PHD, Kent State University, 2021, College of Arts and Sciences / Department of Geography

Cancer epidemiology has a long history of applying geographic thinking to address long-standing place-based disparities. This dissertation adds new knowledge to geospatial approaches to social determinants of cancer outcomes. It establishes a framework consisting of three dimensions in evaluating, identifying, and prioritizing spatially heterogeneous risk factors of cancer outcomes. The first dimension is protection. Using a space-time statistic, the first study evaluated whether a non-spatial healthcare policy, Medicaid expansion, has offered protection targeting spatially vulnerable populations against adverse cancer outcomes such as breast cancer late-stage diagnosis. The second dimension is phenotype. Using a classification and regression tree, the study disentangled how risk factors of late-stage breast cancer diagnosis were conceptualized and capsulized as phenotypes that labeled groups of homogenous geographic areas. It provides a novel angle to uncover cancer disparities and to provide insights for cancer surveillance, prevention, and control. The third dimension is priority. Using a geographic random forest along with several validation methods, the study emphasized the importance of the competing effect among risk factors of cancer mortality that are specific to geographic areas. The findings from this study can be used directly for priority settings in addressing the most urgent issues associated with cancer mortality. This dissertation demonstrated that geographic methodologies and frameworks are useful and are imperative to cancer epidemiology.

Committee: Jay Lee (Committee Chair); Jun Li (Committee Member); James Tyner (Committee Member); Xinyue Ye (Committee Member) Subjects: Epidemiology; Geographic Information Science; Geography; Health; Health Care; Health Care Management; Oncology; Public Health; Public Policy; Statistics

Master of Science (MS), Wright State University, 2024, Computer Science

In today's technological landscape, hardware devices are integral to critical applications such as industrial automation, autonomous vehicles, and medical equipment, relying on advanced platforms like FPGAs for core functionalities. However, the multi-stage manufacturing process, often distributed across various foundries, introduces substantial security risks, notably the potential for hardware Trojan insertion. These malicious modifications compromise the reliability and safety of hardware systems. This research addresses the detection of hardware Trojans through side-channel analysis, utilizing power and electromagnetic signal data, combined with meta-learning techniques, specifically model stacking. By employing diverse base models and a meta-model to consolidate predictions, this non-invasive approach effectively identifies Trojans without requiring direct access to internal circuitry. The methodology demonstrates robust classification capabilities, achieving an accuracy of 88.0\%, precision of 81.0\%, and recall of 95.0\%, even on previously unseen data. The results highlight the superior performance of meta-learning over traditional detection methods, offering an efficient and reliable solution to enhance hardware security.

Committee: Fathi Amsaad Ph.D. (Advisor); Junjie Zhang Ph.D. (Committee Member); Huaining Cheng Ph.D. (Committee Member); Nitin Pundir Ph.D. (Committee Member); Thomas Wischgoll Ph.D. (Other); Subhashini Ganapathy Ph.D. (Other) Subjects: Computer Engineering; Computer Science; Electrical Engineering

MS, Kent State University, 2024, College of Arts and Sciences / Department of Geography

Wildfire and drought are key drivers of shrubland expansion in southwestern US landscapes. Stand-replacing fires in dry conifer forests induce shrub-dominated stages, and changing climatic patterns may cause a long-term shift from coniferous forests to deciduous shrublands. This study assessed recent changes in deciduous fractional shrub cover (DFSC) in the eastern Jemez Mountains from 2019-2023 using topographic and Sentinel-2 satellite data in a random forest model. Sentinel-2 provides multispectral bands at 10 and 20 meters, including three 20 meter red edge bands, which are highly sensitive to variation in vegetation. There is no consensus in the literature on whether upscaling imagery to 20 meters or downscaling to 10 meters is more advantageous. Therefore, an additional goal of this study was to evaluate the impact of spatial scale on DFSC model performance. Two random forest models were built, a 10 and 20 meter model. The 20 meter model outperformed the 10 meter model, achieving an R-squared value of 0.82 and an RMSE of 7.85, compared to the 10 meter model (0.76 and 9.99, respectively). The 20 meter model, built from 2020 satellite imagery, was projected to the other years of the study, by replacing the spectral variables with satellite imagery from the respective year, resulting in yearly predictions of DFSC from 2019-2023. DFSC decreased from 2019-2022, coinciding with severe drought and a 2022 fire, followed by a significant increase in 2023, particularly within the 2022 fire footprint. Overall trends showed a general increase in DFSC despite high interannual variability, with elevation being a key topographic variable influencing these trends. This study revealed yearly vegetation dynamics in a semi-arid system and provided a close look at post-fire regeneration patterns in deciduous resprouting shrubs. Understanding this complex system is crucial for informing management strategies as the landscape continues to shift from conifer forest to shrubland du (open full item for complete abstract)

Committee: Scott Sheridan (Advisor); Christie Bahlai (Committee Member); Timothy Assal (Advisor) Subjects: Ecology; Geography

Master of Science, University of Akron, 2024, Computer Science

Within this study, Random Forest regression models were employed to predict the viscosity of ionic liquids (ILs) using an expanded dataset generated from the cheminformatics software RDKit. The initial dataset comprised over 22,000 experimental observations for 2,068 different ILs. This dataset was meticulously filtered to remove data points with high uncertainty or inconsistencies, ensuring the reliability of the training data. Feature extraction was performed using RDKit, significantly expanding the dataset from 4 initial features to 126 features, including various molecular descriptors such as molecular weights, charges, and structural characteristics. To mitigate the high dimensionality and improve model performance, Principal Component Analysis (PCA) was used to reduce the feature space while retaining 95% of the variance. Feature selection techniques, including SelectKBest, Recursive Feature Elimination (RFE), and a novel Bayesian Feature Selection (BFS), were utilized to refine the feature set further by identifying and removing redundant and less informative features. Hyperparameter tuning via Bayesian optimization was also performed, which systematically explored the hyperparameter space to identify the optimal settings for the Random Forest model. This is then followed with a rigorous cross validation process, involving 10-fold cross-validation, which confirm the model's generalizability and robustness. The model development involves autonomous creation and testing of numerous models with varying features and hyperparameter subsets. Machine learning methods were applied to efficiently predict viscosities and establish accurate structure-property relationships. The resulting predictions showed an average R² value of 0.96 for the training set and 0.76 for the test set, indicating robust predictive performance for viscosity. Overall, this study demonstrates the effectiveness of combining advanced data preprocessing, feature engineering, and machine learning (open full item for complete abstract)

Committee: En Cheng (Advisor); Fardin Khabaz (Committee Member); Zhong-Hui Duan (Committee Member) Subjects: Computer Science; Materials Science

Master of Science in Computer Engineering (MSCE), Wright State University, 2024, Computer Engineering

Hardware components are becoming prone to threats with increasing technological advances. Malicious modifications to such components are increasing and are known as hardware Trojans. Traditional approaches rely on functional assessments and are not sufficient to detect such malicious actions of Trojans. Machine learning (ML) assisted techniques play a vital role in the overall detection and improvement of Trojan. Our novel approach using various ML models brings an improvement in hardware Trojan identification with power signal side channel analysis. This study brings a paradigm shift in the improvement of Trojan detection in integrated circuits (ICs). In addition to this, our further analysis towards hardware authentication extends towards PUFs (Physical Unclonable Functions). Arbiter PUFs were chosen for this purpose. These are also Vulnerable towards ML attacks. Advanced ML assisted techniques predict the responses and perform attacks which leads to the integrity of PUFs. Our study helps improve ML-assisted hardware authentication for ML attacks. In addition, our study also focused on the defense part with the addition of noise and applying the same attack ML-assisted model. Detection of Trojan in hardware components is achieved by implementing machine learning techniques. For this Purpose, several Machine learning models were chosen. Among them, Random Forest classifier (RFC) and Deep neural network shows the highest accuracy. This analysis plays a vital role in the security aspect of all hardware components and sets a benchmark for the overall security aspects of hardware. Feature extraction process plays major role for the improvement of accuracy and reliability of hardware Trojan classification. Overall, this study brings significant improvement in the field of overall hardware security. Our study shows that RFC performs best in hardware classification with an average of 98. 33% precision of all chips, and deep learning techniques give 93. 16% prec (open full item for complete abstract)

Committee: Fathi Amsaad Ph.D. (Advisor); Kenneth Hopkinson Ph.D. (Committee Member); Wen Zhang Ph.D. (Committee Member) Subjects: Computer Engineering; Computer Science; Engineering; Information Technology; Technical Communication; Technology

Master of Science (MS), Wright State University, 2024, Computer Science

Semiconductor microelectronics integrated circuits (ICs) are increasingly integrated into modern life-critical applications, from intelligent infrastructure and consumer electronics to the Internet of Things (IoT) and advanced military and medical systems. Unfortunately, these applications are vulnerable to new hardware security attacks, including microelectronics counterfeits and hardware modification attacks. Physical Unclonable Functions (PUFs) are state-of-the-art hardware security solutions that utilize process variations of integrated circuits for device authentication, secret key generation, and microelectronics counterfeit detection. The negative impact of aging on Static Random \linebreak Access Memory Physical Unclonable Functions (SRAM PUFs) has significant consequences for microelectronics authentication, security, and reliability. This research thoroughly \linebreak examines the effect of aging on the reliability of SRAM PUFs used for secure and trusted microelectronics integrated circuit applications. It initially provides an overview of SRAM PUFs, highlighting their significance and essential features while addressing encountered challenges. The study then covers mitigation techniques, including multi-modal PUFs, that already exist to boost the resilience of SRAM PUFs against aging impacts, highlighting their advantages and the gap in the research addressed in this research. This work proposes a novel AI-enabled security for reliable SRAM PUFs. The proposed approach aims to study and countermeasure the impact of aging on SRAM PUF by analyzing data samples, including Bias Temperature Instability (BTI), Bit Flips, Accelerated aging, and Hot Carrier Injection (HCI) and to study their effects on SRAM PUF cell properties and output. Accelerated aging is a direct result of a change in the environmental temperature and voltage for a few hours. We aim to mitigate the impact of accelerated aging on the reliability authentication and encryption keys of (open full item for complete abstract)

Committee: Fathi Amsaad Ph.D. (Advisor); Wen Zhang Ph.D. (Committee Member); John Emmert Ph.D. (Committee Member) Subjects: Computer Engineering; Computer Science; Electrical Engineering; Information Technology; Technology

PhD, University of Cincinnati, 2023, Engineering and Applied Science: Mechanical Engineering

The aim of this dissertation is to present the development of new approaches to air traffic management for vehicles operating within the national airspace. This dissertation examines air traffic management techniques for both manned commercial aircraft along with small unmanned aerial systems (sUAS). This dissertation starts with the development of a supervised learning algorithm to generate estimated time of arrival for commercial aircraft and covers a unique conflict prediction algorithm for sUAS which is based on near future path predictions. Both contributions assist in air traffic management by minimizing required human intervention. The dissertation culminates with the development of a novel predictive artificial potential field (PAPF) navigation system for sUAS. The PAPF relies upon Interacting Multiple Model (IMM) vehicle predictions and is designed to eliminate the need for human intervention as sUAS operate within the airspace by preemptively rerouting vehicles in real time based on conflicts that will likely happen in the near future. The PAPF system is run completely on-board the sUAS and does not require additional ground based resources. The first element of the dissertation covers the estimated time of arrival (ETA) predictor for commercial aircraft. The ETA predictor focused on aircraft landing at the Dallas Fort Worth airport and used historical flight and meteorological data to train a random forest regressor. The trained algorithm could be implemented to make use of real time data and estimate the remaining flight time once aircraft have entered within a 200 mile radius circle of the airport. The data used is readily available from airport facilities and the aircraft themselves via Automatic Dependent Surveillance-Broadcast (ADS-B). This work was supported by NASA Office of STEM Engagement and used historical flight data for aircraft landing at Dallas Fort Worth airport. The second element discussed in the dissertation consists of a (open full item for complete abstract)

Committee: Manish Kumar Ph.D. (Committee Chair); Michael Alexander-Ramos Ph.D. (Committee Member); David Thompson Ph.D. (Committee Member); Tejas Puranik Ph.D. (Committee Member); Rajnikant Sharma Ph.D. (Committee Member); Zachariah Fuchs Ph.D. (Committee Member) Subjects: Mechanical Engineering

Doctor of Philosophy (Ph.D.), Bowling Green State University, 2023, Media and Communication

This dissertation investigates the effectiveness of counterspeech strategies employed on Twitter in response to anti-Asian hate during the COVID-19 pandemic. This research delves into the communicative strategies, emotional tones, and geospatial distribution of counterspeech, specifically focusing on its effectiveness in the United States. A supervised machine learning was employed to classify counterspeech tweets and counterspeech strategies based on empirical typology. By analyzing 106,388 tweets associated with the hashtag #StopAsianHate collected from November 2021 to May 2022, this research provides insights into the varied effectiveness of counterspeech strategies. The analysis revealed that though counterspeakers were using more negative tones in counterspeech tweets, the tweets with visual media and positive emotional tone received more engagement on Twitter through retweets and favorites compared to those with a negative or neutral tone. This study also breaks new ground by recognizing that higher level of racial diversity does not facilitate higher level of counterspeech against hate speech and hate crime. Additionally, this study highlights the varying degrees of participation in counterspeech across different ethnic groups within Asian American community and underscores the importance of tailored strategies in addressing hate speech. Recognizing this distinction proved essential in crafting evidence-based guidance for community and individual interventions while fostering support from allies of diverse racial backgrounds.

Committee: Louisa Ha Ph.D. (Committee Chair); Lisa Hanasono Ph.D. (Committee Member); Yanqin Lu Ph.D. (Committee Member); William Sawaya Ph.D. (Other) Subjects: Artificial Intelligence; Asian American Studies; Communication; Mass Media; South Asian Studies

Master of Science, The Ohio State University, 2023, Environmental Science

Biophysical modeling has experienced significant advancements over time, transitioning from physics-based approaches to regression modeling and now embracing machine learning techniques alongside vast quantities of environmental data. However, a growing concern in data-driven models is the lack of physical consistency and transparency, leading to the emergence of a field known as hybrid machine learning. In this abstract, we present two studies that leverage the availability of cost-effective in-situ remote sensing systems to address data limitations in surface energy modeling. The first study focuses on developing simple machine learning models to improve model transparency and physical consistency by accurately estimating soil heat flux—a vital component of the energy balance. To achieve this, we utilized two variables, namely radiometric surface temperature (LST) and the normalized difference vegetation index (NDVI), which could capture upwards of 85% of the variability in soil heat flux at half-hourly resolution across four agricultural systems. The performance of these models surpassed that of existing semi-empirical models, while significantly reducing the costs associated with measuring soil heat flux. In the second study, we demonstrate the application of simple machine learning for parameter estimation, serving as a precursor to physical modeling. Specifically, we validated the Surface Temperature Initiated Closure (STIC) model simultaneously over four crop systems in Ames, Iowa. Through this implementation, we discovered that simple machine learning estimation can replace model inputs without introducing additional biases. Furthermore, leveraging the parametric estimations of energy conductance provided by STIC, the study identified variations in stomatal response and drought resilience among the four crop systems. Integrating machine learning and remote sensing techniques shows promise in improving our understanding and predictive capabilities in surface (open full item for complete abstract)

Committee: Darren Drewry (Advisor); John Fulton (Committee Member); Christopher Stewart (Committee Member) Subjects: Agriculture; Biophysics; Environmental Science; Remote Sensing

Doctor of Philosophy (Ph.D.), Bowling Green State University, 2023, Statistics

This dissertation explores the relationship between the local context of a nucleotide in an RNA 3D structure and the extent to which the nucleotide base (A, C, G, U) is conserved across different species. Two datasets are studied, a small dataset from E. coli where the local context is described in terms of human-annotated interactions such as base pairs and base stacking, and a large dataset in which the local context is divided into 675 grid cells and the number of atoms in each cell forms a high-dimensional predictor. The response variable in both cases is the proportion of species having the most common base at that location. For both datasets, random forest, neural network, and other models are fit and evaluated. This makes it possible to identify which predictor variables are most informative, and that in turn tells which features of the local context most constrain base conservation, which are interpreted from both a statistical and biological point of view. Poorly predicted nucleotides are labeled and explained.

Committee: Craig L. Zirbel Ph.D. (Committee Chair); Andrew C. Layden Ph.D. (Other); Umar Islambekov Ph.D. (Committee Member); Junfeng Shang Ph.D. (Committee Member) Subjects: Statistics

PhD, University of Cincinnati, 2022, Medicine: Biostatistics (Environmental Health)

Connecting and predicting health outcomes based on environmental exposures is critical for disease prevention. The ensemble approach, weighted quantile sum regression (WQS) is often used for that purpose. In this work, a novel application of WQS to analyzing microbiome data, an original method for determining the WQS selection threshold parameter t , and the validation of this method with real world chemical exposure biomarker data are used. WQS provided useful estimates of overall effect of exposures to microbial mixtures on the observed presence of respiratory health conditions in children. Additionally, the proposed novel method for determining threshold selection parameter t improved characterization of important predictors within the model. Validation of the t selection procedure expanded the generalizability of ROC optimization for identifying groups of important predictors for relative comparisons.

Committee: Roman Jandarov Ph.D. (Committee Member); Tiina Reponen Ph.D. (Committee Member); Marepalli Rao Ph.D. (Committee Member); Ashley Merianos Ph.D. (Committee Member) Subjects: Biostatistics

Master of Science, The Ohio State University, 2022, Environment and Natural Resources

The population of the Central Ohio region is increasing largely due to better economic prospects. The need for housing and related developments will likely go up as the population grows. Most of Ohio's forests are privately owned, and the anticipated developments could impact the current environment by altering the land use of privately owned woodlands. Landowner-level factors impacting changes in land cover and use are largely neglected while predicting these trends. In the first study, private woodland owners were surveyed in multiple counties in Central Ohio on their ownership characteristics, motivations for owning woodlands, demographic factors, and familiarity with ecosystem services account for those factors. The data was analyzed using a binary logistic regression model to identify key elements influencing woodland owners' willingness to sell their property at various price points. The study found that the choice to sell a property was significantly influenced by the landowners' age and residency on the property. Private woodland owners who owned their properties for hunting and amenity values were more likely to sell them. Additionally, landowners aware of the forest's capacity to clean the air expressed less interest in selling their property. On the other hand, landowners who used woodland for recreational activities were less likely to sell. The second study surveyed private woodland owners to determine their preferences for a hypothetical conservation program utilizing binary choice experiments and best-worst choice profiles. Woodland owners were asked to select the best and worst attributes of different programs and their willingness to enroll. Best-Worst scores, Conditional logistic, and Random Effects logistic regression were used to explain woodland owners' priorities. Best-Worst scores show that the highest revenue ($100 acre/year) was the most selected attribute in all choice profiles. A non-profit program structure and no withdrawal penalty are m (open full item for complete abstract)

Committee: Sayeed Mehmood (Advisor); Roger Williams (Committee Member); G. Matthew Davies (Committee Member) Subjects: Environmental Economics; Natural Resource Management

MS, University of Cincinnati, 2022, Medicine: Biostatistics (Environmental Health)

Kidney transplantation is a crucial clinical treatment for patients with end stage renal disease, chronic kidney disease, or kidney failure. In 2021, 41,355 transplants were performed in the U.S, with organs from both deceased and living donors. A record-setting 24,670 kidney transplants were performed in 2021, representing an increase of 8.1% from 2020. Patients have several comorbidities, but the impacts of comorbidities on kidney transplant outcome have not been well studied. On this research note, we have examined the impacts of comorbidities on kidney transplant outcome and developed quantitative biomarkers based on a select number of comorbidities. The 2016 National Inpatient Sample has data on 7,135,090 patients, of which 3,216 patients have been identified to have received kidney transplants following the Diagnosis Related Group (DRG)'s code, '652' without missing observations. For our study, inpatient records were summarized based on demographic characteristics, patient primary conditions, and patient comorbidities. Patient primary conditions and comorbidities are coded as International Classification of Diseases (ICD-10) codes. The primary conditions for each patient are listed in the data column labeled I10_DX1 and the comorbidities for each patient are listed in the data columns labeled I10_DX2, I10_DX3, ..., I10_DX30. All the comorbidities put together include over 2000 conditions; we have created a column for each comorbidity. A method based on the Random Forest algorithm is used to select comorbidity variables using mean decrease accuracy and mean decrease gini with the response variable, i.e., the outcome of the kidney transplant. The top 27 comorbidity variables, including demographic variables, were selected for further analysis. Logistic regression model is fitted to the data consisting of the response variable and the top 27 predictor variables. The response variable encapsulates the result of a kidney transplant. The logit (open full item for complete abstract)

Committee: Roman Jandarov Ph.D. (Committee Member); Marepalli Rao Ph.D. (Committee Member) Subjects: Biostatistics

Master of Computing and Information Systems, Youngstown State University, 2022, Department of Computer Science and Information Systems

Today data is the most important part of any organization, as data is everywhere around us. Most companies produce large amount of data that is essential for the decision making process. In this context, many machine learning and artificial intelligence methods can be used for analysis and prediction. To understand the data quality and make efficient use of the data, several pre-processing steps are necessary. In various fields of study and industry, machine learning is becoming the dominant problem-solving technique. Machine learning models are now being used to solve a variety of real-world problems in a variety of disciplines, ranging from retail and finance to medicine and healthcare which demands high predictive accuracy. Understanding data quality and feature engineering are some of the most critical parts of any machine learning project. Mostly, companies manage tabular data that needs to be converted into numerical data. However, this improved predictive accuracy has often been achieved through increased model complexity which leads to a lack of transparency. The major disadvantage is that the models' inner workings are hidden from the user because it prevents even an experienced professional from interpreting and understanding the reasoning behind the system and how some decisions are made. The quality and quantity of data used to train machine learning algorithms are directly related to their predicted ability. Quality data leads to accurate predictions that in turn leads to accurate explanations. In many cases, it is important to know how predictions are made. The research is focused on the effect of data quality and feature engineering on training different tabular datasets using different machine learning models and the ranking of features in terms of their importance to the prediction. The results are compared in terms of performance accuracy to find which feature set and which model works best.

Committee: Alina Lazar PhD (Advisor); Feng Yu PhD (Committee Member); Yong Zhang PhD (Committee Member) Subjects: Computer Science

Doctor of Philosophy (PhD), Wright State University, 2021, Interdisciplinary Applied Science and Mathematics PhD

Counterfactual reasoning can be used in task-switching scenarios, such as design and planning tasks, to learn from past behavior, predict future performance, and customize interventions leading to enhanced performance. Previous research has focused on external factors and personality traits; there is a lack of research exploring how the decision-making process relates to both task-switching and counterfactual predictions. The purpose of this dissertation is to describe and explain individual differences in task-switching strategy and cognitive processes using machine learning techniques and linear ballistic accumulator (LBA) models, respectively, and apply those results in counterfactual models to predict behavior. Applying machine learning techniques to real-world task-switching data identifies a pattern of individual strategies that predicts out-of-sample clustering better than random assignment and identifies the most important factors contributing to the strategies. Comparing parameter estimates from several different LBA models, on both simulated and real data, indicates that a model based on information foraging theory that assumes all tasks are evaluated simultaneously and holistically best explains task-switching behavior. The resulting parameter values provide evidence that people have a switch-avoidance tendency, as reported in previous research, but also show how this tendency varies by participant. Including parameters that describe individual strategies and cognitive mechanisms in counterfactual prediction models provides little benefit over a baseline intercept-only model to predict a holdout dataset about real-world task switching behavior and performance, which may be due to the complexity and noise in the data. The methods developed in this research provide new opportunities to model and understand cognitive processes for decision-making strategies based on information foraging theory, which has not been considered previously. The results from this (open full item for complete abstract)

Committee: Ion Juvina Ph.D. (Advisor); Joseph W. Houpt Ph.D. (Committee Member); Valerie L. Shalin Ph.D. (Committee Member); Zheng Xu Ph.D. (Committee Member) Subjects: Applied Mathematics; Cognitive Psychology; Mathematics; Psychology

Doctor of Philosophy, University of Toledo, 2021, Spatially Integrated Social Science

This dissertation bridges the gap between spatial econometrics and machine learning under the theoretical banner of spatial data science. Methodologically, it uses the spatial error model, spatial lag model, and the randomForest algorithm in order to predict Human Development Index (HDI) values within Morocco at the commune scale. This prediction task is done using the Moroccan censuses of 2004 and 2014. The results of this process show that randomForest can outperform the traditional spatial econometric models in terms of numeric accuracy within this specific case. Since spatial thinkers are just as concerned with spatial accuracy as they are with numeric accuracy, post-estimation procedures were developed in order to assess the spatial accuracy of the spatial error model, spatial lag model, and randomForest in the Moroccan case. These post-estimation procedures were developed for both the global and local levels. In both cases, it is shown that randomForest outperforms both of the spatial econometric models in terms of spatial accuracy within the Moroccan case. With the Morocco specific results complete, the dissertation moves to simulated data experiments in order to assess different properties of randomForest vs. the spatial lag model, and randomForest vs. the spatial error model. The simulation experiments are carried out using five different data generation processes. Throughout the experiments bias, consistency, efficiency, and spatial prediction performance are evaluated and compared. These experiments show that when either the spatial lag model or spatial error model are the correct model specification, randomForest is unable to outperform either of them in terms of bias, consistency, efficiency, or spatial prediction performance. Therefore, it is concluded that if randomForest does outperform the traditional spatial econometric models, as happened in the Moroccan case, neither the spatial lag model nor the spatial error model are the correct m (open full item for complete abstract)

Committee: Oleg Smirnov Dr. (Advisor); Neil Reid Dr. (Committee Member); Sujata Shetty Dr. (Committee Member); David Nemeth Dr. (Committee Member); Jack Kalpakian Dr. (Committee Member) Subjects: Geographic Information Science; Geography

Master of Arts, Miami University, 2021, Economics

Political turnover is the change in a government from one ruling political party to another. Turnover often leads to the redaction of policies from the previous administration, creating periods of political as well as economic uncertainty. While turnover can impact an economy, the economy can also impact an electorate's perspective on the ruling political body. This study looks to answer the question: can we predict turnover using economic indicators such as unemployment rate? Using linear probability models, logistic regression, and classification tree models, we examine relationships between factors such as unemployment, inflation, and investment on political outcomes for presidential and congressional elections. We compare the forecasting performance of parametric models to nonparametric models. Current results find that there are strong relationships between economic indicators and political outcomes but lack accuracy in prediction.

Committee: Jing Li Dr. (Advisor); Jonathan Wolff Dr. (Committee Member); Charles Moul Dr. (Committee Member) Subjects: Economics

Doctor of Philosophy, The Ohio State University, 2021, Agricultural, Environmental and Developmental Economics

There are many determinants of health such as individual dietary and health-related habits, constraints such as money and time, as well as market goods and services such as medical care, access to health insurance, and environmental conditions. In this dissertation, I focus on three key elements of household and individual consumption behaviors that are tied to economics of health and nutrition—policy, preferences, and consumption self-control. In the first essay, I demonstrate how receiving subsidized health care services can lead to new patterns of household consumption, specifically, undertaking fewer preventative health measures by the targeted households. This topic has received less attention in the literature. To do this, I investigate the effects of recent Medicaid expansions on eligible households' quarterly food and non-food expenditures using state and time variation in Medicaid expansion. Using an event-study design, and a triple difference-in-differences framework, I find that the Medicaid eligible households from expansion states spent less on fresh produce per adult and more on over-the-counter medications and remedies while not changing their expenses on frozen fruits and vegetables which have similar nutritional value as fresh fruits and vegetables. The robust reduction in fresh produce expenditures and increase in expenditures on over-the-counter medications and remedies suggest that while expanded public health insurance increases formal healthcare activity, it decreases informal preventative non-healthcare expenditures. These findings may begin to shift the focus in the literature on the unintended consequences of Medicaid expansion from sins of commission, i.e., moral hazard responses such as increased smoking, alcohol use and junk food consumption, to sins of omission, i.e., responses in which preventative health habits erode. In the second essay, I focus on healthy eating in institutions such as schools and colleges, which is promoted in (open full item for complete abstract)

Committee: Brian Roe (Advisor); H Allen Klaiber (Committee Member); Zoë Plakias (Committee Member); Wuyang Hu (Committee Member) Subjects: Agricultural Economics

Doctor of Philosophy, The Ohio State University, 2021, Geodetic Science

Anthropogenic excessive groundwater depletion (GWD) is a major problem affecting numerous regions in the world that depend on these precious water resources for drinking, irrigation, industrial and urban needs. Climate change is thought to further exacerbate scarcity and degrade the quality of these freshwater resources globally. The Gravity Recovery and Climate Experiment (GRACE) and its successor, GRACE Follow- On (GRACE-FO) twin-satellite gravimetry missions, have been observing the global temporal variations in Terrestrial Water Storage (TWS) for almost two decades at monthly sampling and spatial resolution longer than 333 km (half-wavelength). Innovative methodologies have enabled the retrieval of Groundwater Storage (GWS) anomalies in the world's large climate-stressed aquifers, or disaggregated the signal from satellite gravimetry observed total TWS by removing the surface hydrologic signals via simulated or assimilated hydrologic model output data, or via hydrologic observations. However, uncertainties during the disaggregation process coupled with the limited spatial resolution (666 km grids) of GRACE/GRACE-FO estimated GWS have limited the use of such data for local-scale assessment of GWS variations and for practical applications of water resources management. In this research, we develop and leverage Machine Learning (ML) approach to estimate decadal or longer GW variations for the Central Valley (CV) in California, USA, and North China Plain (NCP) in China, to a local scale (5 km). These two study regions are among the regions in the world, largely dependent on GW for agricultural irrigation and other usages and are currently undergoing severe GWD due primarily to anthropogenic activities and plausibly exacerbated by an increasingly warmer Earth. First, we developed and implemented the robust Artificial Neural Network (ANN) and Random Forest (RF) ML modeling framework in the Central Valley (CV) to study the severe GWD problem using GRACE-derived TW (open full item for complete abstract)

Committee: C.K. Shum (Advisor); Orhan Akyilmaz (Committee Member); Michael Durand (Committee Member); Wei Feng (Committee Member); Ian Howat (Committee Member) Subjects: Geographic Information Science; Geography; Geological; Geophysical; Remote Sensing

Master of Science, University of Toledo, 2020, Engineering (Computer Science)

In this study, an Intrusion Detection system (IDS) is designed based on Machine Learning classifiers and its performance is evaluated for the set of attacks entailed in the UNSW-NB15 dataset. This dataset is comprised of 2,540,226 realistic network data instances as well as 49 features. Most studies reported in the literature employ a representative subset of this dataset with predefined training and testing subsets, and containing a total of 257,673 records which this study also used. In light of relatively lower than expected performance of Machine Learning or Statistical classification algorithms tested on this dataset and as reported by others in the literature, this dataset was subjected to visual data analysis to explore potential reasons or issues which likely challenge Machine Learning classifiers. The consequent observations demonstrated the presence of class representation imbalance with respect to pattern counts and class overlap in feature space, which makes preprocessing strategies indispensable before this dataset can be meaningfully employed for data-driven model development for intrusion detection. For preprocessing, we implemented min-max scaling in the normalization phase followed by the application of Elastic Net and Sequential Feature Selection (SFS) algorithms. We employed ensemble methods using three base classifiers, namely Balanced Bagging, XGBoost, and RF-HDDT, augmented to address the imbalance issue. Parameters of Balanced Bagging and XGBoost are tuned for the imbalanced data, and Random Forest is supplemented by the Hellinger distance metric to address the limitations of default distance metric. Two new algorithms are proposed to address the class overlap issue in the dataset and applied during training. These two algorithms are leveraged to help improve the performance on the testing dataset by affecting the final classification decision made by three base classifiers as part of the ensemble classifier which employsa majority vote combi (open full item for complete abstract)

Committee: Gursel Serpen (Committee Chair); Ahmad Y. Javaid (Committee Member); Mohammed Niamat (Committee Member); Richard G. Molyet (Committee Member) Subjects: Computer Engineering; Computer Science; Engineering; Mathematics; Statistics