AL-Rafidain Journal of Computer Sciences and Mathematics

The offline voice control systems are becoming more prominent in the smart home environment because of their ability to deliver quick feedback, preserve the privacy of the user, and be able to function even without network connectivity. However, their usage has not been mainstreamed, due to restrictions on vocabulary, resistance to noise, and hardware performance. In this paper, the authors attempt to provide a critical assessment of the recent developments in offline voice control technology in smart homes. On devices with limited resources, it evaluates a variety of datasets, frameworks, embedded platforms, and algorithmic enhancements. Moreover, the paper examines the issues linked to the growth of vocabulary, audio variability, and privacy, as well as interoperability and compares offline and online paradigms in terms of a more detailed analytical framework. Among the emerging trends highlighted include lightweight keyword spotting, embedded spoken language understanding, neural processing units, and on-device personalization, as being crucial in the evolution of future systems. This paper summarizes the current situation and outlines the future trends of the offline voice control technologies as it is applicable to smart home usage by synthesizing recent achievements and pinpointing the current limitations of the technology.

This paper explains the application of Ito-integral formula to finding the moments for nonlinear stochastic differential equations (SDEs) with coefficients (inverse trigonometric function) such as sin^(-1) (x), cos^(-1) (x) and tan^(-1) (x), this method naturally constrains state variables within predefined physical or operational limits. The study derives moment equations to explain and analyze statistical properties, including mean and variance, as well as higher-order moments( the moment generation function ), while addressing the challenges of nonlinear drift, diffusion interactions, and multiplicative noise and the main results show that finite terms such as cos^(-1) (x) enforce stability in robotic angular control, yielding error bounds E[(cos^(-1) (X))^2]≤σ^2/(2κ), while gene regulatory models with the function tan^(-1) (x) ensure protein concentrations remain biologically viable, limiting transcriptional noise to "Var" (tan^(-1) (X))≤〖(π/2)〗^2 , With inherent limitations, multiplicative noise can push moments toward saturation. Systematic trade-offs in moment closure approximations and numerical verification (e.g., Euler-Maruyama, stochastic Runge-Kutta) are critically evaluated, highlighting their effectiveness in biological and engineering applications and the framework is extended to fractional SDEs and multidimensional systems, proposing machine learning techniques to solve moment hierarchies and enhance predictive modeling by uniting theoretical rigor with practical insights and this work advances robust stochastic modeling in constrained systems, providing scalable solutions for gene networks, financial markets, and autonomous navigation under uncertainty.

Background: In a variety of medical specialties, Telemedicine is extensively applied. Nevertheless, it is not evident whether this approach will be critical once COVID-19 is endemic. In this research, we have discussed the opinions of physicians and patients regarding the use of telemedicine.
Methods: The anonymous electronic questionnaire was sent forwarded to physicians working in the colleges of Medicine at the University of Mosul and the University of Nineveh plus to physicians working in the public and private hospitals and also in the private clinics in Mosul randomly without regard to their specialty. The survey contained 18 questions about the feasibility and acceptability of video consultations, ease of use, quality, and other issues. The evaluation of most of the items was on five-point Likert scale, but the perceived benefits and future expectations were open-ended. The survey of patients was also carried out through an electronic connection that was distributed randomly and included 14 questions on the willingness to undergo a video consultation, ease of use, benefits, difficulties, the mode of choice of consultation and the overall satisfaction measured on the same scale.
Results: 114 physicians and 142 patients were involved. Video consultations were positively received by patients, with the reportedly acceptable communication and patient satisfaction, although there were some fears about technical difficulties and privacy. physicians were more reserved, indicating that they did not feel comfortable with virtual communication, decreased trust with remote clinical evaluation, and significant doubts about the lack of physical examination and laboratory results. The questionnaire had a satisfactory degree of reliability (Cronbach alpha = 0.736). There were considerable differences in a number of physician perceptions, but the perception of patients in different groups did not exhibit significant differences.
Conclusion: The physicians and patients found telemedicine to be an appropriate and viable mode of care provision and reception. Technical problem was the key to the successful use.

The exponential growth of neural networks and artificial intelligence (AI) has revolutionized diverse fields, including healthcare, finance, natural language processing, and autonomous systems. These technologies have redefined the boundaries of what is possible, enabling solutions to complex problems across multiple domains. This review critically examines the efficiency of neural networks and AI applications in advanced settings, with a focus on computational performance, scalability, energy consumption, and ethical implications. By evaluating state-of-the-art architectures and application-specific implementations, the study identifies pressing challenges, including environmental sustainability and data privacy, alongside opportunities for improvement. Furthermore, it highlights emerging trends such as Green AI, federated learning, and neurosymbolic AI that are shaping the future of the field. This comprehensive analysis aims to provide actionable insights for researchers and practitioners seeking to optimize AI systems for greater effectiveness and impact. This review uses a structured literature review approach to examine approximately 50 sources, including journal articles, conference papers, preprints, and selected technical reports published up to 2026. The scope is limited to efficiency-oriented neural network and AI applications in healthcare, autonomous systems, natural language processing, finance, and environmental/climate science. The main contribution is a cross-domain comparison of algorithms, efficiency metrics, deployment settings, and open research gaps. The review finds that no single technique is universally efficient: CNNs and vision transformers remain strong for perception tasks, transformer and retrieval-augmented models dominate language applications, graph neural networks and reinforcement learning support relational and decision-making tasks, and compression, federated learning, edge deployment, and hardware acceleration are increasingly required to control latency, energy consumption, and scalability costs.

The importance of intrusion detection systems (IDSs) is part of existing cybersecurity facilities. However, existing machine learning-based IDSs are vulnerable to adversarial attacks and evasion techniques. Generative adversarial networks (GANs) have demonstrated the possibility of using synthetic data augmentation to address the problem of imbalance in network security datasets. Nevertheless, current methods examine performance using clean datasets without considering resilience to adversarial corruption. In this study, a framework of dual GAN (DGF) is proposed, in which one of the GANs produces data to augment the data with synthetic samples, and the other GAN generates adversarial instances to improve the robustness of the model in the training process. The synthesis of synthetic data and the adversarial robustness of intrusion detection have been under-researched. The proposed framework focuses on this aspect. The experimental results on the NSL-KDD benchmark dataset indicate that DGF decreases the false-negative rates by an absolute of 1.11% (11.11% on clean data and 10.00% on evasion conditions), and its relative performance is decreased by 41.5 %–46.0% on augmentation-only baselines. The accuracy of the framework in detecting clean data reached 92.44%, and under adversarial conditions, 92.13% (only 0.31% decreased). This is better than the existing procedures, which have a reduction in accuracy of over 30% in the case of the same perturbations. These findings reveal how dual-purpose GAN architectures can be useful in designing IDSs that work in adversarial network settings.

Social media has modernized advocacy by enabling individuals to mobilize, spread opinions, and champion justice globally. Accurate prediction of social media advocacy campaign success can play a significant role in promoting social and political change. Motivated by the recent #JusticeForNavalAbbas advocacy campaign in Nigeria, this study aims to predict the success rate of posts related to the campaign. By examining how social media platforms including Facebook, TikTok, Instagram, and X (formerly known as Twitter) influence public opinion, this paper investigates how these platforms attract engagement, amplify the voices of the marginalized, and encourage collective action. The present research leverages five machine learning techniques, including Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and XGBoost (XGB). The dataset was preprocessed and split into training and testing subsets comprising 75% and 25% of the data, respectively. Model performance was evaluated using Accuracy, Precision, Recall, F1-Score, and Area Under the ROC Curve (AUC). RF and XGB demonstrated exceptional performance in predicting social media post success, achieving accuracies of 98% and 97%, respectively. The findings highlight the significance of social media platforms in amplifying advocacy campaigns at both local and global levels.

In this paper, we investigate 〖Micro 〗_Z Continuous functions and examine their relationships with various classes of Micro _Z-Continuity. In particular, we introduce several types of Continuity, including μ_(Z )Semi Continuity, μ_(Z ) -pre-Continuity, μ_(Z )-α Continuity, μ_(Z )-regular Continuity, and μ_(Z )-β Continuity. Furthermore, we study a number of fundamental properties of Micro _Z-Continuous functions. The main results are established through rigorous theorems and supported by illustrative examples.

This paper discusses the general boundary integral equations, Riemann mapping function for bounded simply connected regions whose boundary curves are smooth and analytic, that correspond to a real-valued kernel, and a new boundary integral formula is introduced. Based on the multiplication of the Kerzman-Stien Trummer integral formula (briefly, KSTiF) by the penalty function, such that the complex-valued kernel transforms into, we derive a new formula and prove its uniqueness. Numerical results using the Nyström method with the trapezoidal rule yield approximations of high accuracy when compared with exact solutions for test regions.

In arid and semi-arid areas, especially in Iraq, extreme heat has become one of the most significant impacts of climate change, as the long duration of summer heat is becoming a serious threat to human health, water supply and infrastructure. This paper explores the issue of summer extreme temperatures in Iraq based on ERA5-Land daily maximum temperature data of 1980-2024. Descriptive heatwave indicators were then computed after which annual summer maxima were modeled in a Bayesian non-stationary generalized extreme value (GEV) framework in three climatic regions: North, Central, and South Iraq. A linear time effect in the location parameter was used to introduce non-stationarity, whilst the scale and shape parameters were fixed to maintain inferential stability because of the small number of annual maxima to use in each region. The results reveal a clear north–south gradient in baseline extreme heat intensity and a statistically credible warming trend of approximately 0.32 °C per decade. Return-level analysis further indicates that 20-year and 50-year extreme temperature levels increased by about 0.7-0.8 °C between the early period (1980–1999) and the recent period (2000–2024), with the most severe extremes concentrated in southern Iraq. These findings provide probabilistic evidence that extreme heat risk in Iraq is intensifying under a non-stationary climate and demonstrate the value of Bayesian non-stationary extreme-value modeling for regional climate-risk assessment.

With the rapid growth of the Internet of Things (IoT), ever greater quantities of complex data are created. To analyze this data and take better decisions, Artificial Intelligence (AI) especially in the form of Machine Learning (ML) and Deep Learning (DL) has been widely adopted, which has led to the rise of Artificial Intelligence of Things (AIoT). This study is a systematic literature review consisting of 79 peer-reviewed articles published in the last five years (from 2019 to 2024), following the PRISMA guidelines. We thoroughly searched major databases, and used strict quality assessment and inclusion/exclusion criteria for assessing the influence of the AI in five fields, namely, Smart Home, Smart Transportation, Smart Education, Smart Grid, and Robotics. Based on our quantitative results, we notice that the most studied domain is the Smart Transportation, and the most used technologies are recurrent models such as LSTM and convolutional networks (CNN). The review compares the performance of the models critically, by considering the evaluation metrics including accuracy, latency, and computational cost. Moreover, the paper discusses key challenges and limitations in the existing AIoT deployments such as data privacy, cybersecurity threats, power consumption and explainability of the models. This review aims to fill the gap of literature by providing a structured reference towards future research on scalable, secure and energy-efficient AIoT systems.

In this paper, we characterize the texture of tripotent graph of the ring, which is associated by tripotent element of a ring such that, for every two distinct vertices; s and t then, their adjacent if and only if 〖(s+t) 〗^3=s+t. We establish several structural graphical properties of tripotent graph of local ring when it is a path graph and it is spanning subgraph with unit graph and clean graph. Also, we obtain that, the tripotent graph is Hamiltonian graph but doesn't Eulerian. Moreover, we investigate the concept of tripotent graph of a commutative ring R where R≅Z_2r and R≅Z_3r for r is a prime integer as diam(TP(R)), m (TP(R)) and ω(TP(R)) which equals to χ(TP(R)). Furthermore, we prove that the tripotent graph of commutative ring is 4-partite graph and γ_ind TP(R)=r, for R≅Z_2r. Finally, we are concerned about the metric dimension of tripotent graph of some cases of commutative rings, especially; where tripotent graphs are planar and nonplanar.

As we know, cloud computing has become an increasingly popular solution to deliver scalable and effective solutions to the needs of today's modern applications. One of the greatest problems in this context is the efficient allocation of resources to the different needs of numerous users and applications. Reinforcement learning (RL) is an approach that can adapt dynamically to changing environments and is a promising solution. In this paper, the SARSA (state action reward state action) RL algorithm is combined with the whale optimization algorithm (WOA) to propose a new resource allocation approach for cloud computing based on a multi objective optimization. We are going to refer to this method as MO SAWOA. It assists the SARSA algorithm to faster converge and prevents it from being trapped in the local optima. The proposed system handles complex multi objective resource allocation issues by improving load balancing, reducing makespan, saving costs, and enhancing user experience in dynamic cloud environments. It can also be used for real time systems (RTS) like public clouds like Amazon Web Services (AWS).