Nirmal Kumar Sivaraman, Manas Gaur, Shivansh Baijal, Sakthi Balan Muthiah, Amit Sheth

Publications

Epidemics like Covid-19 and Ebola have impacted people's lives significantly. The impact of mobility of people across the countries or states in the spread of epidemics has been significant. The spread of disease due to factors local to the population under consideration is termed the endogenous spread. The spread due to external factors like migration, mobility, etc. is called the exogenous spread. In this paper, we introduce the Exo-SIR model, an extension of the popular SIR model and a few variants of the model. The novelty in our model is that it captures both the exogenous and endogenous spread of …

Exploring Factors That Predict Marketing Of E-Cigarette Products On Twitter: Infodemiology Approach Using Time Series, Nnamdi C. Ezike, Allison Ames Boykin, Page D. Dobbs, Huy Mai, Brian A. Primack

Health, Human Performance and Recreation Faculty Publications and Presentations

Background:

Electronic nicotine delivery systems (known as electronic cigarettes or e-cigarettes) increase risk for adverse health outcomes among naïve tobacco users, particularly youth and young adults. This vulnerable population is also at risk for exposed brand marketing and advertisement of e-cigarettes on social media. Understanding predictors of how e-cigarette manufacturers conduct social media advertising and marketing could benefit public health approaches to addressing e-cigarette use.

Objective:

This study documents factors that predict changes in daily frequency of commercial tweets about e-cigarettes using time series modeling techniques.

Methods:

We analyzed data on the daily frequency of commercial tweets about e-cigarettes collected …

Online Deep Learning From Doubly-Streaming Data, Heng Lian, John S. Atwood, Bo-Jian Hou, Jian Wu, Yi He

Computer Science Faculty Publications

This paper investigates a new online learning problem with doubly-streaming data, where the data streams are described by feature spaces that constantly evolve, with new features emerging and old features fading away. A plausible idea to deal with such data streams is to establish a relationship between the old and new feature spaces, so that an online learner can leverage the knowledge learned from the old features to better the learning performance on the new features. Unfortunately, this idea does not scale up to high-dimensional multimedia data with complex feature interplay, which suffers a tradeoff between onlineness, which biases shallow …

Using Skeleton Correction To Improve Flash Lidar-Based Gait Recognition, Nasrin Sadeghzadehyazdi, Tamal Batabyal, Alexander Glandon, Nibir Dhar, Babajide Familoni, Khan Iftekharuddin, Scott T. Acton

Electrical & Computer Engineering Faculty Publications

This paper presents GlidarPoly, an efficacious pipeline of 3D gait recognition for flash lidar data based on pose estimation and robust correction of erroneous and missing joint measurements. A flash lidar can provide new opportunities for gait recognition through a fast acquisition of depth and intensity data over an extended range of distance. However, the flash lidar data are plagued by artifacts, outliers, noise, and sometimes missing measurements, which negatively affects the performance of existing analytics solutions. We present a filtering mechanism that corrects noisy and missing skeleton joint measurements to improve gait recognition. Furthermore, robust statistics are integrated with …

Facial Landmark Feature Fusion In Transfer Learning Of Child Facial Expressions, Megan A. Witherow, Manar D. Samad, Norou Diawara, Khan M. Iftekharuddin

Electrical & Computer Engineering Faculty Publications

Automatic classification of child facial expressions is challenging due to the scarcity of image samples with annotations. Transfer learning of deep convolutional neural networks (CNNs), pretrained on adult facial expressions, can be effectively finetuned for child facial expression classification using limited facial images of children. Recent work inspired by facial age estimation and age-invariant face recognition proposes a fusion of facial landmark features with deep representation learning to augment facial expression classification performance. We hypothesize that deep transfer learning of child facial expressions may also benefit from fusing facial landmark features. Our proposed model architecture integrates two input branches: a …

Particle Identification And Tracking In Real Time Using Machine Learning On Fpga, F. Barbosa, L. Belfore, C. Dickover, C. Fanelli, S. Furletov, Y. Furletova, L. Jokhovets, D. Lawrence, D. Romanov

Electrical & Computer Engineering Faculty Publications

This project is a multi-disciplinary endeavour between Physics, Electrical Engineering, and Computer Engineering. The purpose is to develop and implement an FPGA(*) based Machine Learning algorithm for real-time particle identification, filtering, and data reduction. This is important research that can be applied to streaming readout systems being developed now at JLab and other facilities. Real-time data processing is a frontier field in experimental physics, especially in HEP. The application of FPGAs at the trigger level is used by many current and planned experiments (CMS, LHCb, Belle2, PANDA). Usually they use conventional processing algorithms. LHCb has implemented ML elements for real-time …

Temporal Disambiguation Of Relative Temporal Expressions In Clinical Texts Using Temporally Fine-Tuned Contextual Word Embeddings., Amy L. Olex

Theses and Dissertations

Temporal reasoning is the ability to extract and assimilate temporal information to reconstruct a series of events such that they can be reasoned over to answer questions involving time. Temporal reasoning in the clinical domain is challenging due to specialized medical terms and nomenclature, shorthand notation, fragmented text, a variety of writing styles used by different medical units, redundancy of information that has to be reconciled, and an increased number of temporal references as compared to general domain texts. Work in the area of clinical temporal reasoning has progressed, but the current state-of-the-art still has a ways to go before …

Estimating Weighted Panel Sizes For Primary Care Providers: An Assessment Of Clustering And Novel Methods Of Panel Size Estimation On Electronic Medical Records, Martin A. Lavallee

Theses and Dissertations

Primary Care is on the frontlines of healthcare, thus they see the most diverse set of patients. In order to achieve high functioning primary care, a practice must establish empanelment, the pairing of patients to providers. Enumeration of empanelment, or estimating panel sizes, helps ensure that the demands of the patients demand the supply of providers and optimize the balance of primary care resources to improve quality of care. Further we can adjust panel sizes by using patient-level data on healthcare utilization and complexity extracted from the electronic medial record to determine the amount of care or burden of work …