College of Science and Engineering Projects and Theses

Permanent URI for this collection


Recent Submissions

Now showing 1 - 20 of 93
  • Item
    Evaluating Sentiment Analysis Mechanism for Labelled Amazon Reviews
    (2023-05-09) Trada, Parth Sureshbhai; Sha, Kewei; Wu, Yalong; Wei, Wei
    Sentiment analysis has become increasingly important in understanding customer opinions, feedback, and preferences towards products and services, particularly on marketplaces like Amazon. Researchers have proposed various techniques and algorithms for sentiment analysis. However, there still lacks a good guidance that can systematically direct data scientists to select appropriate algorithms and models, although a few efforts have been made. This thesis aims to fill the gap by presenting a comprehensive evaluation on different sentiment analysis mechanisms for labeled Amazon reviews. To achieve the above goal, we first prepare an accurately labelled Amazon review dataset through manually labeling. This builds a solid foundation for our evaluation. Then, we evaluate the effectiveness of popular mechanisms used in sentiment analysis, including both data preprocessing techniques such as Bag of Words (BOW), Term Frequency- Inverse Document Frequency (TF-IDF) weighting, spell correction, stemming, and lemmatization, and various sentiment analysis models such as K-Nearest Neighbors (KNN), Logistic Regression (LR), Support Vector Machine (SVM), Artificial Neural Network (ANN), Long Short-Term Memory (LSTM), and Bidirectional Encoder Representations from Transformers (BERT). These mechanisms were selected based on their prominence in the field of sentiment analysis, their potential to yield high-accuracy results, and their representation of different designs. We conducted five experiments using a combination of above data preprocessing techniques and analysis models. Through these experiments, we aim to identify a set of optimal combinations of preprocessing techniques and classification models that demonstrate superior performance in sentiment analysis of labeled Amazon reviews. The experiment results show that the use of BERT with BOW, TF-IDF, Spell Correction, and Lemmatization achieved the highest accuracy of 98.99%, outperforming other combinations. The addition of TF-IDF weighting, spell correction, stemming, and lemmatization improves the accuracy of four analysis models by about 6%, i.e., from 87.34% to 93.4% for KNN, from 86.6% to 94.22% for SVM, from 90.68% to 96.87% for ANN, and from 92.87% to 97.95% for LSTM. However, LR shows a comparatively lower accuracy ranging from 74.32% to 81.09% regardless different preprocessing techniques due to its limitations as a linear model, which may struggle to capture complex patterns and non-linear relationships in the sentiment data. This work provides insights into the effectiveness of different data processing and analysis mechanisms for sentiment analysis of labeled Amazon reviews. The findings can be applied to improve the effectiveness of customer review analysis to help achieve higher level of customer satisfaction, which can be essential in areas such as product and business strategy development.
  • Item
    Federated Learning to Build Sentiment Analysis Models for Amazon Review Datasets without Labels
    (2023-05-09) Kadve, Ritu; Sha, Kewei; Wu, Yalong; Wei, Wei
    In Natural Language Processing, one of the most popular tasks is sentiment analysis which aims to predict the sentiment of a text. It has many practical applications in industries such as marketing and customer service. Performance of sentiment analysis models play a significant role to the success of these applications. To achieve a high accuracy, sentiment analysis usually trains analytical models based on labeled datasets, preferably to large-scale labeled dataset. However, large-size labeled dataset may not be available because of the high-cost in labeling. Therefore, researchers study alternative approaches aiming to learn high accurate and reliable models based on small-scale labeled datasets or using other existing labeled datasets from different categories. A centralized model is a machine learning model that utilizes a large dataset stored on a central server to perform sentiment analysis. Training a centralized model on a small, labeled dataset can result in inaccurate or incomplete predictions. While processing labeled datasets of different categories on a centralized platform, it also comes with many challenges such as data heterogeneity, bias towards the categories that are overrepresented, requirement of large amount of computational power and resources, and the availability of good amount of labeled data for training. In addition, it is difficult to select appropriate data categories to train a reliable model for the new category. In this thesis, we propose a federated learning approach to overcome these challenges. Federated Learning (FL) is a type of decentralized Machine Learning (ML) that lets us train data analytical models on local data without transferring data to a central server. When Federated Learning is applied to sentiment analysis, one server and multiple clients collaborate to train a reliable and accurate sentiment analysis model. In our scenario, each client trains a local sentiment analysis model based on a labeled review dataset of a specific category, and the server makes use of the FedAvg algorithm to aggregate the parameters from the trained client models to build a global model for the new category that has no available labeled dataset. We evaluate the performance of our design based on a prototype implementation using Amazon review datasets. Compared with the centralized sentiment analysis, the proposed FL-based sentiment analysis performance is 10% better. This validates the potential of federated learning in training better data analytical models for categories with no large-scale labeled datasets.
  • Item
    Effectiveness of Drones for Freshwater Turtle Surveys Aimed Toward Detecting the Cryptic Western Chicken Turtle. (Deirochelys Reticularia Miaria).
    (2023-04-28) Nagro, Jason; Guillen, George; Mokrech, Marc; Howard, Cynthia; Gordon, Mandi
    The rise in the use of drones in wildlife research has shown promising results for conservation practices. Few studies have focused on drone surveys for aquatic freshwater turtles. This study evaluated the effectiveness of drones for detecting freshwater turtles with the primary target species being the Western Chicken Turtle (WCT; Deirochelys reticularia miaria). Two drones were employed to investigate their effectiveness for detecting freshwater turtles. 1) Videos and thermal imagery were utilized using a DJI Mavic 2 Enterprise (M2) and 2) static multispectral imagery using a DJI Phantom 4 (P4MS). Binocular aided visual surveys (BAVS) were conducted simultaneously with M2 surveys to compare and contrast methodologies. A total of 20.7 hours of video footage yielded 1916 freshwater turtle detections and 57090 photos with 1915 detections. BAVS had a cumulative time of 58.1 hours with 1096 turtle detections. Six turtle groups were detected with the M2, five with the P4MS and four with BAVS. Groups that were identified by all methods were Slider Turtles (Trachemys sp.), North American Softshell Turtles (Apalone sp.), and Common Snapping Turtles (Chelydra serpentina). A majority of all detected freshwater turtles displayed no reaction to the drones presence. The M2 had a statistically significant (p = 0.015) higher monthly catch per unit effort (CPUE) for freshwater turtles when compared to BAVS. Six WCT were detected using drone surveys (M2 = 5 and P4MS =1) while BAVS failed to detect WCT. Drone surveys were successful at detecting and identifying freshwater turtles such as the WCT when compared to BAVS, but quality data collection relies upon many internal and external factors such as camera resolution and essential habitat features. Drones are powerful tools when surveying freshwater turtles and other wildlife collecting vast amounts of data. Their implementation in future research studies concerning wildlife conservation with freshwater turtles have evident benefits in overall site accessibility, field team safety, and non-invasive rapid data collection.
  • Item
    Parameterizable Design on Convolutional Neural Networks with Chisel Hardware Construction Language
    (2023-04-21) Madineni, Mukesh Chowdary; Yang, Xiaokun; Koc, Hakduran; Unawala, Ishaq
    This thesis presents a parameterizable design generator on convolutional neural networks (CNNs) using Chisel hardware construction language (HCL). Chisel HCL is an open-source embedded domain-specific language (created and maintained by University of California, Berkeley) that inherits the object-oriented feature of Scala for constructing hardware. By parameterizing structural designs such as the streaming width, pooling layer type, and floating-point precision, multiple register-transfer level (RTL) implementations can be created to meet various accuracy and hardware cost requirements. The HCL design can generate the RTL implementations with Verilog, which is synthesizable and implementable on FPGAs (field-programmable gate arrays). The evaluation is based on generated RTL designs including 16-bit, 32-bit, 64-bit, and 128-bit implementations on FPGAs. The experimental results show that the 32-bit design achieves optimal hardware performance when setting the same weights for estimating the quality of results, FPGA slice count, and power dissipation. Although the focus is on CNNs, the approach can be extended to other neural network models for efficient RTL designs.
  • Item
    Talk Code-y To Me: An analysis of speech to text systems for consideration of use in writing software
    (2022-05-16) Tijerina, Isaac Guillermo; Datta, Soma; Helm, James C; Lacher, Lisa L
    This study proposes to create an application to allow ease of Speech to Text (STT) conversion specifically for programmers to make programming more accessible to those with disabilities. Recently there is being a movement of pairing STT with other disciplines now that STT is readily available and reliable. The main questions are how well Apple’s STT performs, is Apple’s STT ready to be integrated with coding, how do programmers interpret and speak code aloud, and how well does a formatting application created for this study to format transcriptions into executable code perform. The study concludes that Apple’s STT transcribes text at an average success rate of 50.1% and in correctly transcribing and interpreting words at an average success rate of 13.12%, whether it is ready to be used in coding is up to the reader, programmers interpret and speak code in wide range of ways, and the application had a success rate of 0% but the data collected will help it to improve in the future.
  • Item
    Swarm intelligence application in solving robot inverse kinematic problems
    (2021-07-28) Danaci, Hasan; Nguyen, Luong; Harman , Thomas L.; Koc , Hakduran
    This dissertation aims to find the inverse kinematic solution for redundant serial manipulators using the meta-heuristic method, Particle Swarm Optimization algorithm (PSO). Primarily this paper focuses on moving the end-effector to any desired pose in cartesian space accurately by converging position and orientation with the PSO algorithm. In order to prove the exactness of the study, the result has been compared with some of other PSO research that only examines converging the position. All demonstrations were performed by using humanoid human-sized with 7 degrees of freedom robot (DOF), Baxter. First, the Denavit-Hartenberg(DH) table of Baxter's left arm is created, and transformation matrices are calculated according to two different setups joint angles to calculate target position and orientation values. Furthermore, joint angles are picked randomly for each particle, and the particles' pose is calculated by applying forward kinematics. In order to obtain subsequent angle values, the PSO algorithm, conversion of quaternion to a rotation matrix, and Jacobian matrices are utilized. This research gives another perspective to solving inverse kinematic by using quaternions instead of Euler angles. The Euclidian function is used to compute the cost function, which estimates the distance between the target pose and particle's pose. In this study, the algorithm is tested with several different concepts. Conclusively, the validity of the algorithm is verified via Gazebo simulation. The result confirms that the algorithm functions well in accuracy and merit of the swarm intelligence in solving the inverse kinematics problem for any serial robotic manipulators.
  • Item
    Reviewtag: Tagging Amazon Negative Product Review with Deep Learning
    (2023-05-19) Kumari, Priyanka; Sha, Kewei; Wu, Yalong; Wei, Wei
    The success of Amazon sellers hinges on high ratings and meeting customer needs with exceptional products and services. However, the large scale of negative reviews pose significant challenges that require careful analysis to identify underlying reasons of buyers concerns. We aim to develop an automated tagging system named ReviewTag to address this challenge. The system uses deep learning models and Natural Language Processing (NLP) techniques to swiftly categorize negative reviews into two broader categories i.e., product issues and seller issues. The system provides further insight into customers' specific issues using subtopic tagging, allowing Amazon sellers to identify areas for improvement and make data-driven decisions to meet evolving customer expectations. We employ five deep learning models to perform topic and subtopic tagging. These models include Bidirectional Encoder Representations from Transformers (BERT), Distilled Bidirectional Encoder Representations from Transformers (DistilBERT), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Recurrent Neural Network (RNN). Based on the evaluation with a prototype implementation of ReviewTag, the BERT model demonstrates high precision, recall, and F1-scores of 0.97, 0.96, and 0.96, respectively, for topic tagging. Additionally, the BERT and CNN models show impressive precision, recall, and F1-scores of about 0.92, for subtopic tagging. These results demonstrate the effectiveness of deep learning models for automatically tagging negative product reviews on Amazon. It helps Amazon sellers take action to improve their product ratings.
  • Item
    Examining the ecological function of small-scale living shorelines in Galveston Bay
    (2022-12-01) Gilbert, Ryan Scott; Howard, Cynthia; Zhang, Chunlong; Imrecke, Daniel; Dobberstine, James
    Living shorelines are an increasingly popular shoreline protection strategy. In contrast to traditional armoring techniques such as concrete bulkheads, living shorelines are designed to provide the many ecological functions and benefits of natural coastal wetlands. Despite a wealth of knowledge on coastal wetland restoration, studies verifying ecological function in living shorelines are limited. The objective of this study was to provide a comprehensive ecological assessment of three living shoreline projects in the Galveston Bay system. This study collected data on stem density, percent cover, and root-biomass to characterize plant communities. Data was also collected on the abundance and community structure of benthic and nekton organisms. Additionally, sediment heavy metal concentrations were examined. Living shoreline data was compared to both natural and armored shorelines ultimately confirming the suspected hypothesis that living shoreline sites function similarly to natural shorelines and improved over armored shorelines.
  • Item
    Shock-Induced Geochemical Variations in the Keplerite-Bearing Assemblages of Tissint and Intergrown Apatite-Merrillite Assemblages of ALH 84001,146
    (2022-12-01) Slabic, Ane; Imrecke, Daniel B.; Sun, Dongmin; Treiman, Allan H.; Filiberto, Justin R.
    Shock metamorphism describes changes (physical, chemical) that occur to materials (structures, rocks, minerals) as a result of shock wave deformation. All Martian meteorites have been subjected to shock, though the degree of shock, and subsequent changes to the physical and chemical characteristics of these rocks, is highly variable. Extensive chemical and structural heterogeneities may occur as a result of shock metamorphism; however, the way in which phosphate minerals respond to shock-metamorphism is not well constrained. Here we present new evidence of shock-induced chemical variations in the keplerite-bearing assemblages of Tissint and the apatite-merrillite assemblages of ALH 84001. Geochemical and structural studies of phosphate phases were investigated using Electron Probe MicroAnalyzer (EMPA), Raman Spectroscopy, Scanning Electron Microscopy (SEM), and Electron Backscatter Diffraction (EBSD) analytical techniques. Phosphates identified included Low-Na (Na# = 2.3) keplerite in Tissint and intergrown Cl-rich-apatite with High-Na merrillite (Na# = 8.3) in ALH 84001. Na-number is the calculated atomic ratio of sodium to the sum of sodium and calcium ([Naatomic/(Naatomic+Caatmoic)]*100). Presence of keplerite (a newly defined, high-temperature, Na-deficient, phosphate) in Tissint suggests it may be common and occur in a broader range of environments (i.e., meteorite groups) than initially established. ALH 84001 apatite was found to be heterogeneously enriched in halogens with Cl between 3.85 and 5.05 wt% and F between 0.30 and 0.92 wt%. Assuming F + Cl + OH = 1.0 structural formula unit (sfu); average atoms per formula unit (apfu) of F = 0.16, while apfu of Cl = 0.63. Therefore, water content of apatite in ALH 84001 is estimated at 0.21 sfu or, ~21% hydroxyl apatite component. Thus, indicating that parental magmas were Cl-rich and OH-poor. Intergrown apatite-merrillite phases were observed as well, with textures suggesting the replacement of magmatic apatite by merrillite in the solid-state following a shock-metamorphic event.
  • Item
    Application of Whole Genome Sequencing and Maldi-Tof to Identification of Bacillus Species Isolated from Cleanrooms at Nasa Johnson Space Center
    (2021-12-15) Mazhari, Farnaz; LaMontagne, Michael G.; Regberg, Aaron B.; Davis, Richard E.; Santiago-Vázquez, Lory Z.
    Astromaterial cleanrooms at NASA Johnson Space Center are built environments that hold samples, such as lunar rocks, from different space exploration missions. Bacillus sp. are frequently detected in routine microbial monitoring of these facilities. Since this, and related genera, can form endospores that can withstand harsh conditions, they could contaminate astromaterials. This could confound searches for extraterrestrial life. Whole genome sequencing (WGS) is widely used for identifying bacterial strains and tracking their source; however, WGS is expensive and time consuming. Matrix-assisted laser desorption ionization– time of flight mass spectrometry (MALDI-TOF) shows promise as a low-cost, rapid method of identifying strains of bacteria, but few studies have compared this proteomics method to WGS. To evaluate a high throughput method of tracking the source of contamination of this built environment, WGS and MALDI-TOF was conducted on 18 bacterial strains isolated from surfaces in astromaterials cleanrooms. WGS identified 14 Bacillus, 2 Paenibacillus, 1 Solibacillus and 1 Alcaligenes strains. These isolates showed similarity to strains commonly observed in spacecraft assembly cleanrooms at other facilities. Cluster analysis of mass spectra generated by MALDI-TOF grouped strains together that were greater than 94% similar to each other in terms of amino acid sequences of single copy core genes, as assessed by WGS. This suggests that MALDI-TOF and WGS results are consistent with each other and MALDI-TOF can rapidly identify strains of Bacillus sp. isolated from cleanroom environments with a resolution comparable to WGS. Based on phylogenomic analysis, these results also suggest the presence of a cosmopolitan class of Bacillus sp. that are more likely to be found in cleanrooms and similar built environments than in natural systems.
  • Item
    Oa1s and Oa2s Radiative corrections for w-boson production at the tevatron and atlas experiment
    (2022-05-31) Teel, Matthew James; Masood, Samina; Garrison, David; Withey, Paul
    In this thesis, we propose an approach for calculating W-Boson Transverse Momentum Dependent (TMD) distributions using the LHAPDF library with nCTEQ15, MSTW2008 and CT10 Parton Distribution Functions (PDF). We utilize ManeParse as a PDF reader in the Mathematica framework for importing and computing relevant cross-sections, luminosity functions, and error analysis, and additionally allow cross-check with a number of sum rules. A proper description of TMD distributions requires resummation of large logarithms responsible for the divergence of the perturbation series in the strong coupling as. Our proposed calculation is for next-to-leading-order (Oa1s) and next-to-next-to-leading-order (Oa2s) large logarithmic perturbative corrections for the differential W  pT partonic cross-sections for processes of the form pp -> W  + X. The goal of large logarithmic corrections with threshold resummation is to improve the accuracy of the pT distribution thus reducing dependence of the cross-section on renormalization (uR) and factorization (uF) scales, and plays a role in Higgs searches, precision measurement of the W-boson mass, as well as for testing perturbative QCD.
  • Item
    Application of Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS) for Differentiating Sources of Fecal Pollution
    (2022-05-27) Kmiecik, Caroline; LaMontagne, Michael G; Santiago-Vazquez, Lori; Howard, Cindy
    Since the passage of the Marine Mammal Protection Act in 1973, pinniped populations in coastal waters of the United States have increased exponentially. These high populations of large mammals could contribute to fecal contamination of recreational waters. Enterococci species counts are used to assess the degree of fecal contamination and elevated counts of this fecal indicator bacteria (FIB) force managers to close beaches; however, contribution of pinnipeds to high Enterococci counts is not known. This may reflect the high cost of methods of tracking the source of microbial contamination. Matrix Assisted Laser Desorption Ionization – Time of Flight Mass Spectrometry (MALDI-TOF MS) is a time and cost-effective way to identify bacteria through protein mass spectra analysis. MALDI-TOF can distinguish strains of bacteria of the same species but has not been evaluated as a tool for tracking Enterococci isolated from pinnipeds. In this study, Enterococci isolates were cultured from fresh seal scat samples taken from a pen housing Harbor seals. Enterococci were isolated by plating serial dilutions onto Rapid Enterococci ChromoSelect agar to generate a library of isolates from captive seals. Libraries of Enterococci were also generated from dog scat and a composite sample of human waste from a wastewater treatment plant, respectively. These isolates were identified with a MALDI-TOF system and cluster analysis was performed of all Enterococci isolates showed overall clusters between Enterococci isolated from a single source. Cluster analysis of isolates from harbor seal and dog scat reliably identified as E. faecalis with Bruker MALDI Biotyper system (> 2.3) was performed for source discrimination comparison. 22 E. faecalis isolates clustered into four separate MTUs, with two MTUs being harbor seal-specific. The distinct source-dependent MALDI-TOF MS clusters suggest that MALDI-TOF MS may be a valuable tool in microbial source tracking.
  • Item
    Foraging Ecology of Bottlenose Dolphins (Tursiops Truncatus) in Galveston Bay, Texas
    (2022-05-31) McDaniel, Sherah Ann; Guillen, George J; Howard, Cindy; Zhang, Chunlong (Carl)
    The overall goal of this research was to develop a better understanding of the trophic ecology of the Galveston Bay common bottlenose dolphin (Tursiops truncatus) stock that would provide critical data needed to manage this species. The specific objectives of this study were to: (1) estimate areas used by dolphins for foraging, (2) estimate factors contributing to foraging behaviors of dolphins, and (3) estimate proportions of different prey consumed by bottlenose dolphins in Galveston Bay. From 2015-2017, two survey methods (behavioral data from photo-identification surveys and stable isotope data from biopsy surveys) were used for objective one and two, while stable isotope (δ13C and δ15N) mixing models were used for objective three. Dolphins were observed foraging for 62.3% of all photo-ID sightings. Dolphins were observed foraging most often in the channel (52.3%), followed by open bay (41.4%), and nearshore (6.3%) but there is no evidence that the odds of observing foraging behavior was different between the habitats. For the sightings where foraging behavior was observed, 68.7% of the time a trawler was present. It is estimated that when approaching a shrimp trawler during a survey, the probability of observing dolphins patrolling is 60.8% of the time (95% CI: 55.6% to 100.0% , one-sided, one-sample porportion test, p-value <0.05). Foraging significantly decreased as time passed throughout the day in sightings from 2015-2017 (beta regression: pseudo R2= 0.8726, p-value <0.05). Potential prey of dolphins were collected in 2015 and 2016 for stable isotope analysis. Data from those sampling events and select nekton from Barcenas (2013) were used to model proportions of prey consumed by dolphins using A Bayesian isotope mixing model, Stable Isotope Analysis in R (MixSIAR version 3.1.10). Ward’s hierarchical cluster analysis was used to group 19 nekton species into six groups based on their mean C and N isotopic values. Overall, group six which contained only one species, White Mullet (Mugil curema), was estimated to contribute to the highest proportion of nekton prey consumed by dolphins (median: 25.3%) based on MixSIAR analyses. The second highest proportion consumed by dolphins overall was group two (Atlantic Brief Squid [Lolliguncula brevis], Hardhead Catfish [Ariopsis felis], and Striped Mullet [Mugil cephalus]) at 21.0%. There was a significant difference between the δ15N (‰) values in Upper Galveston Bay (UGB) and Lower Galveston Bay (LGB) (Wilcoxon Rank Sum Test, W=105.5, p-value≤0.05). This difference may suggest that dolphins in UGB and LGB are foraging on different prey or may support the notion that the upper portions of the bay are more heavily influenced of elevated anthropogenically produced δ15N (‰). This research contributes to baseline data that can be used for further analysis in future studies. The results from the stable isotope analysis may be used in combination with mercury and organochlorine contaminant analysis to examine trophic level biomagnification in the Galveston Bay ecosystem.
  • Item
    Application of MALDI-TOF MS for multiple source tracking from sewage and septic tanks along with identification of antibiotic-resistant pathogenic Escherichia coli
    (2022-05-26) Mirani, Akshita; LaMontagne, Michael G; Santiago-Vàzquez, Lory Z; Labonté, Jessica; Allen, Michael
    The intensity of extreme flooding events, driven by tropical cyclones and sea-level rise, may increase dramatically this century. These extreme weather events can spread untreated sewage from wastewater treatment plants and onsite wastewater treatment systems (OWTS, septic tanks), which creates a possibility of outbreaks of water-borne diseases. Human waste represents a particular threat because it is laden with antibiotic-resistant bacteria. In particular, bacteria in waters that appear to be contaminated with human waste show a high level of resistance to the antibiotic polymyxin. To assess the risk of disease spread, managers need tools to track the source of this contamination. Escherichia coli (E. coli) is widely used as a fecal indicator bacteria (FIB). Elevated levels of this FIB suggest microbial contamination but not the source. Commonly used microbial source tracking tools, are time-consuming and expensive. Matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF MS) is time and cost-effective way to identify bacteria. This proteomics method can distinguish strains of bacterial species but has not yet been widely used for microbial source tracking. In this study, E. coli strains were isolated from wastewater treatment plants and a sample composited from OWTSs. For comparison, a library of isolates was also generated from seal scat and dog feces. Isolates were then identified by MALDI-TOF MS, and cluster analysis was performed of mass spectra to determine if this technique could differentiate the sources of these FIB. To get knowledge on E. coli resistance towards antibiotics, a disc diffusion assay was implemented to screen representative isolates for sensitivity to broad spectrum antibiotics, and polymyxin B. Colistin resistance was checked by PCR using primers mcr-1, mcr-2, and mcr-3. MALDI-TOF MS distinguished E. coli strains isolated from sewage and OWTSs from E. coli isolated from animal sources. Antibiotic resistance assays indicated that E. coli strains isolated from all sources were resistant to ampicillin, streptomycin, and gentamicin. Bacteria isolated from sewage and OWTS showed resistance to colistin and polymyxin B. This suggests that MALDI-TOF MS could be applied to track the sources of fecal contamination of waterways. This could improve risk assessment and point to mitigation strategies.
  • Item
    Habitat Use, Distribution, and Density of Dwarf Seahorse (Hippocampus zosterae) Populations along the Texas Gulf Coast
    (2022-05-24) Lesher, Story C; Guillen, George; Oakley, Jenny W; Howard, Cynthia
    The Dwarf Seahorse (Hippocampus zosterae) is one of the smallest species of seahorse and resides in shallow waters throughout the Gulf of Mexico, Atlantic Coast of Florida, and Caribbean. They rely on seagrass beds for feeding, spawning, and refuge, rarely traveling far from the bed in which they were spawned. Information on density, distribution, and habitat associations in Texas are needed to address knowledge gaps and inform future management initiatives. We evaluated seagrass beds along the Texas Coast (from Galveston Bay to the Lower Laguna Madre) to determine what factors correlated with Dwarf Seahorse presence and density. Dwarf Seahorse (n=79) were captured at 30 of 80 (36.3%) sites. They were detected in all sampled bay systems except Galveston Bay. Dwarf Seahorse were found in association with all seagrass species found in Texas. Variables significantly correlated with Dwarf Seahorse presence included average seagrass biomass, turtle grass (Thalassia testudinum) percent cover, and nekton species abundance, evenness, and richness. Variables significantly correlated with Dwarf Seahorse density included number of seagrass species present, the presence of turtle grass, and nekton species evenness, and richness. Nekton species associated with the presence of Dwarf Seahorse included grass shrimp (Palaemonetes spp.), Penaeid shrimp, Code Goby (Gobiosoma robustum), and Rainwater Killifish (Lucania parva). There were an estimated 1,116,356 Dwarf Seahorse in Aransas Bay, 2,769,686 in Upper Laguna Madre, and 7,899,652 in Lower Laguna Madre throughout the sampling time frame. Results from this study will inform continued monitoring, development of a habitat suitability index, and management of this species and their essential habitat.
  • Item
    Method of Using Blockchain for Hardware Development Process
    (2022-05-10) Ejaz, Mudassir; Unwala, Ishaq; Lu, Jiang; Yang, Xiaokun
    Modern microprocessors contain many millions of gates and finding any hidden malicious Hardware Trojan (HT) is difficult. Additionally, HTs may not need any additional external input pins to activate. Multiple solutions have been proposed to find these HTs, but none of these solutions have promising result due to their limitations. Moreover, pre-silicon verification and post-silicon testing don’t address the issue of HTs. In this thesis we present a methodology based on blockchain technology to limit the possibility of inserting HTs into the design. Blockchain based technology limits the opportunity for insertion of HT, avoiding them in the design and fabrication process. We proposed a solution by monitoring hardware development process files, and maintaining integrity and trustful relationship using encryption and smart contracts in a blockchain network.
  • Item
    Influence of Freshwater Inflow in the Brazos River Estuary
    (2022-05-24) Jaeger, Joshua James; Guillen, George J; Howard, Cynthia L; Sun, Dongmin
    The influence of freshwater inflow on estuarine ecology is a topic that receives a lot of attention due to estuaries acting as medians to oceanic and freshwater environments. Freshwater inflow has varying effects depending on the morphology of the estuary, and the Brazos River is one of only three rivers in Texas with a riverine estuary that discharges directly into the Gulf of Mexico. The Environmental Institute of Houston (EIH) has been collecting data to categorize the ecology and hydrology within the estuary since 2012 when the Brazos Basin and Bay Area Stakeholder Committee (BBASC) formally addressed the need for research assessing the impacts of freshwater inflow. Project objectives included: (1) Describing the temporal variation in hydrology in the lower Brazos River, (2) evaluating the relationships between nekton community, freshwater inflow, seasonality and water quality using graphical and statistical methods, (3) characterizing nekton abundance, diversity, and community composition, (4) identifying focal species for different sites, flow tiers and seasons, and (5) identifying any future research needs. Data acquired from the United States Geological Survey (USGS) gage #08116650 was used in order to assess the normal trends of variation in discharge (cfs), as well as determine if data utilized far upstream from the estuary could be used to accurately predict discharge and water quality downstream. Automated monitoring loggers were used to collect long-term data for salinity (psu), temperature (°C), dissolved oxygen (DO) (mg/L), and depth (m) beginning in 2014 in order to create predictive models from daily average discharge. The results indicated that discharge in the Brazos River does exhibit predictable seasonal patterns of increased flow during the spring and reduced flow during the summer, yet still exhibits a huge degree of variation within seasons and between years. The Brazos River is also subject to extreme flow conditions at an increasing rate despite having an average annual discharge of only 7,400 cfs. The regression analysis from the automated loggers and USGS discharge data indicates that salinity and water depth are strongly correlated to flow and react in the form of exponential decay and sigmoidal growth respectively. Temperature proved to not be significantly correlated with flow, but multiple linear regression analysis with DO data demonstrated a complex relationship with flow and temperature. Two-way ANOVAs were used to determine if variations in water quality variables- temperature, salinity, DO, pH, turbidity (NTU), thalweg depth (m), and Secchi disk transparency (m) could be explained by sites, TCEQ flow tiers, or an interaction between both. The results showed that all variables exhibited significant differences (p ≤ 0.05) between flow tiers and sites- the only exception being temperature. The only two variables tested with a significant interaction effect between flow tier and site were salinity and pH. Principal component analysis was also used to determine whether season, flow tier, or spatial differences could explain the variability in water quality between samples. The results indicated that season is the primary driver of variability due to temperature having the highest eigenvector coefficients for each principal component calculated for surface and bottom profile. Pearson correlation analysis was also used to determine how water quality could be used to predict nekton community diversity. Salinity proved to be the most significantly correlated to nekton communities sampled using both an otter trawl and beam trawl. Nekton community metrics in the form of total catch, species richness, Shannon-Wiener diversity, Shannon evenness, Margalef Richness and catch per unit effort (CPUE) were subjected to nonparametric tests in order to test for significant differences between season, site, and flow tier. The results indicated the larval fish community sampled with the beam trawl near the shore was significantly affected by season, while spatial differences and flow tier explained the variation in nekton communities sampled mid-channel with the otter trawl. The results of this study corroborate many of the conclusions drawn from earlier studies on the ecology of the lower Brazos, as well as provide additional evidence of the highly dynamic nature of the Brazos estuary, while also providing further justification for continued and expanded research.
  • Item
    Deep Neural Network & Dynamic Functional Connectivity Analysis of Functional MRI Data
    (2022-05-05) Mishra, Amaresh Kumar; Sakoglu, Unal Z; Mubaid, Hisham A; Lu, Jiang
    DEEP NEURAL NETWORK & DYNAMIC FUNCTIONAL CONNECTIVITY ANALYSIS OF FUNCTIONAL MRI DATA Amaresh Kumar Mishra University of Houston-Clear Lake, 2022 Thesis Chair: Unal ‘Zak’ Sakoglu, PhD This thesis work presents a dynamic functional connectivity (DFC)-based classification analysis of an already collected and completely de-identified functional magnetic resonance imaging (fMRI) dataset from two groups, veterans with Gulf War Illness (GWI), vs matched controls. Neuroimaging or brain imaging is the use of various techniques to either directly or indirectly image the structure, function, or pharmacology of the nervous system. fMRI is a neuroimaging technique which is used to measure brain activity by detecting changes associated with blood oxygenation level dependence (BOLD), which is an indirect measure of neural activity, and it helps obtain three spatial dimensional (3D) brain activation maps associated with certain stimulus and/or a task, depending on the experiments performed during the fMRI scan. Whole-brain resting-state fMRI (rsfMRI) data which were scanned from 23 GWI veterans (mean age 49.4) and 30 normal control (NC) veterans (mean age 49.8) were used for analyses. A computational method using DFC features, deep learning, and machine learning techniques were used to correctly classify GWI vs NC. Results show that, support vector machine (SVM) -based machine learning technique, combined with simple t-test method for feature extraction (using the DFC), performed better than convolutional neural network (CNN) deep learning method, in terms of classification accuracy (upwards of 98% accuracy for the former vs. upwards of 60% accuracy for the latter).
  • Item
    Deriving the initial conditions of the electroweak and QCD phase transitions and testing a numerical relativity code.
    (2021-08-05) Barrera, Joshua Eulece; Garrison, David; Withey, Paul; de la Torre, Rafael
    The field of Numerical Relativity (NR) has been primarily driven by the study of largescale dynamics involving binary systems of black holes and neutron stars. Due to the nature of the underlying theory, NR also has the ability to simulate relativistic fluids. Being radiation dominated, the early universe can be modeled as a relativistic plasma, and the proper Stress-Energy tensor can be utilized with Einstein’s Field Equations to evolve the conditions of the Early Universe over time. These simulations can give us key insights pertaining to the development of the universe and the formation of large-scale systems. Magnetogenesis is of particular interest, as characterizing this phenomenon could shed light on the seeding and formation of galaxies. Addition-ally, these techniques can be used to derive gravitational-wave spectra from the events that took place during these time periods. This thesis aims to derive the conditions present in the early universe during the Electroweak (EW) and Quantum Chromodynamic (QCD) phase transitions. These conditions are prerequisites for SpecCosmo, a NR code being developed at the University of Houston-Clear Lake, which utilizes the techniques of NR to model the evolution of the early universe. This thesis will also investigate the ability of SpecCosmo to handle relativistic shocks in its current state. The shock capturing ability of the code will be gauged using a suite of tests from work by Komissarov [1, 2]. Once the shock capturing ability of SpecCosmo has been analyzed, SpecCosmo will be ready to accept the initial conditions calculated here that can then be used in simulations of the early universe to study its development and characteristics.
  • Item
    Computational Analysis of Various Reactions Between Hydroxyl Radicals and Organic Mercury Species
    (2021-07-29) Perry, Cody James; Su, Yi; Dubrovskiy, Anton; Wang, Daniel
    The interactions between hydroxy radicals and both dimethylmercury and methylmercury hydroxide have been studied using ab-initio methods. The density functional theory (DFT) method M06-2X was used to optimize the structures of the stationary points in each reaction channel. The single point energies were determined using coupled-cluster theory (CCSD(T)). For the reaction with dimethylmercury there are three possible channels, with the favored product as methylmercury hydroxide with the smallest energy barrier of 7.3 kcal mol-1 (k(TST)=1.23x10-12 cm3 molecule-1 s-1). For the reaction between hydroxyl radical and methylmercury hydroxide there are six possible channels with the most probable channel being a hydrogen abstraction to form MeHgO radical. This channel has the lowest energy barrier of 2.5 kcal mol-1 with a rate constant of 3.87x10-9 cm3 molecule-1 s 1. Based on these results, dimethylmercury will undergo demethylation due to hydroxyl radicals whereas methylmercury hydroxide will not.