Authors - Febrian Nasrullah, Abdul Mukti Soma Abstract - This study examines the impact of risk perception, financial selfefficacy, and financial literacy on the actual usage behavior and intention to use "Buy Now Pay Later" (BNPL) services among Generation Z in Indonesia. Adopting a quantitative approach, the study surveyed 385 Gen Z individuals who use BNPL services such as Kredivo or Akulaku. Data were analyzed using SEM-PLS with the aid of SmartPLS 4. The results indicate that financial selfefficacy and financial literacy contribute to actual usage behavior and intention, whereas risk perception has a negative impact on both. Furthermore, intention contributes to actual usage behavior and mediates the effects of the other variables. These findings indicate that financial management skills, risk perception, and individual confidence in financial capability play a pivotal role in shaping BNPL usage behavior among Generation Z in Indonesia.
Authors - Byron Albuja-Sanchez, Miguel Angel Lema Carrera, Luis Antonio Ortiz Parra Abstract - This study focuses on evaluating the capabilities of different large language models chatbots in the task of designing a PID controller for a third-order transfer function of a real-world vehicle’s cruise control system. Chatbots received a detailed prompt containing the system’s transfer function and the design’s goals in the form of overshoot and settling time constraints. Chatbots only received simulation-response information as feedback during the tuning process to test their predisposition to fix the errors without being specifically asked to do so. Results showed that chatbots have a good level of knowledge regarding basic control theory and basic tuning methods for PID controllers. Preferred tuning methods involved pole placement with dominant second order dynamics, Ziegler-Nichols and heuristic methodologies. Simulation results compared the controllers designed by chatbots with a PID tuned with ant lion optimizer algorithm, none of the evaluated chatbots outperformed the optimization-based benchmark controller. However, Gemini 3 Flash designed a controller which performance was close to the ant lion optimizer results. Chatbots’ underperformance was attributed to the following facts: no expert feedback was given to them to fix the observed flaws in the proposed designs, no specific methodologies were asked to be used in order to improve the results, and no specific instructions to redesign the controllers were given to chatbots in order to test their disposition to fix their errors. Results suggest that LLMs can assist in preliminary controller design tasks, although their effectiveness remains limited without expert-guided iteration and explicit optimization-oriented prompting.
Authors - Aman Kumar, Kathan Nitin Patel, Aviral Sharma Abstract - Diabetes mellitus is perceived as a disease that significantly impacts a nation’s social, human, and financial expenditures. Concurrently, it is imperative to lower the prevalence rate and address the misunderstandings surrounding diabetes. An improved model that employs machine learning techniques to identify the behavior of diabetes in an individual. We have employed the parameters observed in the typical lifestyle, as well as the individual's emotional states and physical activities in the elderly age group. For a variety of test parameters, the proposed model implements a network classifier. It has been noted that this methodology yields effective results in the diagnosis of diabetes mellitus when the appropriate dataset is provided. The dataset utilized in this reseacrh study is the Indian PIMA dataset from the UCI Machine learning database. The detection of diabetes is contingent upon the presence of eight features in this dataset. The proposed Machine learning model has been implemented using a multilayer neural network that has been trained on backpropagation and feed-forward network simulation.
Authors - Nallappagari Venkatarami Reddy Abstract - Business Process Automation (BPA) has become an essential requirement of modern enterprise environments owing to the need for operational efficiency, process agility, and smart decision making. Traditional methods of BPA mostly depend on rules-based approaches which are not able to adapt to the needs of a dynamic environment as these methods do not incorporate the element of adaptive intelligence and autonomous orchestration. To overcome such limitations, this study aims to develop an intelligent orchestration framework named IWOF for Autonomous Business Process Automation. In the proposed solution, PIEL, AWOE, DRAM, and PDOU have been used. Also, two new algorithms named AWIO and PARDO are developed for optimizing the process sequencing, resource assignment, and decision support tasks respectively. Experimental evaluation was performed by applying the proposed framework to datasets consisting of processes related to employee onboarding, payroll management, procurement approvals, recruitment workflow, and finance transactions. With the help of the IWOF model, Process Automation Accuracy, Workflow Completion Rate, Resource Utilization Efficiency, and Autonomous Business Process Automation Score (ABPAS) were measured to be 98.7%, 98.2%, 97.1%, and 98.9%, respectively, surpassing all other available models such as OSMAS and BPA-SME.
Authors - Abeer Tag, Tahani Almarri, Abir Sidilemine, Rowaa Khaled, Loay Ismail Abstract - Medication non-adherence among elderly and chronically ill patients remains a critical global health challenge, leading to severe complications, hospital readmissions, and reduced quality of life. This paper presents REMEDI, a smart mobile medication dispensing robot that integrates autonomous indoor navigation, biometric patient authentication, automated pill dispensing, pill verification, and real-time adherence monitoring into a unified platform. The system combines a TurtleBot3 Waffle Pi mobile base with a custom-designed three-cylinder dispensing mechanism controlled using Raspberry Pi 5 and Arduino Nano. Patient verification is performed using facial recognition with MobileFaceNet embeddings and liveness detection, achieving an overall verification accuracy of 83.3% and zero false accepts during experimental testing. Autonomous navigation is implemented using LiDAR-based SLAM and A* path planning, enabling map-based movement between predefined indoor patient locations. Post-dispensing verification uses a custom-trained YOLOv11 object detection model integrated with OpenCV for pill detection and counting. A companion Android application allows caregivers to enroll patients, schedule medications, and monitor adherence in real time. Experimental results show successful integrated operation, including dispensing delays below 5 seconds, navigation success rates of 84–92%, 95% dispensing reliability, and functional multi-patient queue management. Although pill verification achieved only 69% real-world accuracy, the results demonstrate the feasibility of integrating mobility, secure authentication, dispensing, and monitoring in one user-centered prototype. REMEDI aims to bridge the gap between stationary home medication dispensers and large institutional delivery robots.
Authors - Simon Kloker, Alex Cedric Luyima, Matthew Bazanya Abstract - This paper presents WASHtsApp, a WhatsApp-based mHealth chatbot that supports clean water, sanitation, and hygiene (WASH) education in rural African settings. The chatbot uses Retrieval-Augmented Generation (RAG) to reduce out-of-context responses and improve answer relevance. Following a Design Science Research approach, we evaluated the artifact in two steps: expert content validation (four WASH experts) and community acceptance validation (n = 71). Expert ratings classified 86% of responses as perfect or sufficient, while community results showed high perceived usefulness, ease of use, and intention to use. The findings indicate that WhatsApp is a viable delivery channel for WASH education and that a constrained RAG setup can provide useful localized guidance. We also discuss privacy, safety, and future improvements, including local-language support.
Authors - K. N. Subramanya, Padmashree T., Manojith Bhat V., Manasvini G. Padmasali Abstract - In an era of rising digital dependence and ever-evolving cyber threats, cybersecurity has become a concern for education institutions of all sizes. Higher education institutions (HEIs) are prime targets for cyberattacks since they manage massive volumes of sensitive data related to students, faculty, and research. Protecting this data is important to avoid major consequences such as disruptions in academic services, reputational harm, legal or financial ramifications. This survey consolidates current research on cybersecurity practices in HEIs, analyzes critical digital assets and infrastructure. It also reviews selected cybersecurity frameworks adopted globally. The survey further explorers the threat landscape confronting HEIs, examining various cyberattacks by identifying possible entry points, attack pathways, and potential consequences. The study emphasizes the necessity of adaptive cybersecurity approaches that can evolve alongside emerging technologies and pedagogical models in academia.
Authors - Roberts Dargis, Arturs Znotins, Ilze Auzina, Maris Golubovskis, Mikelis Gulbis, Normunds Gruzitis Abstract - Operational radio communication is a challenging application domain for automatic speech recognition (ASR) despite advances in multilingual foundation models. We investigate the applicability of modern Latvian ASR models to operational communication and evaluate whether speech enhancement techniques improve recognition quality under realistic conditions. To support the study, we created a specialized corpus of authentic Latvian operational radio communication. The corpus captures acoustic and linguistic phenomena largely absent from general speech corpora, including narrow-band transmission, radio-channel artifacts, environmental noise, domain-specific terms, and fragmented utterances. Using this corpus, we evaluate state-of-the-art adaptations of the massively multilingual Whisper and MMS models in combination with several audio preprocessing methods. The results reveal a substantial performance gap between the conventional Latvian ASR benchmarks and operational communication data. While some preprocessing methods improve perceived audio quality, they provide limited benefit for downstream recognition and often even degrade ASR performance. Voice activity detection, however, yields the most consistent improvements. The findings indicate that domain mismatch, rather than acoustic degradation alone, is the dominant source of recognition errors and highlight the need for representative domain-specific data when adapting general-purpose ASR models for the operational communication environment.
Authors - Mark Fedorchenko, Olena Kopishynska, Yurii Utkin, Igor Sliusar, Leonid Flehantov, Olha Barabolia, Nadiia Protas, Tetiana Dugar Abstract - Crop yield forecasting based on small official statistics is different from forecasting with dense satellite, field, or weather datasets: the sample is short, temporal leakage is easy to introduce, and machine learning (ML) should not be accepted unless it beats transparent baselines. This paper presents a baseline-first and reliability-aware workflow for farm management and regional advisory systems. Wheat, maize, and sunflower are evaluated for Poltava, Vinnytsia, Cherkasy, and national-level Ukraine data for 2010-2024. ElasticNet, XGBoost, and LightGBM are compared with naive lag-1, linear-trend, LINEST, and Autoregressive Integrated Moving Average (ARIMA) baselines under a forward temporal design. The contribution is a decision layer that recommends ML only after it clears a practical mean absolute error (MAE) margin and then reports empirical validation-residual bands, test coverage, feature-group diagnostics, and compact farm management systems (FMS)-compatible forecast cards. The Poltava workflow recommends FORECAST.LINEAR for wheat (MAE 0.49 t/ha), LightGBM for maize (MAE 0.69 t/ha), and LightGBM for sunflower (MAE 0.04 t/ha). Across the external check, ML is recommended in 7 of 12 region-crop cases. The results show that ML can help in small official-statistics settings only when checked against simple baselines and reported with reliability diagnostics.
Authors - Ingy Emara, Rawan Waleed, Sherry Emad Abstract - This study analyses speech errors in individuals with Down syndrome (DS) in both Arabic and English, with a particular focus on errors that reduce intelligibility for automatic speech recognition (ASR) systems. It compares the speech errors that most significantly affect intelligibility in each language and investigates the factors underlying differences in ASR accuracy across Arabic and English DS speech. The findings indicate that ASR systems perform less accurately with Arabic DS speech, highlighting the need for larger and more diverse Arabic DS speech datasets for system training. The study also identifies several physiologically related speech errors that negatively affect intelligibility in both languages, including devoicing of stop consonants, lateralization of /r/, reduced pressure in /s/, and deletion of consonants and consonant clusters. In addition, certain errors were found to be language specific, such as the mispronunciation of uvular and pharyngeal sounds in Arabic and the frequent omission of /r/ and vowel centralization in English. These findings have important implications for speech therapy by identifying priority areas for intervention, and for the ASR industry by emphasizing the need to expand labelled DS speech datasets across languages to improve recognition accuracy.
Authors - Hiep Nghia Phan, Duy Nguyen Ngoc Abstract - Low-Rank Adaptation (LoRA) is widely used for parameter-efficient fine-tuning of large language models because it is lightweight, modular, and easy to distribute. However, the growing practice of sharing third-party LoRA modules also creates security concerns. A malicious adapter can introduce hidden behaviors, backdoors, or other risks while appearing to function normally. Existing research has largely focused on detecting whether a LoRA module is malicious. In practice, deployment decisions often require a more nuanced assessment of risk. This paper presents a measurable framework that evaluates LoRA security across four dimensions: supply-chain integrity, static weight characteristics, dynamic behavior, and deployment-time observations. The resulting indicators are normalized and combined into a composite risk score that supports comparison and prioritization of LoRA modules. The framework was evaluated using benign and backdoored LoRA modules attached to a frozen base language model. The results show a clear separation between the two groups even when their task performance remains similar. Dynamic behavioral testing and static weight analysis contribute the most useful signals, while deployment-time monitoring provides additional evidence of long-term operational risk. The proposed framework provides a practical mechanism for integrating security assessment into LoRA selection, governance, and deployment workflows.
Authors - Mayen Ben-Koko, Emmanuel Waribo Otiti Abstract - Nigeria loses more new-borns in the first month of life than almost any other country in the world, yet no machine learning tool has been built specifically for this context. This paper proposes a framework for predicting neonatal mortality risk in Nigeria using indicators from the 2023–24 Nigeria Demographic and Health Survey — the most current national health dataset available. Five classification algorithms are compared: Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Machine. Random Forest performed best, with an AUC-ROC of 0.89. The three strongest predictors were whether a skilled health worker attended the birth, the gap between pregnancies, and the number of antenatal visits. The framework is reproducible and designed to be extended as fuller microdata becomes available or adapted for routine clinic records across Nigeria's six geopolitical zones.
Authors - Rui Liu, Neng Zeng Abstract - The Bank for International Settlements' Project Leap Phase 2 trial demonstrated that post-quantum cryptography (PQC) can be functionally integrated into the Eurozone T2 real-time gross settlement (RTGS) system, reporting an average PQC signature verification time of ≈209.9 ms against ≈28.1 ms for the traditional baseline. The report, however, explicitly leaves two questions open for "future testing phases": how the observed timing translates into an SLA-aware deployment plan, and how the system should be architected for a migration-safe transition that NIST IR 8547 recommends but Leap did not test (hybrid signature, hybrid KEM, and a non-modifying deployment path on top of the existing ESMIG/NSP stack). This paper contributes the analytical answer to the first question and audits Leap's framework for the second. We adapt the classical TCP-style timeout bound to a closed-form Watchdog inequality Ttimeout ≥E[Tcompute] + 2 · TRTT + k · σjitter, derive a sensitivity table that maps the safety multiplier k to four financial-grade SLA tiers, and a closed-form capacity-planning bound whose ratio between asynchronous and synchronous throughput is parametric in the FPGA parallelism. We then audit the Leap report against its own admitted limitations on hybrid signature testing, hybrid KEM, and the "modified ESMIG connector" workaround, and we attach to each gap a bounded fix direction expressed entirely within the cost-model envelope. We complement the analysis with a formal EUF-CMA reduction sketch for the nested PQC–RSA signature (degrading gracefully under attacks on either layer) and a production-grade C empirical anchor on a single self-contained library (the LK LEGO PQC platform, native RSA, no OpenSSL): on a commodity x86-64 cloud VM, Dilithium-5 verify takes ≈0.71 ms at the median (σjitter ≈130 µs), the nested PQC–RSA verify 0.77 ms, and the hybrid ML-KEM-768 + RSA-2048-OAEP decapsulation 1.66 ms. These refine Leap's 209.9 ms PQC verify into a fast cryptographic core (0.34%) plus a slow protocol envelope (99.66%) and show both untested hybrid constructions fit inside a single-millisecond budget; a throughput cross-check (2115 verify/s single-core, 83% scaling at 2 threads) validates the independent-cycles assumption. The measured software-only path already meets the legacy 500 ms ceiling, so FPGA acceleration is an optional optimisation rather than a requirement. Production-grade RTGS measurements with HSM transport and full ISO 20022 parsing remain future work.
Authors - Shakeeb Abdullah, Jim Hjartarson, Rony E. Amaya Abstract - Currently the fastest tried, tested, and reliable electro-optical transceivers operate at 56 Gbaud (56 Gbit/s NRZ or 112 Gbit/s PAM-4) communication speeds per lane, and while some companies have finally started rolling out their 112 Gbaud (224 Gbit/s PAM-4) line of products for commercial use, not all companies have caught up; nor is their much research presented on 112 Gbaud TIAs. Higher speeds such as 400 Gbit/s are usually obtained by transmitting data through multiple parallel lanes of 100 Gbit/s PAM-4. The optical industry has been pushing for the next generation of 112 Gbaud (112 Gbit/s NRZ or 224 Gbit/s PAM4) rates for devices, however, it had stalled (hitting many roadblocks) in the past six years or so. Currently, the popularized TIA architectures are producing diminishing returns in terms of their bandwidth performance for every incremental improvement in its designs (or heavily relying on DSP); for this reason, a new paradigm construct is required to overcome such obstacles and meet the new standards of the next generations of TIAs. This brief proposes a different approach in designing TIAs for 112 Gbaud speeds or higher. Estimated criterion dictates a 3-dB BW of 78.4 GHz to process clean eyes at 112 Gbaud. Proposed TIA architecture in this paper utilizes distributed approach instead of usual common gate or feedback mode to convert the incoming photodiode current into an electrical voltage. Post-layout electromagnetic simulations show that these amplifiers can process PAM-4 eyes with 40 mV pk-pk outputs at -3 dBm of input optical power.
Authors - Nirmal Prabhu K, Sisira M S, Kanagaraj S, Kirthika P, Ashwin C Abstract - The proliferation of information and communication technologies has brought ICTs to the forefront as important enablers of social inclusion, governance, and rural development. However, the existence of inequalities over the years concerning the use, benefit, and application of ICTs, among others, has brought about a noticeable digital divide, especially among emerging nations such as India. This paper presents the underlying causes for the existence of the digital divide among rural populations in India as part of a comprehensive research review on the matter, adopting the Resource and Appropriation Theory by Van Dijk. A structured search of studies was conducted using the Scopus database, identifying articles related to rural India, both quantitative and qualitative studies. Findings indicate that the problem of the digital divide is not solely found among the rural populations of India, which lack the necessary infrastructure, as it is also ingrained among the socio-economic, geographical, and linguistic population groups, among others of India, including women as the most vulnerable section of society. The study concludes comprehensive multidimensional approach is required to bridge the digital divide.
Authors - Sowmyashree N, Madhu Sunkanur, Impana M, Suchithra B S, Hemalatha P G Abstract - This paper will outline a cold storage technique that utilizes solar power for the conservation of agricultural produce in rural and non-grid connected areas. This system includes a solar photovoltaic cell together with a backup battery that is used to guarantee continuity of electricity supply. The charge controller manages the electrical input into the circuit. The cooling process is done using a TEC-12706 Peltier unit controlled through an Arduino Uno microcontroller. Sensors are employed to monitor the temperature and voltage. The collected data is fed to the LCD screen, while good insulation ensures that cool temperatures are maintained. A performance assessment has been conducted on the proposed design, which proved its efficacy in lowering dependence on traditional energy resources while maintaining the consistency of cooling efficiency. Implementation of the suggested technology will help reduce losses from post-harvests, boost the financial state of farmers, and promote the adoption of renewable energy technologies. Additionally, the design will enhance environmental sustainability by minimizing greenhouse fuel emissions
Authors - Francka Sakti Lee, Christian Haposan Pangaribuan, Liem Bambang Sugiyanto, Sulistyowati, Jovann Kurniawan, Henry Nugraha Abstract - Voluntary employee attrition presents a systemic challenge to organizational stability, yet predictive modeling is frequently constrained by the accuracy paradox and algorithmic opacity. This study proposes an Explainable Artificial Intelligence (XAI) framework integrating eXtreme Gradient Boosting (XGBoost) with Shapley Additive exPlanations (SHAP) to transform attrition analysis into prescriptive intelligence. By implementing a Random Over-Sampling (ROS) protocol, the model successfully neutralized extreme class imbalances, significantly enhancing the detection sensitivity of latent resignation signals. The novelty of this research lies in its SHAP-driven demographic bifurcation, exposing critical asymmetries in risk triggers between young and senior employees. Empirical findings identify a "Burnout Triad" comprising compensation, overtime, and job satisfaction. Crucially, junior cohorts exhibit hypersensitivity to immediate transactional factors, whereas senior cohorts are driven by intrinsic role actualization. This framework culminates in a Decision Support System (DSS) enabling surgical retention interventions, shifting human capital management toward strategic, evidence-based governance.
Authors - Giordana Castelli, Ida Giulia Presta, Marialucia Camardelli, Mariagiulia Di Lizia, Davide Donato Russo, Giovanni Felici Abstract - Contemporary cities are increasingly shaped by climate change, digital transformation, demographic growth, socio-economic inequalities, and environmental uncertainty. These transformations challenge traditional static planning approaches and require new governance paradigms capable of dynamically addressing urban complexity. This paper discusses the Urban Intelligence paradigm as an integrated framework for adaptive and cognitive urban governance. Within this framework, an Urban Digital Twin is not just as a digital replica of the city, but a cognitive infrastructure capable of integrating datasets, simulation models, real-time monitoring systems, and participatory processes into a unified environment for knowledge production and decision-making. We explore the different technical and multidisciplinary challenges that derive from this approach, with special focus on Information and Communication Technologies that enable the effective realization of Urban Intelligent Systems, providing examples of current work in Italian Cities. We conclude by presenting the 4C Model, a conceptual model for Urban Governance designed to support the path towards resilient and cognitively enabled cities that learn from uncertainty and promote sustainability, inclusion, transparency, and collective well-being.
Authors - Janset Shawash, Henri Liu, Alicia Sudlerd, Leevi Rantala Abstract - Simulation-based learning gives healthcare students safe, repeatable practice before clinical placement; virtual reality (VR) makes it more accessible and affordable. This paper presents Aino, an artificial-intelligence-driven virtual patient for an occupational therapy (OT) home-visit showering assessment, built in Unity for the standalone Meta Quest 3. Where most of existing OT VR tools rely on fixed-viewpoint, pre-recorded 360-degree branching video, Aino is a conversational 3D patient whom students address in unconstrained natural speech (Finnish or English) while moving freely within a single continuous scene that follows a dynamic clinical narrative. The technical core is a hybrid control architecture that decouples a scripted, data-driven clinical narrative from free-form conversational responses: a section-based state machine runs twenty-nine designer-authored sections, each with an explicit completion contract that reconciles deterministic clinical progression with variable-length AI dialogue, behind an AI-provider-agnostic interface. OT educators use observation-based scenarios, and thus, the patient narrates her actions and reactions to keep her performance legible; this narration pattern and its calibration are examined as a transferable design lesson. The showering task additionally involves intimate personal care that cannot be ethically rehearsed in live role-play yet is staged safely in VR. Formative findings from educator co-design, educator try-out sessions, and play-testing are reported, and planned student evaluations are outlined.
Authors - Gilmara Santos, Pedro V. Matias, Yan W. Martins, Jose R. Santos Junior, Joao V. Fernandes, Rodrigo O. Jesus, Ueller B. Silva, Lidia Roque, Klinsman Goncalves, Laisa Paiva, Edjair Mota Abstract - The Amazon River basin, home to one of the world’s largest freshwater reserves and unparalleled biodiversity, silently suffers from a vast environmental disaster caused by illegal mining, during which mercury is discharged into its waters. This contamination threatens aquatic ecosystems and poses serious risks to highly vulnerable populations. In response, this paper provides clues for a resilient and scalable system architecture for real-time water quality monitoring, tailored to the environmental and infrastructural challenges of the Amazon region. A detailed systematic literature review assesses state-of-the-art Internet of Things (IoT)-based monitoring techniques, focusing on key variables such as mercury concentration, temperature, turbidity, and pH. Special emphasis is given on communication technologies suitable for diverse settings—from Wi-Fi-enabled urban areas to remote rainforest regions where LoRa, NB-IoT, and WiLD (Wi-Fi over Long Distance) present viable alternatives. The study also highlights the role of the application layer in enabling data analysis, real-time alerts, and remote visualization of environmental conditions. This research contributes to developing time-efficient and sustainable monitoring strategies to support public health initiatives and ecological conservation by bridging technological innovation with the urgent environmental needs of one of the planet’s most critical biomes.
Authors - Laura Alma Diaz-Torres, Alma Delia Torres-Rivera, Mario Leonardo Nieto Antolinez, Fabian Leonardo Alfonso Sabogal Abstract - Mexico City faces a constant need for high-quality public transport systems capable of reducing passenger waiting times, improving travel comfort, and maintaining the economic viability of private operators. In this context, demand studies are essential both before the concession stage and during service operation, since they support route planning, fleet allocation, schedule adjustments, and operational decision-making. Two similar but distinct methodologies are compared for the estimation of load polygons. The first methodology assigns telemetry events to official stops using spatial proximity and route reconstruction through directed graphs. This approach provides higher operational traceability, since demand is linked to formal routes, directions, and stops. Nevertheless, it may underestimate demand that occurs outside the official route structure. The second methodology uses heat maps and 300-meter-radius polygons to identify functional demand areas based on observed passenger activity. This approach captures real operational behaviour more flexibly, but may lose direct correspondence with formal stops, especially when polygons overlap or include stops from different directions. The comparison shows that neither methodology is sufficient on its own. The graph-based method is useful for formal operational analysis, while the heat-map method is more sensitive to actual demand behaviour. Based on these findings, the paper proposes, as future work, the development of a multicriteria integration approach that combines both methods. Such an approach could reduce structural and observational biases, improve the processing of boarding and alighting data, and generate clearer maps, graphics, and analytical outputs to support expert decision-making in public transport operations.
Authors - Sayee Patil, Vaidehi Pathak, Purva Nalawade, Rupali Vairagade, Nilakshi Jain Abstract - Despite the high classification accuracy of ML-based Network Intrusion Detection Systems (NIDS) achieved on the widely used NIDS benchmarks, there is still limited understanding of the robustness of these systems against adversarial perturbations and whether and how such perturbations transfer between separate models trained on independent datasets. In this paper, an empirical study is conducted to determine the ability of adversarial examples generated in one model to attack another model with a different structure and a different training dataset. We create adversarial examples with two commonly used benchmarks, CICIDS2017 and UNSW-NB15, and train four models (Random Forest, XGBoost for both benchmarks). BoundaryAttack is a black-box decision-based attack suitable for non-differentiable tree ensemble classifiers. We build a complete 4×4 matrix of Attack Success Rate for all source-target model pairs. From our results, we can see that the crossmodel transferability within-dataset is very high (89–100%), meaning that the robustness of the models is not significantly increased by their diversity if they are trained on the same data distribution. Conversely, cross-dataset transferability decreases significantly (5–44%) even when the feature space is limited to 10 harmonized features semantically shared between the two datasets. PCA analysis of the harmonized feature space reveals substantial manifold separation between datasets, explaining the observed transfer degradation. We propose that the disparity between feature spaces is a natural and meaningful obstacle to adversarial transferability, and directly influence the design and testing of adversarially robust NIDS deployments.
Authors - Yisel Clavel-Quintero, Ernesto Gongora-Rodriguez, Melissa Carmenaty-Ramirez Abstract - The Internet has signicantly transformed the business landscape, particularly in the tourism industry, by removing geographical constraints and time restrictions, while enhancing accessibility for consumers. Nowadays, users tend to search online for destinations and opinions from other travelers, make reservations, and share their own assessments. Therefore, customer reviews have become a valuable source of information for companies seeking to evaluate service quality and improve their products, advertising strategies, and overall performance. In this context, opinion mining and sentiment analysis have gained relevance, particularly in platforms such as TripAdvisor, which rely on usergenerated content. A key challenge in polarity detection is the correct interpretation of irony, as it can alter the intended meaning and sentiment of an expression. However, there are still few available TripAdvisor datasets, and, to the best of our knowledge, none are labeled for irony. We propose the creation of a dataset of TripAdvisor reviews annotated with both polarity and irony, alongside an experimental study to identify a model capable of eectively classifying the polarity of ironic TripAdvisor user reviews. Transfer learning was applied by adapting models trained on two source datasets for irony detection, and the best-performing model was subsequently used to annotate a TripAdvisor dataset with irony. Furthermore, experiments for polarity classication were conducted. The logistic regression model achieved the best performance in both tasks. The dataset obtained oers a valuable resource for future research on sentiment analysis and opinion mining in the tourism domain.
Authors - Wilma G. Villacis, Enith J. Mejia, Judith A. Silva, Carlos I. Nunez, Julio E. Cuji, Edder D. Naranjo Abstract - Immersive virtual reality environments have gained increasing attention in language education due to their potential to provide authentic and contextualized opportunities for communication. Despite this growing interest, limited attention has been given to the systematic design and validation of the dialogue scripts that support interaction within these environments. This study aimed to develop and validate CEFR-aligned dialogue scripts for A1-level learners of English as a Foreign Language. A material design and validation approach were adopted, combining expert feedback and quantitative evaluation. Through the integration of CEFR descriptors, communicative functions, and useful language, nine dialogue scripts were developed across two scenarios: a university campus and a shopping center. The scripts were evaluated through a two-round Delphi process involving five experts in Applied Linguistics and English language teaching. Quantitative data were analyzed using descriptive statistics, while qualitative feedback was examined through thematic categorization. Findings from the first Delphi round identified issues related to linguistic level alignment, naturalness, and interactional authenticity, leading to targeted revisions. The second round demonstrated a high level of expert agreement regarding the appropriateness of the revised scripts for A1 learners. The study provides a structured and transferable framework for the development of dialogue-based materials and contributes to the pedagogical design of immersive language-learning environments.
Authors - Emmanuel Tuyishimire Abstract - The fourth Industrial Revolution(4IR), together with the COVID-19 pandemic have made a loud call for digitizing diagnosis processes. The world is now convinced that it is imperative to digitize the diagnosis of long standing diseases such as malaria for more efficient treatment and control. It has been seen that malaria control would benefit a lot from digitising its diagnosis processes such as data gathering. We propose, in this paper, the architecture of a digital data collection system and how it is used to gather data for malaria awareness. The system is formally specified using Z notation, and based on the capability of the system, the malaria determinants are defined and their retrieving mechanisms are discussed.
Authors - Sara Sadiq Jawad, Dheyaa Jasim Kadhim Abstract - Software-defined networks (SDNs) suffer from dynamic congestion due to the nature of the data traffic they transmit. This includes both aggregated mobile network activity (such as video streaming, social media access, browsing, messaging, and mobile app usage) and real backbone internet traffic traces (which contain diverse packet flows from broadband services, cloud computing systems, downloads, server connections, and large-scale network transactions). This congestion reduces service quality and leads to poor resource allocation. Therefore, traffic prediction is considered a smart and efficient transition for SDNs. This paper proposes a deep learning-based predictive system for forecasting incoming traffic using two types of data (periodic Milano and bursty MAWI dataset). It also examines the impact of periodic and bursty traffic types on the prediction models used by the proposed system; and on the integration mechanisms used to translate predictions into actionable data. The results show that periodic Milano traffic requires temporal learning, while bursty MAWI traffic requires clipping, alignment, log scale, and robust prediction. Using MAWI traffic also required alignment between the training and evaluation phases through the application of cross-domain adaptation, unlike the Milano traffic which showed a direct response. The proposed system also demonstrated improved throughput, reduced congestion, and more stable decision-making.
Authors - Quoc-Anh Nguyen, Thanh-Nghi Doan, Huu-Hoa Nguyen Abstract - Crop productivity has been and continues to be influenced by both beneficial and harmful insect species. The classification of these insects plays a critical role in identifying threats and implementing crop protection measures. This paper presents a multimodal insect dataset for multimodal classification, utilizing both images and supplementary textual descriptions. The dataset is enriched to provide comprehensive information about various insect species. The study illustrates an integrated approach for feature extraction, similarity analysis, and insect classification. Furthermore, the research also introduces a model interpretation mechanism for the deep learning-based feature extraction process.
Authors - Petros Papagiannis, George Pallaris, Pantelitsa Leonidou Abstract - Predicting student academic performance presents a persistent challenge for higher education institutions. This paper presents a machine learning study at Cyprus College, Cyprus, using 311 studentcourse records across three semesters from 13 Computer Science courses. Four assessment components—midterm examination, final examination, assignments, and participation—alongside absence counts for 95 unique students are used as features. Seven algorithms are evaluated—Random Forest, XGBoost, SVM, Logistic Regression, K-Nearest Neighbours, Decision Tree, and Naive Bayes—using stratified five-fold cross-validation across three tasks: regression, binary pass/fail classification, and multiclass grade band prediction. SHAP analysis (applied to the full feature set) identifies feature contributions, while early-warning experiments exclude the final examination score to simulate mid-semester prediction. Results show that midterm and assignment scores predict final outcomes with R2=0.746 before the final examination, whilst Random Forest and XGBoost achieve 97.4% pass/fail accuracy. Participation contributes zero predictive signal despite a 10% grade weighting, with direct implications for assessment design at small higher education institutions.
Authors - Sukuse Abe Abstract - According to Einstein’s theory of relativity, the space we inhabit is distorted. Grigori Perelman solved the geometrization conjecture, which states that this space can be classified into eight spaces. Of these eight types, the classification of hyperbolic manifolds remains unresolved. Solving the volume conjecture would greatly advance the classification of hyperbolic manifolds. While the volume conjecture is one of the open problems in knot theory for knots in general, we successfully prove it for the family of oriented twist knots. The proof use the method of steepest descent and the theory of functions of several complex variables on the basis of colored Jones polynomials.
Authors - O. Aina, C.J. Van Staden, P. Makgato-Khunou Abstract - The role of leadership in the acceptance and integration of mobile technology for teaching and learning at a Private Higher Education Institution (PHEI) in South Africa have been explored. Mobile technologies have become widely used, but their application in higher education has been sporadic. The attitude towards mobile technology for teaching and learning are conditioned by the institutional environment which is in turn influenced by the leadership’s adoption. The study applied unified Theory of Acceptance and Use of Technology (UTAUT), Technology Acceptance Model (TAM) and Diffusion of Innovations (DOI) theories. A quantitative case study design was used with academic and institutional leaders. The study established a significant positive relationship between the construct of leadership perception and the acceptance of mobile technology for teaching and learning. The study concludes that leadership influences perceptions about acceptance of mobile technology for teaching and that enablers for the integration depend on the presence of appropriate communication channels and enabling conditions.
Authors - Samira Boulahbel-Bachari, Hind Dib-Slamani Abstract - This study examines digital governance transformation trajectories across fourteen Middle East and North Africa (MENA) countries between 2010 and 2024. Rather than classifying countries as simple “leaders” or “laggards,” it adopts a multidimensional framework covering digital governance, digital infrastructure, inclusion, institutional capacity, and economic capacity. Entropy weighting derives indicator weights, TOPSIS ranks countries according to their proximity to the best observed transformation profile, while hierarchical clustering identifies shared trajectory patterns. Robustness is assessed through VIKOR and principal component analysis. The results reveal marked regional heterogeneity. Saudi Arabia leads the ranking, followed by Türkiye, Oman, and the United Arab Emirates, while Morocco shows a balanced trajectory despite more limited economic resources. Other countries display differentiated progress across connectivity, online services, and institutional conditions, with Tunisia and Lebanon occupying the lowest relative positions. The findings show that progress in aggregate e-government scores does not necessarily reflect coherent digital governance development across all dimensions. The study advances a trajectory-based view of digital governance and offers a practical basis for regional benchmarking and policy prioritization in heterogeneous contexts.
Authors - Mohammad Arafat Ullah Abstract - Fault Detection and Classification (FDC) plays a critical role in semiconductor manufacturing by identifying defective wafers before subsequent processing stages, thereby reducing manufacturing cost, material waste, and production time. Traditional Statistical Process Control (SPC)-based FDC systems are widely used in semiconductor fabrication; however, machine learning techniques can significantly improve defect detection and process monitoring efficiency. In this research, the SECOM semiconductor manufacturing dataset collected from Kaggle was analyzed using multiple machine learning approaches. Several classification techniques including custom Support Vector Machine (SVM), kernel-based SVM, custom K-Nearest Neighbor (KNN), and Random Forest were implemented and compared for defective wafer detection. In addition, pseudo time-series semiconductor signals were reconstructed from static process features. Exponentially Weighted Moving Average (EWMA) smoothing and temporal feature extraction were then applied for signal-based fault analysis. Experimental results show that SVM-based approaches achieved strong classification performance on the SECOM dataset, while temporal signal reconstruction provided additional insight into semiconductor process behavior. The study presents a comparative analysis between conventional feature-based learning and reconstructed temporal feature-based learning for semiconductor fault detection applications.
Authors - M. A. M. P. Wanigaratne, K. B. H. M. T. T. Bandaranayake, C. S. Mohottala, J. V. Pannilage Abstract - Windows Subsystem for Linux (WSL) enables Linux command-line workflows to run directly on Windows endpoints, but this hybrid execution model creates security visibility and interpretation challenges. Host-side monitoring can identify that WSL was launched, but it may not provide sufficient Linux-side command context for threat investigation. This paper presents an explainable WSL command threat detection approach that combines machine learning-based risk scoring with Retrieval-Augmented Large Language Model (LLM) reasoning. The machine learning layer uses wrapper-aware and structure-aware command features to classify WSL-style command activity and convert model output into operational risk scores. The reasoning layer processes suspicious and malicious events using retrieved cybersecurity knowledge to generate analyst-readable explanations, MITRE ATT&CK mappings, confidence reasoning, and suggested defensive actions. The ML component was evaluated using a hybrid command dataset containing 8,028 samples, while the reasoning component was evaluated using 120 sanitized command level scenarios. Results show that the Calibrated SVM achieved 0.96 accuracy and 0.96 malicious class F1-score. Retrieval-augmented reasoning improved MITRE ATT&CK mapping accuracy from 52% to 87% and reduced hallucinated statements from 31% to 12%. The results indicate that combining ML risk scoring with grounded LLM reasoning can improve both alert prioritization and analyst understanding for WSL enabled endpoints.
Authors - Pavel E. Zhukov Abstract - The paper analyzes the problem of growth of public debt in developed countries with the compound interest approach, initially proposed with the Sargent-Wallace model. It is concluded that since 2002, when central banks began to apply the New Keynesian Model in monetary policy, governments have been widely using deficit financing of fiscal expenditures in order to stimulate economic growth. Based on the experience of 2002-2025, the parameters of the exponential growth of the debt-to-GDP ratio and the dangerous values of the budget deficit are assessed. General conclusions are made about the ineffectiveness of the dominant model of fiscal policy and the need to revise it, as well as the need to consider the growth of the money supply in monetary policy. General recommendations proposed. First: it is obvious that the United States and Japan have to introduce the VAT and do not increase customs duties. Second: In order to accelerate economic growth, it is necessary to shift fiscal policy priorities from the development of infrastructure and social programs to R&D, which will increase labor productivity. Third: Generally, all the social programs have to be audited. Specific for the United States, health insurance reform and limiting the growth of the budget deficit due to the Medicare and Medicaid programs are urgently needed. Fourth: Perhaps "national" companies with a high degree of localization should be stimulated with tax incentives for corporate income tax and shareholder income tax.
Authors - Unnati Parmar, Jatin Modh Abstract - Due to the high degree of infusion and morphology, Gujarati is regarded as a low-resource language in the Natural Language Processing field. The key reason why Gujarati can be classified as such is the lack of computing tools and annotated digitized corpus. Every single dialect of the language has its own morphological, lexical, and orthographic peculiarities since the language is extremely diverse. It includes the most diversified dialect – Kutchi – alongside Kathiawadi, Surti, Charotari, and Pattani dialects. The current paper focuses on the evolution in the sphere of Gujarati dialect identification via texts from 2021 till 2025. Morpheme segmentation, parts of speech identification, regional idioms identification, and neural machine translation model adaptation will be analyzed throughout this paper. This study looks at the transition from traditional grammar-based systems to modern deep learning algorithms. Performance metrics from the latest literature are used to identify research gaps. They include excellent results in the detection of idioms and high F1-scores for morphological tagging. Performance metrics of various models, such as transformers and Bidirectional Long Short-Term Memory network, are compared with DFA techniques. A framework for hybrid language models, combining both linguistics and neural networks, is proposed in the conclusion section of this literature review. Neural network models have been found to offer significant improvements in morphology when compared to traditional methods. In this paper, we address an inadequacy in the processing of informal language through the identification of disparities in resources between geographic variations. We propose a combination model that maintains geographic identity in modern-day computerized environments.
Authors - Athanasios Angelakis, Gabriele De Vito, Eleni-Myrto Trifylli, Filomena Ferrucci Abstract - Advanced fibrosis is a major determinant of liver-related morbidity in metabolic dysfunction-associated steatotic liver disease (MASLD). FIB-4 is widely used as a first-line non-invasive test (NIT), but its fixed formula may underuse non-linear diagnostic information contained in age, aspartate aminotransferase (AST), alanine aminotransferase (ALT), and platelet count (PLT). We evaluated whether machine-learning-enhanced NITs (MLE-NITs) can improve advanced fibrosis detection while preserving the clinically accessible FIB-4 variable space. We used three biopsy-validated MASLD cohorts from China, Malaysia, and India (n = 784). The Chinese cohort was split into 486 training and 54 internal validation/tuning patients; final performance was reported only on the Malaysian (n = 147) and Indian (n = 97) external cohorts. Models used five variables: age, FIB-4, AST, PLT, and ALT. We compared FIB-4 with a shallow-deep neural network (s-DNN), TabPFN, and gpt-4o-2024-08-06 in zero-shot and fine-tuned settings. FIB-4 achieved external thresholded ROC-AUCs of 0.75 and 0.60 in Malaysia and India, respectively. TabPFN achieved 0.69 and 0.66, fine-tuned GPT-4o achieved 0.75 and 0.63, and the s-DNN achieved 0.77 and 0.67. The s-DNN contained only 354 trainable parameters, compared with 7,244,554 parameters for TabPFN, and provided the most balanced fixed-threshold operating profile. External diagnostics showed s-DNN Brier scores of 0.18 and 0.22, with AST and FIB-4 as dominant permutation-importance variables. Exploratory decision-curve analysis showed cohort-dependent clinical utility, favoring TabPFN in Malaysia and s-DNN in India.
Authors - Ali Raheman, Asad Khan, Tejas Bhagat, Fazal Raheman Abstract - The emergence of quantum computing challenges digital infrastructure to achieve both quantumresilient security and energy-efficient performance. This is particularly critical for decentralized finance (DeFi), where layered software stacks and exposed authentication mechanisms increase complexity, overhead, and attack-surface exposure. This paper proposes Quantum Ledger Technology (QLT), a blockchain-agnostic architectural framework that integrates Zero Vulnerability Computing (ZVC), Solid-State Software-on-a-Chip (3SoC), and Quantum-Resilient User-Evasive Cryptographic Authentication (QRUECA). Together, these mechanisms reduce software-mediated trust, minimize exposed authentication surfaces, and shift trust enforcement toward hardware-rooted execution environments. A hypothesis-driven evaluation framework is introduced to assess whether architectural simplification can reduce authentication latency, memory usage, energy overhead, and attack-surface complexity while preserving functional equivalence with existing blockchain systems. Preliminary results indicate that hardware-rooted authentication and reduced trusted software complexity can provide a promising foundation for scalable, energy-aware, and quantum-resilient digital infrastructure. The proposed framework aligns with the goals of secure, sustainable, and intelligent future computing systems.
Authors - Shamsa AlNasri, Muna Ali AlShamsi, Mariam AlNuaimi, Hanae Ouahhabi, Gurdal Ertek Abstract - This study presents an analytics framework for analyzing and benchmarking sales transactions data of e-commerce products across multiple countries. The framework consists of an integrated multi-faceted application of a carefully selected portfolio of data analytics techniques. Specifically, the framework combines (a) statistical distribution fitting to well-known probability distributions (Gamma, Normal, Weibull, Lognormal), (b) box plot analysis followed by statistical hypothesis testing (Kruskal-Wallis and Dunn tests) and visualization of pairwise comparison results, and (c) text mining (Latent Dirichlet Allocation (LDA) and word clouds). Although many studies in the literature report on the analysis of e-commerce product sales, this is the first study that combines the mentioned techniques within a multi-faceted yet also unified approach. The results obtained for a case study on the Gulf Cooperation Council (GCC) countries reveal regional differences in consumer behavior, pricing, and preferences. The insights obtained can be used to improve the marketing and engagement of the selected case with the selected products and countries. However, more importantly, the primary contribution of the study is the generalizable analytics framework presented that can be adopted and applied to any product set and country selection with similar data attributes.
Authors - Deepak Mane, Ashwanth Nair, Nihar Gundale, Om Khamkar, Tanmay Kulkarni, Ranjeet Bidwe, Amol Kamble, Suraj Sawant Abstract - There many areas in which the organization can make use of technologies that will make decision making easy. ai (artificial intelligence) is one of the most useful and innovative technologies that is used in many fields of organization to help in business management and decision making. in the recent years, HR department has became very important in the organization, since the quality and skills of the worker in directly proportional to the performance of the organization. After ai is being used in many ways in the organization like in marketing and sales department, now its starting to guide HR department for employee related decision. The purpose of using ai in HR department in to support decision that are based on objective data analysis, not on subjective aspects. The goal of this work is to analyse influence on employee attrition and objective factors. In order to identify the main causes that contribute to a workers decision to leave a company, and identify the employee that about to leave a company. After training, the obtained model for the prediction of employs attrition is tested on real dataset provided by IBM analytics, which has 35 features and about 1500 samples. Results are obtained in terms of classical metrics and the algorithm that produced the best results of the dataset is the gaussian naïve bayes classifier. It has best recall rate of 0.54, since it measures the ability of classifier to achieves an overall false negative rate equal to 4.5% of the total observations.
Authors - Deepak Mane, Ashwanth Nair, Nihar Gundale, Om Khamkar, Tanmay Kulkarni, Ranjeet Bidwe, Amol Kamble, Suraj Sawant Abstract - Accurate emotion recognition remains a significant challenge in affective computing, particularly when relying on unimodal approaches such as facial expression analysis. These systems are inherently limited because individuals can deliberately mask their emotions, and visually similar expressions such as fear and surprise often lead to misclassification. Such limitations highlight the need for more robust methods that incorporate complementary sources of information. The proposed system uses a multimodal framework which combines the Circumplex Model of Affect through its visual and physiological cues to achieve better reliability. The FER-2013 dataset provides data for a Convolutional Neural Network which estimates emotional valence based on facial expressions captured through standard camera systems. The MAX30102 photoplethysmography sensor measures heart rate and heart rate variability through its connection with an Arduino to determine emotional arousal. The rule-based fusion engine combines these modalities to determine the final emotional state which it then categorizes into joy, stress, anxiety, and calmness. The system uses physiological data to clarify between emotional states which appear similar and it also identifies hidden emotional states which facial expressions cannot express. The system offers health monitoring, human computer interaction, and psychological assessment fields a dependable and efficient solution.
Authors - Caio Tertuliano Ribeiro, Lilian Berton Abstract - This paper studies a Disparate Impact (DI)-oriented variant of fairness-aware hyperparameter optimization for XGBoost in banking and credit settings. The method combines Design of Experiments (DoE), Response Surface Methodology (RSM), and NBI-style sampling to jointly tune XGBoost hyperparameters, the decision threshold, and the positive-class weight under a fixed evaluation budget. Unlike the preliminary composite-fairness draft, the final experiment optimizes a DI-only objective while keeping Statistical Parity Difference (SPD), Equal Opportunity Difference (EOD), and Average Odds Difference (AOD) as audit metrics. Experiments on Bank Marketing, German Credit, and Default of Credit Card Clients with 30 replicas per dataset show that the proposed utopia selector is competitive with evaluation-matched random search and consistently superior to the XGBoost default configuration. Relative to the default baseline, it improves Balanced Accuracy by +0.059, +0.020, and +0.019, while reducing the DI gap by -0.375, -0.174, and -0.141, respectively. The main takeaway is that DI-only optimization provides a finance-oriented and reproducible way to navigate fairness–performance trade-offs rather than a universal domination claim over random search.
Authors - Syeda Fatima Rafique, Mohammed Abobaker Baobaid, Majid Shaher Ebrahim Tayfour, Hamed Marhoun Khamis Alsaedi, Ibrahim Alfaki, Gurdal Ertek Abstract - With the invention of Large Language Models (LLMs) and the development and increased usage of generative AI platforms, “vibe coding” (AI-assisted coding, coding with AI assistants) has become an integral part of software development, documentation, and maintenance. Although there are multiple detailed studies on the experiences of developers with vibe coding for software development, no earlier work was encountered on the vibe coding of analytics dashboards in particular. However, in an era of exponential growth in data volume, variety, and velocity, data analytics, and in particular, analytics dashboards, are highly relevant and can serve as competitive leverage for every organization. This paper is the first attempt in the literature to answer the following research question: “What are the practical project experiences of developers during vibe coding of analytics dashboards, especially in terms of challenges faced?” In this paper, experiences in two case study projects on analytics dashboard development are shared as lessons learned to guide developers, product managers, and project managers.
Authors - Marcos Paulo Jeronimo Francisco, Carlos Hideo Arima, Napoleao Verardi Galegale, Joshua Onome Imoniana Abstract - The increasing complexity of digital systems and the growing sophistication of cyber threats have intensified the need for proactive security assessment methods. Threat Modeling is a structured practice for identifying potential vulnerabilities, attack paths and mitigation strategies during the software development lifecycle. However, its manual application is often time-consuming, subjective and dependent on scarce cybersecurity expertise. In this context, Large Language Models (LLMs) may support security teams by generating threat hypotheses, classifying risks and recommending controls. This study evaluates the effectiveness of three LLM-based tools — ChatGPT, Gemini and Manus — in cybersecurity threat modeling for a real-world backend information system. A controlled computational experiment was conducted using standardized prompts applied to the three models, with three independent executions per prompt. The evaluation considered five dimensions: threat identification coverage, technical depth of analysis, quality of risk classification, assertiveness of control recommendations and result consistency. To consolidate the comparison, a Final Effectiveness Metric (FEM) was proposed. The results show different performance profiles among the evaluated models. Manus achieved the highest FEM score, with stronger threat coverage, technical depth, control recommendations and consistency. ChatGPT presented intermediate performance, with structured and detailed analyses, while Gemini showed lower threat coverage, but satisfactory technical reasoning in specific tasks. The findings also indicate that LLMs can enhance threat modeling activities by expanding analytical capacity and supporting DevSecOps practices. Nevertheless, their outputs require human validation, especially regarding risk classification, framework alignment and false positive analysis.
Authors - Jeanne Roux Ngo Bilong, Python Ndekou Tandong Paul, Bakary Kone, Dethie Dione, Ibrahima Toure, Boris Sourou ZANNOU, Hamidou Dathe, Mamadou Diarra, Olga Ngangmo Kengni, Mamadou Thiam, Cheikh Amed Diloma Gabriel Traore Abstract - Breast cancer is a non-communicable disease that causes thousands of deaths each year worldwide. Early detection of breast cancer in women is a public health priority in developing countries. Based on data collected from patients’ breasts, machine learning models can help predict the risk of developing breast cancer.We used three machine learning algorithms (SVM, decision tree, and random forest) for predicting the risk of developing breast cancer, taking into account the physiological factors of breasts. Data collected from 568 patients was used to train the machine learning algorithms. An evaluation of the performance of the three algorithms showed that the random forest algorithm had the highest F1 score, which led to the selection of this algorithm for creating a computer application to diagnose breast cancer risk. The results of this algorithm show 97% accuracy with 94% recall in predicting high breast cancer risk and an F1 score of 96%. The result obtained indicates the model’s excellent ability to correctly predict high cancer risk across the entire risk probability spectrum. The good performance of the model using the Random Forest algorithm could be useful as first-level medical support for breast cancer screening. The developed model is capable of providing the probability of the risk of developing breast cancer for each female patient.
Authors - Sergey Kubinski, Emil Hadzhikolev Abstract - This paper presents a Smart Fitness Assistant system for generating personalized workout and dietary recommendations using machine learning and domain-informed physiological feature engineering. The proposed approach incorporates indicators such as Basal Metabolic Rate, Total Daily Energy Expenditure, and target caloric intake derived from user data. The recommendation task is formulated as a multi-class classification problem for exercise and diet planning. Decision Tree, Multilayer Perceptron, and Random Forest models are evaluated using both baseline and enriched feature sets. Experimental results demonstrate that the inclusion of physiological features improves predictive performance, with Random Forest achieving the highest accuracy. The developed system is implemented within a modular software architecture that supports user interaction, recommendation generation, data management, and progress tracking.
Authors - Madhulika Gajjala Abstract - The fast-paced development in digital transformation strategies has considerably transformed human resource management systems towards cloudbased, smart and efficient workforces management tools. As current HR systems face various problems like data repository fragmentation, poor connectivity options, slow synchronization processes, scalability issues, and lack of advanced workforce intelligence, the current study aims to develop an AFDHRP framework, which would comprise a CSIL, AOIE, WIMM, and ADHOU components. In addition, this research proposes the implementation of Cloud-Integrated HR Data Synchronization and Interoperability (CHDSI) algorithm and Adaptive Workforce Intelligence and HR Optimization (AWIHO) algorithm into an innovative solution. A comparison was made with the two state-of-the-art models, Deep-Hill and Deep Learning-Based ERP Optimization System, on several performance metrics including human resource integration accuracy, interoperability efficiency, synchronization reliability, workforce intelligence metric, decision support capability, cloud scalability, API ecosystem performance, and FutureReady HR Platform Effectiveness Metric. According to the results, the AFDHRP performed excellently scoring an accuracy of 99.5%, 99.2% for interoperability efficiency, 99.4% for synchronization reliability, and 99.8% for overall platform effectiveness.
Authors - Jose Hernandez Cortaza, Arturo Corona Ferreira, Pablo Payro Campos Abstract - The research adopts a qualitative approach through a descriptive-explanatory case study, supported by source triangulation, integrating direct observation and expert judgment in the requirements engineering process, with the purpose of identifying critical points in the traditional digital document certification workflow at a Mexican public higher education institution. Based on this approach, a digital transformation model was designed alongside a system architecture aimed at guaranteeing the integrity, decentralization, and verifiability of digital documents, integrating Blockchain, IPFS, and smart contracts, which enabled a multilayer verification scheme. The results demonstrate that the proposed system allows multilayer verification, prevents document duplication, and strengthens transparency and trust in academic issuance and certification processes. In terms of performance, the system presents an estimated Gas per transaction of 327,423, with an average processing time of 17.94ms, during which the entire document issuance and certification process is fully executed. The main contribution of this work is the proposal of a replicable digital transformation model that combines emerging technologies (Blockchain, IPFS, and smart contracts) with a qualitative organizational analysis, providing a practical framework for the secure management of academic documents in public higher education institutions, significantly contributing to the improvement of document management. This model can be adopted and adapted by institutions seeking to strengthen the integrity and interoperability of digital issuance and certification processes.
Authors - Shayma W. Nourildean, Yousra Abd Mohammed, Nahida Naji Kadhim Abstract - The growth of the Internet of Things (IoT) equipment has changed many industrial and social applications, but it has made the IoT network vulnerable to a wide range of malicious activities by increasing the number of potential attack points. Traditional IDS has a problem with scalability, adaptability and accuracy when facing new and complex cyber threats. In this study, a strong ensemble machine learning framework that integrates Decision Tree (DT), Random Forest (RF), and XGBoost through confidence-based soft voting, had been validated across individual datasets. The proposed system (DTXG-RF) uses different forces and weaknesses of these algorithms to improve the accuracy of the detection and reduce the chances of causing it a false alarm. Two Benchmark IoT data sets, CIC-IoT2023and IoTID20, were used. These data sets showed a wide range of scenarios in the real IoT attack. To reduce overfitting risk and data leakage, strict train–test separation, stratified splitting, and pipeline-based preprocessing were enforced, and additional cross-validation experiments were conducted to verify model stability across folds and datasets. Assessment results showed that DTXG-RF ensemble voting model consistently improves traditional machine learning models such as DT, XGBOOST, KN, logistic regression, Nave Bayes and Catboost. The model accuracy of CIC -IoT-2023 and IOTID20 were 94.03% and 99.99%, respectively, with AUCs of 0.95025 and 0.9994. These results indicated that the ensemble IDS was lightweight and could achieve high detection accuracy with low latency and memory overheads, which is also suitable for low-latency IoT edge deployment.
Authors: Manuela Moreno-Arcila, Luz Marcela Restrepo-Tamayo, Gloria Piedad Gasca-Hurtado Abstract. Productivity in software development teams is a multidimensional phenomenon that cannot be reduced to delivery metrics or technical activity indicators. The social and human factors that influence collective performance are consistently overlooked in the measurement frameworks used in practice. The lack of a tool that integrates social, human, and process dimensions with execution results in a structured way prevents a holistic and actionable measurement of software teams' productive capacity. The lack of a tool that systematically integrates social, human, and process dimensions with execution results prevents a holistic and actionable assessment of software teams' productivity. The PCI was designed following the Design Science Research (DSR) paradigm, through a process consisting of two interconnected components: 1) the design and content validation of the Team Capacity Measurement Instrument (TCMI), through expert judgment and calculation of the Content Validity Coefficient (CVC), and 2) the conceptual construction of the index, including the definition of socio-technical dimensions, standardization of variables, conceptual weighting, aggregation, and interpretation using bands and alerts by dimension. The PCI generates a standardized composite score on a [0–100] scale using a weighted aggregation formula with theoretically justified weights, accompanied by a classification scheme that divides the system’s health into four bands and a mechanism for generating actionable alerts by dimension.
