Digital Commons Network^™

Llms In Network Intrusion Detection – A Comprehensive Analysis, Sudharshan Balaji

USF Tampa Graduate Theses and Dissertations

Network Intrusion Detection Systems (NIDS) play a critical role in identifying and mitigating malicious activities within computer networks. With the rapid evolution of natural language processing (NLP), Large Language Models (LLMs) have emerged as transformative tools across various domains. LLMs, such as OpenAI’s GPT series and Meta’s LLaMA models,have demonstrated remarkable performance in tasks like language generation, reasoning, and classification. Their ability to understand and process vast amounts of data has enabled groundbreaking advancements in areas like healthcare, finance, and cybersecurity. Recent trends highlight their potential to handle unstructured data, perform complex reasoning, and adapt to a wide range of …

Learning Soft Tissue Deformation From Incremental Simulations, Nathan Lampen, Daeseung Kim, Xuanang Xu, Xi Fang, Jungwook Lee, Tianshu Kuang, Hannah H Deng, Michael A K Liebschner, Jaime Gateno, Pingkun Yan

Faculty, Staff and Students Publications

Background: Surgical planning for orthognathic procedures demands swift and accurate biomechanical modeling of facial soft tissues. Efficient simulations are vital in the clinical pipeline, as surgeons may iterate through multiple plans. Biomechanical simulations typically use the finite element method (FEM). Prior works divide FEM simulations into increments to enhance convergence and accuracy. However, this practice elongates simulation time, thereby impeding clinical integration. To accelerate simulations, deep learning (DL) models have been explored. Yet, previous efforts either perform simulations in a single step or neglect the temporal aspects in incremental simulations.

Purpose: This study investigates the use of spatiotemporal incremental modeling …

Enhancing Diagnostic Accuracy Of Drug-Resistant Tuberculosis On Chest X-Rays Using Data-Efficient Image Transformers, Joan Mnyambo, Amir Aly, Shang-Ming Zhou, Yinghui Wei, Stephen Mullin, Emmanuel Ifeachor

School of Engineering, Computing and Mathematics

Tuberculosis is an infectious disease with increasing fatalities around the world. The diagnosis of the diseaseis a major challenge to its control and management due to the lack of adequate diagnostic tools, contributingsignificantly to the prevalence of drug-resistant tuberculosis. Convolutional Neural Network (CNN) modelshave recently been developed to detect drug-resistant tuberculosis by analyzing chest radiograph images fromthe TB portal, but the classification results are low. This is because CNNs struggle to capture complex globaland overlapping features in medical imaging, such as chest radiographs of drug-resistant tuberculosis. Incontrast, transformers excel in these areas by utilizing self-attention mechanisms that detect inherent subtle …

Personalized Drug Recommendation System Using Wasserstein Auto-Encoders And Adverse Drug Reaction Detection With Weighted Feed Forward Neural Network (Waes-Adr) In Healthcare, J Omana, P. N Jeipratha, K Devi, S Benila, K Revathi

Journal of Informatics and Web Engineering

In recent years, the use of deep learning approaches in healthcare has yielded promising results in a variety of fields, most notably in the detection of adverse drug reactions (ADRs) and drug recommendations. This paper promises a breakthrough in this field by using Wasserstein autoencoders (WAEs) for personalized medicine recommendation and ADR detection. WAEs' capacity to manage complex data distributions and develop meaningful latent representations makes them ideal for modeling heterogeneous healthcare data. This study intends to improvise the precision and efficiency of drug recommendation systems while also improving patient safety by combining WAEs and early ADR detection strategies. Previous …

Rapid Prediction Of Coastal Flooding With Deep Neural Networks, Ali Shahabi, Navid Tahvildari

Graduate Student Government Association Research Conference

With the increasing impact of climate change and relative sea level rise, low-lying coastal communities face growing risks from extreme storm tides and recurrent nuisance flooding. Thus, timely and reliable predictions of coastal water levels are critical to resilience in vulnerable coastal areas. Over the past decade, enormous efforts have been made to utilize machine learning (ML) based data-driven models for the emulation and prediction of storm tides. However, flood advisory systems still rely on running computationally demanding real-time hydrodynamic models. because developing highly reliable ML-based models suitable for real-time forecasting and capable of capturing any surge levels is challenging. …

Bridging Expert Knowledge With Deep Learning Techniques For Just-In-Time Defect Prediction, Xin Zhou, Donggyun Han, David Lo

Research Collection School Of Computing and Information Systems

Just-In-Time (JIT) defect prediction aims to automatically predict whether a commit is defective or not, and has been widely studied in recent years. In general, most studies can be classified into two categories: 1) simple models using traditional machine learning classifiers with hand-crafted features, and 2) complex models using deep learning techniques to automatically extract features from commit contents. Hand-crafted features used by simple models are based on expert knowledge but may not fully represent the semantic meaning of the commits. On the other hand, deep learning-based features used by complex models represent the semantic meaning of commits but may …

Deep Learning-Based Auto-Segmentation For Liver Yttrium-90 Selective Internal Radiation Therapy, Jun Li, Wookjin Choi, Rani Anne

Department of Radiation Oncology Faculty Papers

The aim was to evaluate a deep learning-based auto-segmentation method for liver delineation in Y-90 selective internal radiation therapy (SIRT). A deep learning (DL)-based liver segmentation model using the U-Net3D architecture was built. Auto-segmentation of the liver was tested in CT images of SIRT patients. DL auto-segmented liver contours were evaluated against physician manually-delineated contours. Dice similarity coefficient (DSC) and mean distance to agreement (MDA) were calculated. The DL-model-generated contours were compared with the contours generated using an Atlas-based method. Ratio of volume (RV, the ratio of DL-model auto-segmented liver volume to manually-delineated liver volume), and ratio of activity (RA, …

Real-Time Cbct Imaging And Motion Tracking Via A Single Arbitrarily-Angled X-Ray Projection By A Joint Dynamic Reconstruction And Motion Estimation (Dreme) Framework, Hua-Chieh Shao, Tielige Mengke, Tinsu Pan, You Zhang

Faculty, Staff and Student Publications

Objective: Real-time cone-beam computed tomography (CBCT) provides instantaneous visualization of patient anatomy for image guidance, motion tracking, and online treatment adaptation in radiotherapy. While many real-time imaging and motion tracking methods leveraged patient-specific prior information to alleviate under-sampling challenges and meet the temporal constraint (< 500 ms), the prior information can be outdated and introduce biases, thus compromising the imaging and motion tracking accuracy. To address this challenge, we developed a framework dynamic reconstruction and motion estimation (DREME) for real-time CBCT imaging and motion estimation, without relying on patient-specific prior knowledge.

Approach: DREME incorporates a deep learning-based real-time CBCT imaging and motion estimation method into a dynamic CBCT reconstruction framework. The reconstruction framework reconstructs a dynamic sequence of CBCTs in a data-driven manner from a standard pre-treatment scan, without requiring patient-specific prior knowledge. Meanwhile, a convolutional neural network-based motion encoder is jointly trained during the reconstruction …

An Intelligent Robotic System For Multi-Sensory Cognitive Fatigue Detection To Assist Persons With Paralysis In Activities Of Daily Living, Enamul Karim

Computer Science and Engineering Dissertations - Archive

Assistive robotics is a promising area for improving the quality of life of people with paralysis, specifically through assistance in Activities of Daily Living (ADLs). Current state-of-the-art assistive robotic systems do not have the capability to dynamically modulate their functionality according to the cognitive fatigue level of the user, which can negatively impact their effectiveness and usability in real-life settings.

This dissertation explores an adaptive robotic framework that adjusts its behavior depending on the cognitive fatigue level of users. The system operates in three different modes, and switches between Fully Controlled, Semi-Autonomous, and Fully Autonomous modes. The overall goal is …

Improving Vulnerable Road Users' Safety Through Computer Vision-Based Crossing Direction Prediction And Multimodal Large Language Model-Based Accident Scene Description, Younggun Kim

Graduate Thesis and Dissertation post-2024

Vulnerable Road Users (VRUs), such as pedestrians and cyclists, are among the most at-risk participants in traffic, making their safety a key priority for intelligent transportation systems (ITS). Accurate perception and understanding of VRU behavior are essential for proactive accident prevention and the design of human-centric mobility systems. Computer vision has become a core component of ITS, enhancing VRU safety. This thesis introduces a human behavior-aware transformer-based framework for VRU intention prediction and presents VRU-Accident, a large-scale benchmark enabling systematic assessment of multimodal large language models (MLLMs) in understanding accident scenarios involving VRUs.

First, the proposed framework leverages multi-modal cues, …

Energy Considerations For Large Pre-Trained Neural Networks, Leo Mei

Master's Projects

In recent years, neural network models have achieved phenomenal performance due to the increasing parameters and complexity of model architectures. However, these advancements come with high environmental costs as they require massive computational resources and consume substantial amounts of electricity, leading to high carbon emissions. Previous studies have demonstrated that substantial redundancies exist in large pre-trained models, and reducing these redundancies through compression would not compromise model performance. While these studies focused on retaining comparable model performance, the direct impact of compression on energy consumption when training models appears to have received little attention. By quantifying the energy usage associated …

Large Language Models For Bacterial Genomic Analysis, Manvendra Chavan

Master's Projects

Identification of bacterial gene sequences with agricultural applications has the potential to transform agricultural biotechnology. These genes can be used in environmentally friendly pest control strategies. One such use case is identifying genes with potential insecticidal properties. With an increasing number of genomic information and decreasing numbers of available annotated sequences, finding new insecticidal genes has become more challenging.The traditional methods relying on sequence alignment and annotated databases are not effective in detecting functionally relevant genes lacking close homology to known cases. This project investigates the data-driven classification of genes by sequence modeling. This research is focused on learning DNA …

Advancements In Human Action Recognition In Manual Material Handling Using Machine Learning, Jesus Jimenez, Joni Mccawley, Francis A. Mendez Mediavilla, Abhimanyu Sharotry

17th IMHRC Proceedings (Trondheim, Norway-2025)

No abstract provided.

Assessing Deep Learning Models And Hyperparameter Optimization For Stable Time-Series Electricity Load Forecasting, Sukma Patrya, Aji Prasetya Wibawa, Aripriharta Aripriharta

Knowledge Engineering and Data Science

Long-term electricity load forecasting plays an important role in ensuring system reliability, optimizing energy management, and making operational plans in face of continuously rising electricity demands. This study suggests a complete deep learning method for univariate forecasting of future electricity loads based on climatology and electricity consumption data for the period between 2019 and 2023. The initial dataset was cleaned, normalized, and partitioned chronologically into train/test datasets. Four train/test split cases (20/80, 40/60, 60/40, 80/20) were considered to explore the impact of different levels of historical data availability on the performance of the suggested framework from data-poor to data-rich situations. …

Simple Vs. Complex Human Activity Classification Via Hybrid Machine Learning Models, Anusha Kukreja

Master's Projects

In-depth understanding of the complexity of daily human activities is crucial for building responsive health monitoring and assistive technologies. However, limited research has focused on distinguishing activities based on their involvement level, as most existing work classifies only the type of activity performed. In this thesis, we address this gap by proposing a method to classify human activities as either simple or complex using sensor data from the Opportunity dataset. We define complex activities as those involving object interactions or multiple coordinated movements (e.g., drinking from a cup, cleaning a table), and simple activities as static or low-effort postures (e.g., …

Prediction Of Audit Findings Using Deep Learning With Financial And Non-Financial Data: A Case Study In Province X, Fery Yohan Setiawan, Eko Mulyanto Yuniarno, Reza Fuad Rachmadi

Knowledge Engineering and Data Science

The implementation of the audit from the local government financial statements by The Audit Board of The Republic of Indonesia (BPK RI), especially for the Province X representative, are frequently faced by the various limitations, one of them being the required audit time. At this moment, the BPK RI representative of Province X doesn’t have the tools that are able to help the accurate of sample determination for the pick test, which resulted in this study proposing the application of multi-label classification to predict the findings of financial statement (Laporan Keuangan, LK) audits based on financial and non …

Traditional Vs Deep Learning Approaches For Efficient Electricity Consumption Prediction., Lucas Sant’Ana

ICT

Accurate forecasting of electricity demand is critical for reliable energy planning, resource allocation, and policy design. Traditional statistical models, such as ARIMA, SARIMA, and ARIMAX, have been widely applied but remain constrained by linear assumptions, limited temporal memory, and inflexibility in handling multiple exogenous drivers. In this study, we systematically compare these approaches with multivariate Long Short-Term Memory (LSTM) networks, which are capable of capturing nonlinear dependencies, long-term temporal dynamics, and multivariate interactions. Historical electricity consumption data were combined with weather variables, including temperature, wind speed, and rainfall, and pre-processed through cleaning, scaling, and temporal alignment. Statistical baselines and deep …

Leaf Classification Using Convolutional Neural Networks And Vision Transformers, Louis Wilkie

ICT

This study investigates deep learning techniques for plant image classification using four publicly available datasets: Folio, Flavia, Swedish, and Sugarcane. Three modelling approaches—Convolutional Neural Networks (CNN), Vision Transformers (ViT), and hybrid CNN-ViT ensembles—are evaluated. Models leverage PyTorch implementations and pretrained weights from the Hugging Face hub, with hyperparameter tuning applied to optimize accuracy. Transfer learning with pretrained ViT models achieved the highest performance, attaining 100% accuracy on Folio, Flavia, and Swedish datasets, and 94.07% on Sugarcane. CNN and CNN-ViT models showed comparable results, but pretrained CNN-ViTs were more resource-efficient. The findings highlight the importance of evaluating multiple pretrained architectures to …

Advancing The Characterization Of Geophysical Signals Through Array Processing And Artificial Intelligence, Miro Ronac Giannone

Earth Sciences Theses and Dissertations

In geophysics, seismic and infrasound observations are routinely employed to constrain the nature and origin of events. Seismoacoustics, as a discipline, is built upon the simultaneous detection and integrated analysis of these data types. This joint approach is critical not only for advancing scientific understanding but also for supporting global monitoring efforts in hazard mitigation and nuclear explosion treaty verification. The data analyzed in this dissertation were recorded by array deployments, which consist of multiple sensors arranged in predetermined configurations to enhance signal detection, resolve directionality, and quantify waveform coherence. Leveraging these array recordings, I introduce novel approaches which combine …

Deep Target Recognition: Semi-Supervised Annotation, Sensor Fusion And Super-Resolution, Shoaib Meraj Sami

Graduate Theses, Dissertations, and Problem Reports (ETD)

Despite the recent expansion of machine learning algorithms to cover a wide range of disciplines, several areas of automatic target recognition (ATR) remain underexplored. This dissertation presents tools developed to improve performance in three significant aspects of ATR: semi-supervised annotation, sensor fusion, and image super-resolution. The aim of the semi-supervised methods is to automatically annotate targets in scenarios where labeled data are scarce in the target domain but available in the source domain. Secondly, to address the limitations of individual image sensors and enhance robustness under different environmental conditions and man-made constraints, a sensor fusion algorithm was developed to improve …

Applications Of Deep Learning For Optimizing Fingerphoto And Latent Fingerprint Biometrics, Amol Sanjay Joshi

Graduate Theses, Dissertations, and Problem Reports (ETD)

Fingerprint-based biometric recognition remains one of the most dependable and widely adopted approaches for identity verification due to its permanence and distinctiveness. Recent advancements in mobile and contactless imaging have extended fingerprint acquisition beyond controlled environments into unconstrained, real-world conditions through fingerphotos, contactless fingerprint images captured by digital or smartphone cameras. While this paradigm shift enhances accessibility and user convenience, it introduces significant technical challenges. Variations in illumination, focus, and motion blur often degrade ridge patterns, making accurate feature extraction and matching more difficult. Similarly, in forensic applications, latent fingerprints, incomplete or smudged prints lifted from surfaces, pose unique challenges …

Semi-Supervised Multimodal Multi-Instance Learning For Aortic Stenosis Diagnosis, Zhe Huang, Xiaowei Yu, Benjamin S. Wessler, Michael C. Hughes

Computer Science Faculty Research & Creative Works

Automated interpretation of ultrasound imaging of the heart (echocardiograms) could improve the detection and treatment of aortic stenosis (AS), a deadly heart disease. However, existing deep learning pipelines for assessing AS from echocardiograms have two key limitations. First, most methods rely on limited 2D cineloops, thereby ignoring widely available Spectral Doppler imaging that contains important complementary information about pressure gradients and blood flow abnormalities associated with AS. Second, obtaining labeled data is difficult. There are often far more unlabeled echocardiogram recordings available, but these remain underutilized by existing methods. To overcome these limitations, we introduce Semi-supervised Multimodal Multiple-Instance Learning (SMMIL), …

Automated Classification Of Cardiac Anomalies Through The Analysis Of Electrocardiogram Signals Using Convolutional Neural Networks., Yhosely Isset Villegas Baldiviezo

ICT

Electrocardiography (ECG) is a widely used non-invasive method for monitoring cardiac activity and detecting heart abnormalities. However, manual interpretation can be time-consuming and prone to human error, particularly for non-specialists. This study investigates the use of deep learning techniques, specifically convolutional neural networks (CNNs), to automate ECG signal classification using a large-scale dataset. A classical machine learning model is also included as a baseline for comparison. The aim is to evaluate the potential of deep learning as a clinical decision-support tool that can assist healthcare professionals in improving diagnostic accuracy and supporting early detection of cardiac conditions.

Theoretical Analysis Of Cnns For Automatic Seizure Detection In Eeg Signals, Jackson T. Small

Honors Undergraduate Theses

Epilepsy is a common brain disorder where neurons in the brain rapidly fire, causing recurring seizures. The brain activity during a seizure can be detected by electroencephalogram (EEG) signals; however, this process is not only labor-intensive and time-consuming but is also subject to inter-rater variability, with a study showing only moderate agreement when diagnosing patients, even among experts. Convolutional Neural Networks (CNNs) are often proposed to detect seizures automatically, achieving high performance. The focus on performance comes at a cost of losing interpretability, leaving the model as effective but seen as a ’black box’. This thesis confronts the interpretability knowledge …

Knowledge And Ontology Enhanced Approach To Natural Language Understanding (Koe-Nlu) In Computational Social Media And Healthcare, Naga Usha Gayathri Lokala

Theses and Dissertations

Natural Language Understanding (NLU) faces both opportunities and challenges as the amount of social media and healthcare data grows. This is particularly evident in context-sensitive applications such as evaluating cognitive health, identifying mental health symptoms, and monitoring drug abuse. Even though traditional NLU models work well for processing language in a wide range of areas, they often lack the ability to understand language in a specific domain, reason in context, and incorporate structured external knowledge. This dissertation talks about the Knowledge and Ontology Enhanced Approach to Natural Language Understanding (KOE-NLU), a new framework that is meant to make NLU systems …

Automated Treatment Planning Framework For Cervical Cancer In Brachytherapy, Suman Gautam

Theses and Dissertations

Brachytherapy (BT) is a critical treatment modality for cervical cancer that delivers ionizing radiation directly to the tumor using intracavitary (IC) and/or interstitial (IS) applicators. In high dose-rate (HDR) BT, a remote afterloader guides the radioactive source to specific dwell positions for defined durations, known as dwell times, allowing for a conformal dose to the tumor while minimizing exposure to surrounding organs at risk (OAR). However, due to the inherent nature of ionizing radiation, some dose to OARs is unavoidable, creating a fundamentally conflicting objective in treatment planning: maximizing tumor coverage while minimizing OAR dose. This characterizes the problem as …

Leveraging Distributed Semantics From Deep Learning Architectures For Literature-Based Discovery, Clint A. Cuffy

Theses and Dissertations

Literature-based discovery (LBD) is a scientific process that introduces methods to automatically identify novel insights between non-interacting sets of literature. To date, numerous statistical and machine learning-based methods have been applied in the biomedical domain to find treatments for diseases such as Raynaud's disease, Parkinson's disease, and Multiple Sclerosis. However, the lack of standardized practices and creation of bespoke methodologies produces a scenario where the adoption of LBD remains challenging in real-world systems. Our work addresses these concerns through the improvement of five critical areas: 1) error propagation within LBD's a priori dependent tasks, 2) exploring the integration of modern …

Using Deep Neural Nets In Writer Identification & Analysis, Aditya Majithia

Dissertations and Theses

This thesis focuses on developing automated deep learning methods for writer identification and writer attribute prediction from handwriting. It introduces two novel architectures: Convolutional Transformer Encoder (CTE) and Convolutional Swin Encoder (CSE). CTE is designed for determining authorship from handwritten text, be it modern or historical handwriting. CTE tracks subtle features and cues from handwriting strokes to distinguish authorship. It is the first of its kind capable to operate on historical handwritten fragments, setting it apart from existing methods that rely on entire document pages. CSE is designed to determine multiple attributes of an author such as authorship, gender, age …

Unlocking The Secrets Of High-Frequency Financial Data: Innovative Approaches To Modeling And Analysis, Lu Zhang

Graduate Research Theses & Dissertations

High-frequency trading (HFT) has emerged as a pivotal innovation in modern financial markets, characterized by rapid, data-driven decision-making processes that capitalize on granular, time-stamped market data. This dissertation examines the predictability of intraday stock price movements during the final 30 minutes of U.S. trading, employing a Bayesian regression model with Student-t error terms to address the limitations of traditional Gaussian-based methods. The analysis reveals a decline in established predictors and the emergence of new dynamics, such as post-Federal Reserve "tug-of-war" effects. The research further advances high-frequency financial modeling by introducing a comprehensive data engineering pipeline and applying cutting-edge deep learning …

Intelligent Forecasting Of Solar Atmospheric Disturbances Via Capsule Neural Networks And Space Weather Data, Eman Turki Mahdi, Mohammed E. Seno, Abdullah I. Abdulghafar

Mesopotamian Journal of Big Data

Solar flares represent a major challenge to satellite communications, navigation systems, and terrestrial power grids, making accurate forecasting essential for mitigating their disruptive effects. This study aims to improve the reliability of solar flare prediction by developing a deep learning framework based on Capsule Networks (CapsNet). The proposed approach integrates feature engineering, data preprocessing, and imbalance-handling techniques such as SMOTE and Focal Loss. Using NASA space weather data, we constructed both binary (storm/no storm) and multi-class (C, M, X) classification models across forecasting windows of 6, 24, and 48 hours. The 48-hour binary model achieved 96% accuracy with a True …