Digital Commons Network^™

Llamoco: Instruction Tuning Of Large Language Models For Optimization Code Generation, Zeyuan Ma, Yue-Jiao Gong, Hongshu Guo, Jiacheng Chen, Yining Ma, Zhiguang Cao

Research Collection School Of Computing and Information Systems

Recently, combining the strength of large language models (LLMs) and Evolutionary Computation (EC) has shown promising results for addressing optimization problems. It typically involves either iterative next-step solution seeking or directly prompting LLMs to generate critical optimization codes. However, these methods often suffer from low computational efficiency, high sensitivity to prompt design, and a lack of domain-specific knowledge. We introduce LLaMoCo, the first instruction-tuning framework designed to adapt LLMs for solving optimization problems in a code-to-code manner. LLaMoCo features a comprehensive instruction set that includes code-style problem descriptions as input prompts and robust optimization codes from expert EC optimizers as …

Realign: Text-To-Motion Generation Via Step-Aware Reward-Guided Alignment, Wanjiang Weng, Xiaofeng Tan, Junbo Wang, Guo-Sen Xie, Pan Zhou, Hongsong Wang

Research Collection School Of Computing and Information Systems

Text-to-motion generation, which synthesizes 3D human motions from text inputs, holds immense potential for applications in gaming, film, and robotics. Recently, diffusion-based methods have been shown to generate more diversity and realistic motion. However, there exists a misalignment between text and motion distributions in diffusion models, which leads to semantically inconsistent or low-quality motions. To address this limitation, we propose Reward-guided sampling Alignment (ReAlign), comprising a step-aware reward model to assess alignment quality during the denoising sampling and a reward-guided strategy that directs the diffusion process toward an optimally aligned distribution. This reward model integrates step-aware tokens and combines a …

Explainable Physics-Based Constraints On Reinforcement Learning For Accelerator Optimization, Jonathan Colen, Malachi Schram, Kishansingh Rajput, Armen Kasparian

Data Science Faculty Publications

We present a reinforcement learning (RL) framework for optimizing particle accelerator experiments that builds explainable physics-based constraints on agent behavior. The goal is to increase transparency and trust by letting users verify that the agent’s decision-making process incorporates suitable physics. Our algorithm uses a learnable surrogate function for physical observables, such as energy, and uses them to fine-tune how actions are chosen. This surrogate can be represented by a neural network or by an interpretable sparse dictionary model. We test our algorithm on a range of particle accelerator optimization environments designed to emulate the Continuous Electron Beam Accelerator Facility at …

Slm With Swarm Intelligence For Efficient Representation Of Medical Claims, Mohamed Ahmed Abo El-Enen, Ravi S. Sharma, Mustafa Abdulrazek, Amril Nazir, Reem Muhammad, Ahmed Talat Sahlol

All Works

Healthcare industry faces significant challenges due to fraudulent medical insurance claims, which result in substantial financial losses. We propose an automated system using domain-specific Small Language Models (SLMs) with a narrower scope and smaller parameter count than general-purpose Large Language Models (LLMs), combined with optimization algorithms to improve fraud detection. Our approach integrates numerical features, such as age and claim amount, with textual descriptions, including diagnoses and procedures, into a unified textual representation for each medical activity. This representation captures complex patterns, enhancing the model’s predictive ability. SLMs fine-tuned on medical corpora transform these textual inputs into fixed-dimensional numerical embeddings, …

The Deepfake Litmus Test: A Multimedia Authenticity Mechanism, Amna Alzaabi, Hessa Alqubaisi, Fatima Alzaabi, Richard Ikuesan

All Works

Deepfake technologies have made it increasingly difficult to distinguish authentic video content from manipulated media. This paper presents a forensic detection framework, referred to as the Litmus Test, which focuses on structural analysis of MP4 container files to detect signs of tampering. Unlike conventional AI-based approaches that operate as black boxes, this method examines the atomic composition of video containers to identify anomalies. The proposed method performs atom-level inspection of MP4 file hierarchies and structural markers to uncover anomalies indicative of synthetic manipulation. Evaluations using datasets such as CelebDF, UADFV, and DeeperForensics reveal that the framework can identify inconsistencies common …

Modelling Route-Level Interzonal Travel Time Using Gps Trajectories, Muhammad Faiq Ahmed, Tom Bellemans, Fatma Outay, Muhammad Ahmed, Afzal Ahmed, Feng Liu, Muhammad Adnan

All Works

Activity-based models (ABMs) require accurate travel-time estimates for accessibility calculations, yet many implementations rely on static routing outputs that fail to capture temporal congestion dynamics due to limited high-resolution data. This paper develops route-level travel-speed prediction models using GPS trajectory data from 48 vehicles in Flanders, Belgium. GPS trajectories are integrated with OpenStreetMap and land-use data through destination-based segmentation, in which trips from fixed origins are cumulatively segmented at zone crossings. To capture behavioural differences by trip length, separate Gamma regression models are estimated for short (≤5 km) and long (>5 km) trips using temporal, network, and spatial variables. …

Smart Health Care Application For Predicting Complications Risk In Type 2 Diabetes Management Using Personalized Digital Twins: A Focus On Early Intervention And Prevention Strategies, Haifaa Alkaabi, Ahed Abugabah

All Works

The study examined the application of Personalized Digital Twins (PDTs) to prevent complications during the management of Type 2 Diabetes, especially in early intervention and prevention plans. Based on a high-quality dataset related to the CDC Behavioral Risk Factor Surveillance System (BRFSS) data, we tested multiple predictive models such as the Random Forest, Gradient Boosting machines (GBM), and Extreme Gradient Boosting (XGBoost). We developed a composite risk indicator from established clinical risk factors (hypertension, dyslipidemia, elevated BMI) to stratify complication risk. The Random Forest model achieved 99% accuracy (AUC: 0.98) at the population-level risk classification. The GBM model was optimized …

Adaptive Multi-Agent Learning For Infrastructure-Aware Its: The Imer Data-Processing Approach, Mayssa Hamdani, Nafaa Jabeur, Ansar Yasar, Fatma Outay, Li Li

All Works

The performance of Intelligent Transportation Systems (ITS) critically depends on accurate and efficient road-condition monitoring. This paper presents IMER (Inspect–Map–Eliminate–Reduce), a novel AI-driven data-processing framework that extends the traditional Map-Reduce paradigm for infrastructure maintenance. IMER integrates confidence-based validation, redundancy elimination, and severity prioritization to enhance data quality and decision efficiency. Implemented within a multi-agent architecture, IMER enables autonomous agents to inspect, classify, and fuse multi-source road data in real time, supporting predictive and adaptive maintenance planning. Simulation results using augmented pothole datasets demonstrate a 39.9 % reduction in redundant reports and 39.8 % fewer false positives. These findings highlight IMER’s …

A Design Science Research Architecture For Xr-Based Pre-Visit Cultural Heritage Learning Applications, Mousa Al-Kfairy, Omar Alfandi, Saed Alrabaee

All Works

Pre-Visit preparation plays a critical role in shaping visitors’ learning and engagement in cultural heritage sites; however, existing approaches largely rely on static and passive materials that fail to foster meaningful understanding before the physical visit. Extended Reality (XR) technologies offer new opportunities to address this gap by enabling immersive, narrative-driven pre-visit learning experiences. This paper proposes a conceptual architecture for XR-based pre-visit cultural heritage learning applications, grounded in Design Science Research (DSR). Drawing on museum pedagogy, experiential learning, and XR interaction design, the study identifies key educational and technical requirements and translates them into a layered, modular system architecture. …

Mapping Llm Misuse In Computing Education: A Survey-Based Risk Analysis Of Faculty And Student Contexts, Noura Alzaabi, Mohamed El-Attar, Sarah Kohail, Mahmood Niazi

All Works

Large Language Models (LLMs) have become deeply embedded in computing higher education, yet the misuse risks they introduce for faculty and students remain insufficiently understood from a cybersecurity and data privacy perspective. This paper presents an empirical study in which a structured survey of 105 participants at a computing college was used to identify and systematically risk-score thirteen LLM misuse cases across faculty and student contexts. Using a Likelihood × Impact scoring model, the resulting taxonomy classifies misuse cases as Critical, High, or Medium severity, with over-reliance and skill atrophy, academic integrity violations, and research integrity risks emerging as the …

Evaluating Chatgpt-5 For Misuse Case Diagram Generation: An Empirical Evaluation, Alia Alzarooni, Yasser Khan, Hassan Alsayegh, Mohamed El-Attar, Rima Grati

All Works

Misuse case diagrams are a widely adopted technique in security requirements engineering, enabling analysts to model adversarial threats and derive countermeasures early in the software development lifecycle. However, manual construction of these diagrams is prone to incompleteness and subjectivity, requiring significant security expertise. Large language models (LLMs) such as ChatGPT present a promising opportunity to automate this process, yet their effectiveness for generating structured security modeling artifacts remains largely unexplored. This paper presents an exploratory study evaluating ChatGPT-5's ability to generate misuse case diagrams directly from textual security requirements, using 12 case studies of varying complexity spanning small, medium, and …

Aoi-Aware Agentic Federated Mixture-Of-Digital-Twin Experts For 6g Vehicular Edge Intelligence, Asadullah Tariq, Mohamed Adel Serhani, Ikbal Taleb, Shayma Alkobaisi, Tariq Qayyum, Irfan Ud Din

All Works

Digital twin-enabled vehicular edge intelligence is expected to become a fundamental service paradigm for sixth-generation (6G) intelligent transportation systems. However, the performance of such systems depends not only on model accuracy, but also on the freshness of digital twin states, timeliness of inference, privacy-preserving model training, and efficient use of heterogeneous edge resources. Existing DT-assisted federated learning and edge mixture-of-experts solutions optimize digital twin synchronization, distributed learning, and sparse inference largely independently, without allowing digital twin states to actively govern expert specialization, expert refreshing, and distributed orchestration. Nevertheless, the joint problem of how digital twins should guide federated expert specialization, …

Drone Authentication System Using Radio Frequency Fingerprinting, Jamila Muhsen Alnuaimi, Shamma Ghaleb Almansoori, Noura Ahmed Alrumeithi, Richard Ikuesan

All Works

The widespread integration of unmanned aerial vehicles (UAVs) across domains such as logistics, surveillance, and emergency response has introduced critical security challenges, particularly unauthorized access, identity spoofing, and drone cloning. Traditional software-based authentication methods, including GPS tracking and encryption, have proven inadequate against advanced cyber-physical threats. This paper proposes a secure and automated drone authentication framework based on Radio Frequency (RF) fingerprinting, leveraging intrinsic hardware-level signal imperfections to generate unique and unclonable drone identities. Using Random Forest classifiers, the system captures, preprocesses, and analyses RF features to distinguish between authorized and unauthorized UAVs. Validation with real-world RF datasets demonstrates high …

Human Factors In Visual Attention: Gender Differences In Engagement With Male-Oriented Ads, Mohamed Basel Almourad, Emad Bataineh, Zelal Wattar, Mohammed Hussain

All Works

In today's competitive market, it is increasingly important to understand how visual design shapes customer behaviour. This study examines the decision drivers influencing female customers' purchase choices when buying male-oriented products as gifts, identifying the visual components of advertisements that attract them by analysing the relationship between purchase intention and visual attention. Results show that participants with higher purchase intent focused more on product imagery and branding, indicating that visual appeal, perceived quality, and brand familiarity significantly guide their decisions, with brand awareness speeding up decision-making by reducing the need for repeated visual checks. Conversely, those with low purchase intent …

Towards A Context-Aware Driving Assistance System (Ca-Das): Advancing Intelligent Vehicular Safety Through Multimodal Context Integration, Fatma Outay, Siham Farrag, Anjum Zameer, Ansar Yassar

All Works

Driving-related behavioural factors are responsible for 90% of traffic collisions. The rapid growth of urbanization and the complexity of the traffic conditions demand a smart, efficient, and flexible transportation system. The advancement of transportation through technologies such as the Internet of Things (IoT) and AI have reshaped the way that drivers interact with their vehicles and the surrounding environment. In this paper, we propose a comprehensive Context-Aware Driving Assistance System (CA-DAS) that employs sensor fusion, semantic context modelling, along with a machine-learning-based approach to provide personalised and proactive driving assistance across dynamic scenarios. The proposed CA-ADS was developed using a …

From Image To Insight: Evaluating Llm Accuracy In Understanding Uml Use Case Diagrams With Claude, Mohamed El-Attar, Yasser Khan, Mahmood Niazi, Sajjad Mahmood, Mohammad Alshayeb

All Works

UML use case diagrams are a prominent artefact of requirements engineering, capturing the functional scope of a software system in terms of actors, use cases, and their stereotyped relationships. The emergence of multimodal large language models with image understanding capabilities raises the question of whether such models can reliably extract structured construct-level information from use case diagram images. This paper reports an empirical evaluation of Claude on the task of counting 14 notational construct types from a corpus of 78 computer-generated UML use case diagrams, assessed against manually verified ground truth annotations. Results reveal a strongly differentiated accuracy profile: Claude …

Pxaa: A Root-Cause Decision-Support System For Airline Passenger Complaints Using Sentiment Analysis And Complaint Classification, Adélia Acacio Da Silva, Stefani Rabelo De Campos Nunes

ICT

PXAA (Passenger Experience Analytics Assistant) is a machine learning and decision-support prototype developed to analyse airline passenger complaints using sentiment analysis and complaint classification techniques. The project follows the CRISP-DM methodology and uses airline passenger reviews collected from AirlineQuality.com through Python web scraping. Natural language processing (NLP) techniques were applied to clean and preprocess the textual data, while logistic Regression was used to classify passenger sentiment into positive, negative and neutral categories.

The project also implemented rule-based complaint classification to identify common complaint themes such as delays, baggage issues, customer service problems, comfort issues, payments/refund complaints and communication failures. These …

Design And Evaluation Of A Supervised Machine Learning-Based Intrusion Detection System., Tarquin Qazi

ICT

This project involves designing and evaluating a supervised machine-learning-based intrusion detection system (IDS) using the CIC-IDS2017 dataset. The project is structured around the cross-industry standard process for data mining (CRISP-DM) framework. The dataset comprises over 2.8 million labelled instances and was cleaned and reduced from 78 to 63 features via redundancy analysis. These data were used to train and tune three supervised classifiers: Logistic Regression algorithm as a linear baseline, a Decision Tree algorithm and a Random Forest algorithm. Hyperparameter optimisation was conducted using RandomizedSearchCV with 3-fold cross-validation, optimising for the F1-macro scoring metric.

While the random forest performed similarly …

Comparative Economic Impact Modelling Of Floods Vs Drought On A Global Scale, Diogo Lemos Aguiar, Paulo Henrique Oliveira Machado

ICT

Climate change is generating unprecedented economic losses through extreme weather events, yet comparative predictive analysis between different hazard types remains limited. This report presents a Machine Learning–Based Comparative Economic Impact Modelling System designed to predict and compare the economic damages caused by floods and droughts on a global scale. The system integrates historical disaster records from the EM-DAT International Disaster Database with country-level socioeconomic indicators from the World Bank's World Development Indicators, applying supervised regression modelling to a unified dataset of 1,301 disaster events across 133 countries from 2000 to 2025. XGBoost and a Multi-Layer Perceptron Neural Network were evaluated …

Predicting Co₂ Emissions From Cattle-Driven Deforestation In The Amazon Rainforest Using Machine Learning, Daniel Ambrosio, Suelen Rocha

ICT

Deforestation has become a major environmental issue worldwide, especially in the Amazon Rainforest, which is known as one of the world’s largest carbon sinks. This region has been significantly impacted by cattle farming, contributing to increased CO₂ emissions and land degradation. As Brazil is one of the world’s largest beef producers and exporters, the environmental impact of livestock production has attracted increasing attention from organisations, governments, and sustainability analysts.

This project aims to analyse the relationship between cattle farming, deforestation, and CO₂ emissions in the Amazon using Machine Learning. Using the CRISP-DM framework, the project explored each stage of the …

Healthcare Digital Twins: A Methodological Literature Review On Integrating Iot And Ai For Personalized Medicine And Predictive Care, Sara Shahnazinia, Mahsa Tavasoli, Abdolhossein Sarrafzadeh, Ali Karimoddini

Electrical & Computer Engineering Faculty Publications

Digital Twin (DT) technology has the potential to revolutionize healthcare delivery and enhance patient outcomes through personalized and precision medicine, simulation models for operations and interventions, and drug discovery. However, successful implementation of DTs in Internet of Things (IoT) and artificial intelligence (AI) healthcare is contingent upon addressing key challenges such as privacy, ethics, and robust data security. This paper presents a methodological literature review of DT applications in healthcare, systematically analyzing the current state of research, key enabling technologies, and implementation challenges. The review summarizes DT categorization approaches (application-based, technology-based, and real-time function-based); delineates core DT components such as …

Gem-Can: A Real-World Dataset Of Can-Bus Attack Scenarios On An Autonomous Vehicle For Intrusion-Detection Research, Mahsa Tavasoli, Abdolhossein Sarrafzadeh, Ali Karimoddini, Tienake Phuapaiboon, Milad Khaleghi, Daniel Tobias

Electrical & Computer Engineering Faculty Publications

This paper presents GEM-CAN, a labelled Controller Area Network (CAN) dataset captured from an autonomous GEM e6 platform under both normal operation and controlled cyber-attack conditions.

The dataset contains ∼143 K frames comprising (i) ∼ nominal autonomous operation (∼100k messages), (ii) DoS floods using arbitration ID 0 × 00000000 (∼41 K messages), and (iii) data-tampering injections that reuse legitimate IDs for brake and steering-lock (∼1.3 K messages). Each record includes timestamp, arbitration ID (11/29-bit), DLC, eight payload bytes, and a Normal/Attack label. A companion metadata file enumerates attack windows, PCAN bus-load traces, bitrate, and test conditions. Data were collected with …

Flair- Flow-Level Anomaly Intrustion Recognition, Joseph Dumond

2026 Research Poster Competition

Industrial control systems (ICS) and industrial Internet of Things (IIoT) networks support critical infrastructure such as power generation, water treatment, and manufacturing. As these systems become more connected, they are increasingly exposed to cyber threats. However, many existing intrusion detection approaches rely on detailed packet inspection or supervised machine learning techniques that require labeled attack data and extensive tuning, making them difficult to deploy in real industrial environments.ICS networks typically generate highly regular and predictable communication patterns due to periodic control logic and deterministic device behavior. Because of this structure, abnormal or malicious activity often appears as a disruption to …

When Relevant Isn’T Readable: Evaluating Cognitive Accessibility In Search, Kaija Frierson

2026 Research Poster Competition

According to the World Health Organization, more than one billion people worldwide experience some form of disability. Many of them may affect cognitive processing and reading comprehension rather than vision or mobility, which creates barriers when accessing written information online. Web search is a critical modern literacy skill used for learning, decision-making, and everyday problem solving, yet most search systems are designed with the assumption that all users can process complex written language equally. As a result, people with dyslexia, other cognitive disabilities, low literacy levels, or who are English language learners may struggle to access relevant information even when …

P4 Driven Data Plane Analytics For Industrial Control Network Security, Haden Fowler

2026 Research Poster Competition

Industrial control systems manage critical infrastructure such as power grids, water treatment plants, and manufacturing facilities. These systems rely on specialized network protocols to send measurements and commands between sensors, controllers, and operator workstations. Protecting these networks from cyberattacks is essential because a successful intrusion could disrupt services that millions of people depend on daily. Most current security monitoring approaches analyze copies of network traffic after it has already passed through the system. This delay means that malicious commands may reach their targets before any alarm is raised. More importantly, when alerts do occur, operators often lack the evidence needed …

Solidfit: A Decentralized Mobile Health-Tracking App Using Personal Online Data Stores (Pods), Evan Meyers

2026 Research Poster Competition

Personal fitness data is increasingly collected through mobile applications, yet most current systems store this information in centralized platforms that limit user control, long-term ownership, and data portability. This project addresses the question of whether a decentralized data model can be used to support everyday fitness tracking while preserving user autonomy and privacy. As individuals become more aware of how their health data is collected and shared, there is a growing need for systems that allow users to retain ownership of their personal information without sacrificing usability or functionality.This research presents SolidFit, an Android-based fitness tracking application designed around the …

Artificial Elections: How Artificial Intelligence Can Affect Elections And Decide Our Future, Ryan D. Granger

Honors Undergraduate Theses

This thesis examines the growing role of artificial intelligence (AI) in democratic elections, highlighting both its transformative potential and its associated risks. Drawing on a qualitative analysis of existing literature, the study explores how AI is increasingly integrated into political campaigns, election administration, and voter engagement. Key benefits include enhanced data analysis, personalized political messaging, and improved efficiency in campaign operations. AI also supports real-time fact-checking and more accurate vote tabulation, which can strengthen transparency and trust in electoral processes. However, the thesis emphasizes that these advantages are accompanied by significant challenges. AI technologies enable the rapid creation and dissemination …

Modeling The Impact Of Projected Water Demand On Groundwater Flow And Transport In The Mahomet Aquifer, Christabel Chinonye Abugu

Theses and Dissertations

The Mahomet Aquifer is a vital groundwater resource in east-central Illinois, supporting municipal, agricultural, and industrial water use. With projected changes in water demand driven by population growth and future socioeconomic conditions, assessing the long-term response of the aquifer is critical for sustainable groundwater management. Increased pumping to meet future demand may contribute to groundwater-level decline, intensified cones of depression, and altered groundwater-flow paths relevant to contaminant migration. This study addressed two research questions: (1) how projected changes in groundwater demand affect groundwater levels in the Mahomet Aquifer, and (2) how projected demand influences groundwater-flow paths relevant to contaminant migration. …

Challenges And Applications Of Fine-Grained Temporal Action Understanding: Modeling Temporal Granularity And Data Efficiency, Halil I. Helvaci

Theses and Dissertations--Electrical and Computer Engineering

Fine-grained Temporal Action Segmentation (TAS) has become a cornerstone of video understanding, offering dense frame-level predictions essential for clinical assessment, surgical skill evaluation, and human-computer interaction. While TAS methods have delivered strong results on coarse-grained benchmarks, two fundamental challenges persist: (1) global attention mechanisms dilute boundary information critical for subsecond precision, a phenomenon we term the temporal granularity bottleneck, and (2) dense frame-level annotation remains prohibitively expensive, with most datasets requiring exhaustive labeling of lengthy untrimmed videos. These challenges are particularly pronounced in medical domains, where sub-second primitives define clinical outcomes while expert annotation remains scarce. In this dissertation, we …

Generative Ai Peer Tutoring To Support Peer-Reviewed Source Identification And Evaluation, Rae Mair, Michelle Kelley, Taylar Wenzel, Andrea C. Borowczak, Yuqing Li, Mike Borowczak

Faculty Scholarship and Creative Works

This chapter explores the design and evaluation of a generative artificial intelligence peer tutor prompt to support college students in identifying and evaluating peer-reviewed sources for academic research. Grounded in literature on peer tutoring, Socratic dialogue, and AI-supported learning, the authors describe an iterative prompt engineering process designed to transform large language models (LLMs) into Socratic-style peer tutors capable of scaffolding student reasoning without completing tasks for them. Five guiding criteria for an effective peer tutor shaped development and evaluation: cognitive congruence, step-by-step guidance, avoiding giving answers, adaptability to student level, metacognitive transparency, and following assignment directions. Across multiple human-centered …