The paper investigates an initial-boundary value problem of the Sturm type for the heat equation with a piecewise-constant coefficient and a Caputo derivative of order 0 < a < 1 with respect to time. The domain under consideration is an interval (0, l) containing a strictly internal point of discontinuity x = x₀. From a physical point of view, this problem models the process of temperature field propagation in a thin rod of length l. The rod is composite and consists of two sections (0, x₀) and (x₀, l), possessing different thermophysical characteristics. At the contact point of the two media x = x₀ , perfect contact conditions are prescribed, implying continuity of temperature and continuity of heat flux when passing from one medium to another. The main goal of the work is to substantiate the solution of the posed initial-boundary value problem by the method of separation of variables (Fourier method). Application of this method leads to the necessity of investigating the corresponding spectral problem for an ordinary differential operator with a discontinuous coefficient at the highest derivative.

The paper studies the existence of solutions to a nonlocal boundary value problem for the Poisson equation in the unit ball. The boundary condition is nonlocal and relates the value of the unknown function to its normal derivative at diametrically opposite points on the unit sphere. A necessary and sufficient solvability condition is derived in the form of an orthogonality relation for the prescribed right-hand side and the boundary function. Uniqueness of the solution is proved up to an additive constant. An integral representation of the solution is derived using a specially constructed Green’s function for the nonlocal problem, which is expressed through the Green’s functions of related auxiliary local boundary value problems. In addition, the corresponding spectral problem is investigated. For its analysis, two auxiliary spectral problems with Neumann and Robin boundary conditions are introduced, and their eigenvalues and eigenfunctions are described in detail. It is demonstrated that this approach yields a complete orthonormal set of eigenfunctions for the original nonlocal problem, which forms a basis L₂(Ω).

The paper considers the application of Physics-Informed Neural Networks (PINN) to solving a two-dimensional Poisson equation with inhomogeneous mixed boundary conditions. The PINN method combines neural network approximations and physical laws, including the residual of the equation and errors at the boundary, into a single loss function. The model is implemented in the PyTorch environment and trained on a sample generated by the Latin hypercubic sampling method. To verify the accuracy, a comparison is made with an analytical solution, as well as with traditional numerical methods. The influence of key hyperparameters is systematically studied: the number of training points, weights in the loss function, network architecture, and training step. It is found that increasing the weight on the boundary conditions improves convergence and accuracy, and optimal architectures with 4-5 hidden layers and 64 neurons provide a relative error of less than 0.1-0.2%. Unlike numerical schemes requiring grid discretization, PINN has demonstrated versatility and high accuracy when dealing with sparse data and complex conditions, making it a promising tool for computational mathematics and modeling of physical processes in conditions of complex geometry and incomplete information.

The paper studies a local analogue of the strict order property inside a cut for linearly ordered structures, arising in the theory of order-stable models. A natural attempt to formulate this property in terms of finite c hains o f s trictly n ested d efinable su bsets in tersected wi th re alizations of a cu t is considered. It is shown that such a formulation is not elementary: it may hold in an elementary extension but fail to be preserved in the original model and therefore does not provide a correct local criterion. To overcome this drawback, the notion of strict inclusion of convex components of definable s ets i s i ntroduced, together with its uniform version for families of formulas with parameters. It is proved that the strict order property inside a cut is equivalent to the existence of the uniform property of inclusion of convex components. The obtained result clarifies t he l ocal n a ture o f t his p r operty, s hows i ts d i fference fr om th e cl as sical lo cal strict order property, and makes it applicable to the study of order-stable theories and their models. Thus, the paper complements the criterion of order-stability.

This paper studies a control problem for a one-dimensional pseudo-parabolic equation with a nonlocal boundary condition of the Bitsadze-Samarskii type. The control function appears as a distributed source term on the right-hand side of the equation, and the ob jective is to steer the system from the zero initial state so that the weighted average temperature follows a prescribed trajectory. The equation contains a third-order mixed derivative term that accounts for memory or viscoelastic effects i n h eat conduction, which is particularly relevant for materials with internal structure such as polymers, biological tissues, and granular media. Using spectral decomposition and a biorthogonal basis, the problem is reduced to a Volterra integral equation of the first kind. The kernel of this integral equation is proved to be p ositive and continuous. Under suitable assumptions on the weight function and the spatial distribution of the control, the existence of an admissible control is established for a class of desired trajectories with sufficiently small norm. The Laplace transform method is employed to solve the integral equation and to obtain bounds on the control function. The results quantify the influence o f t he n onlocal b oundary c ondition a nd the memory parameter on the controllability of the system.

In this paper, we consider an inverse problem for a parabolic equation in which a time­dependent coefficient is in cluded in th e bo undary co nd ition. Ad ditional in formation is gi ven by an overdetermination condition that relates the solution to a given function of time. To solve the problem, the spectral method is used, which is based on the expansion of the solution in terms of eigenfunctions of the corresponding boundary value problem with homogeneous boundary conditions. Due to this expansion, the original problem is reduced to a system of ordinary differential equations for the expansion coefficients. Af ter so lving th e ob tained sy stem an d su bstituting the representation of the solution into the integral condition, the problem is reduced to a nonlinear Volterra integral equation with respect to the unknown coefficient. Fo r th e ob tained eq uation, the corresponding operator is introduced and its properties are studied. The existence and uniqueness of the solution on the interval [0, T ] are proved under the contraction condition for the constructed operator. It is shown that the presence of an unknown coefficient in th e bo undary co ndition leads to a change in the structure of the resulting integral equation compared to classical formulations of inverse problems. The proposed approach can be applied to a wider class of inverse problems for parabolic equations with nonclassical boundary conditions.

The paper investigates a two-point boundary value problem for a loaded differential e quation with a Caputo fractional derivative. The Dzhumabaev parametrization method is applied to study the problem. Within this approach, the initial interval is divided into a finite n umber of s ubintervals, a nd additional parameters and new functions are introduced, which makes it possible to reduce the original boundary value problem to an equivalent multipoint boundary value problem with unknown parameters. The equivalence of the original and transformed problems is established. Based on the obtained integral representations, a system of algebraic equations with respect to the introduced parameters is constructed. Sufficient conditions for the existence and uniqueness of a solution are established and related to the invertibility of the corresponding matrix. An algorithm based on the method of successive approximations is developed to determine the unknown parameters and functions. A numerical example demonstrating the application of the proposed algorithm and confirming its e ffectiveness is presented. The ob tained results de monstrate the efficiency of the pa rametrization me thod for in vestigating bo undary va lue pr oblems for lo aded fractionalorder differential e quations and c onstructing t heir a pproximate solutions.

This paper investigates a boundary value problem belonging to a family of linear differential equations with nonlinear two-point boundary conditions. For each fixed v alue of t he f amily p arameter, t he origi­nal problem is transformed into a nonlinear two-point boundary value problem for ordinary differential equations. Since such problems are associated with nonlocal boundary value problems, including systems of hyperbolic equations, determining the conditions for their solvability is of considerable interest. The study employs the parameterization method developed by D.S. Dzhumabayev, which is based on reducing the problem to a system of nonlinear algebraic equations with respect to the parameter and solving an initial value problem for a parameter-dependent differential equation. An iterative method is used for the solution process, and the conditions ensuring its convergence are established. The proposed approach combines the numerical solution of Cauchy problems with the solution of a system of nonlinear equations. The developed algorithm is tested on specific benchmark examples, demonstrating its effectiveness. In particular, it is shown that the solution of the corresponding system converges through the iterative pro­cess for a given parameter value. The considered approach provides an efficient method for determining solutions of boundary value problems for families of linear differential equations characterized by nonlinear two-point boundary conditions. The practical efficiency of the proposed method is confirmed by the results of numerical solutions of specific problems.

This paper investigates a steady-state heat conduction problem in a multilayer cylindrical domain with nonhomogeneous boundary conditions. Two approaches for solving the direct problem are considered: an analytical method based on separation of variables and Fourier series expansion, and a numerical method based on the finite difference approximation combined with the Successive Over-Relaxation (SOR) algorithm. An exact analytical solution of the Laplace equation in polar coordinates is derived for a threelayer cylindrical system with continuity conditions at layer interfaces and convective heat exchange at the outer boundary. A finite d ifference sc heme is de veloped to ap proximate the go verning eq uation, wh ile the resulting system of linear algebraic equations is solved iteratively using the SOR method.The accuracy of the numerical approach is evaluated through comparison with the analytical solution and experimental temperature measurements. The obtained results demonstrate good agreement between the analytical and numerical solutions. The mean absolute error between the two approaches is 0.134K, the root mean square error is 0.141K, and the maximum absolute error does not exceed 0.209K. Experimental validation shows that both models correctly reproduce the overall temperature distribution in the multilayer cylindrical system. The mean absolute errors relative to experimental data are 2.509K for the analytical solution and 2.638K for the SOR solution. The performed analysis confirms the good a ccuracy and s tability of the developed finite difference algorithm and demonstrates its applicability for modeling heat transfer processes in multilayer cylindrical structures. The proposed approach can be employed in the thermal analysis of pipelines, geothermal systems, insulation structures, and other engineering applications involving layered media.

The necessity of studying boundary value problems for elliptic equations is motivated by their numerous practical applications in the theoretical study of processes in hydrodynamics, electrostatics, mechanics, heat conduction, elasticity theory, and quantum physics. For example, the distribution of the potential of an electrostatic field is described by the Poisson equation, while the vibrations of thin plates with small deflections are described by biharmonic equations In this paper, we study the solvability of one boundary value problem for the polyharmonic equation in the unit ball. In this problem, the normal derivatives of the unknown function are prescribed consecutively from the first-order normal derivative up to the (m 1)-th order derivative, while the (m + 1)-th order derivative is additionally prescribed as the last condition. To solve this problem, we reduce the original problem to the Neumann problem by means of a special integro-differential operator. Then, using another integro-differential operator, the Neumann problem is reduced to the Dirichlet problem. We use the fact that the solution of the Dirichlet problem can be represented by means of the Green function and the Almansi representation. As a result of these representations, the necessary and sufficient condition for the solvability of the original problem is obtained.

Selecting appropriate information security systems is a complex task influenced by organizational risks, infrastructure characteristics, evolving cyber threats, and regulatory requirements. Traditional approaches based on manual analysis and expert judgment often lack adaptability, scalability, and consistency in dynamic cybersecurity environments. This study proposes a hybrid intelligent decision-support system that combines fuzzy logic, machine learning, and ISO-based cybersecurity standards for risk assessment and security action recommendation. The fuzzy inference module evaluates cybersecurity events under uncertainty using parameters such as attack frequency, vulnerability level, and defense efficiency. Based on the calculated risk level, the system generates standardized response actions aligned with international ISO/IEC standards. In addition, a Random Forest machine learning model is used to evaluate whether the selected actions are likely to mitigate the attack successfully. Experimental evaluation performed on a synthetic dataset demonstrated that the proposed hybrid architecture achieves high predictive performance and improves both interpretability and adaptability of cybersecurity decision support systems in complex operational environments.

The rapid development of the Internet of Things (IoT) and embedded systems necessitates cryptographic solutions tailored for resource-constrained devices. Lightweight cryptography provides a balance between security, performance, and low power consumption, which makes it suitable for sensors, RFID tags, medical devices, and other IoT components. However, in practice, such algorithms are vulnerable to side-channel attacks that exploit leakage of power consumption, electromagnetic emissions, or timing characteristics to recover secret keys. The most threatening among them are differential and correlation power analysis techniques, which can compromise an implementation with a relatively small number of measurements. This article presents an overview of side­channel attacks targeting lightweight algorithms such as PRESENT, SIMON, SPECK, and PRINCE. Key protection techniques are examined, including masking, hiding, operation shuffling, threshold implementations, and hardware countermeasures. A comparative analysis of their effectiveness and overhead is provided, highlighting the strengths and weaknesses of each approach. Particular attention is given to the practical applicability of protective mechanisms under the resource constraints of IoT devices. The results demonstrate that no single countermeasure is universal; the most effective solutions are combined approaches that integrate masking with hiding techniques. The study emphasizes the need for further research aimed at developing robustness evaluation methodologies and improving protective mechanisms for lightweight cryptography.

This paper examines the dual role of generative artificial intelligence in cybersecurity: as an enabler of attacks and as a practical defense component. It is shown that realistic generation of text, images, and audio lowers the barrier for abuse (phishing, deepfakes, impersonation), while machine learning systems face a distinct risk class­small but targeted input perturbations. The practical part is implemented as a reproducible benchmark around an MNIST handwritten-digit classifier. After verifying stable performance on clean inputs, adversarial examples were generated using FGSM with multiple values of e. The results confirm that a visually subtle perturbation can flip a confident prediction to an incorrect class (in a representative case, a “7” was forced to be classified as “3”). Two reconstruction-based defenses were then compared: a standard autoencoder (AE) and a denoising autoencoder (DAE) trained to recover clean signals from noisy inputs. As e increases, DAE reduces the misclassification rate more noticeably than AE, which is consistent with its training objective of separating high-frequency noise from meaningful strokes. To make the defense actionable in an operational setting, a simple anomaly detector based on reconstruction energy E(x) was also introduced. This adds a second layer: the DAE attempts to restore the input for correct classification, while E(x) provides an explicit alert signal suitable for logging and incident review.

Integer division remains one of the most computationally demanding arithmetic operations in digital computing systems. Conventional sequential dividers produce one quotient bit per clock cycle, requiring n cycles for an n-bit division. This paper proposes an architecture for a sequential integer divider capable of simultaneously generating three quotient bits per clock cycle. The proposed device comprises a dividend register (DR), a divisor multiples generator (DMG), a partial remainder and quotient digit generator (PRQDG), and a synchronization unit (SU). Unlike classical SRT dividers, which rely on a redundant quotient representation and quotient-digit selection tables, the proposed architecture performs an exact comparison of the partial remainder against the precomputed divisor multiples B-7B, thereby eliminating the growth of the selection table as the radix increases. The architecture is described in Verilog HDL and verified on a Xilinx Artix-7 FPGA using the Vivado design suite. The prototype for a 12-bit dividend and a 6-bit divisor occupies 158 LUTs and 50 flip-flops, attains a maximum clock frequency of 134.6 MHz, and is confirmed by a suite of ten tests; a division is completed in two clock cycles. The architecture is intended for use in high-speed digital computing units and specialized processors.

Early identification of patients at high risk of in-hospital mortality is a persistent challenge in clinical practice, particularly during infectious disease outbreaks such as COVID-19, when decisions must be made at the time of hospital admission using incomplete information. Although machine learning methods have been widely applied to mortality prediction, many existing models are developed and assessed under conditions that limit their usefulness for early triage, including insufficient attention to probability reliability, temporal generalization, and decisionoriented assessment. This work examines early-stage mortality risk prediction using routinely available clinical data collected at the time of hospital admission. An ensemble of gradient boosting models is developed and analyzed under strict temporal separation to reflect real-world deployment conditions. Particular emphasis is placed on the reliability of predicted probabilities, with post-hoc isotonic calibration applied to improve alignment between predicted risks and observed outcomes. Model performance is assessed using complementary discrimination and calibration metrics, while clinical relevance is examined through decision curve analysis across plausible operating thresholds. Evaluation on an independent, temporally held-out test set shows that ensemble aggregation improves the stability of risk estimates, while calibration yields more consistent threshold-based behavior without altering ranking performance. Decision curve analysis indicates that calibrated predictions provide higher net benefit than default decision strategies across a broad range of early triage thresholds. These findings highlight that, in early clinical decision-making, probability reliability plays a critical role alongside discrimination. The presented framework offers a methodologically robust approach to early in-hospital mortality risk assessment under realistic clinical and temporal constraints.

This paper describes the design and testing of a modular docking station for a fixed-wing vertical take-off and landing drone to be used in conjunction with an unmanned ground vehicle. The solution enables autonomous engagement and disengagement of payloads through a controlled rail-guided system and visual docking with a quick response marker using a binocular camera module. A novelty of the proposed module lies in the fact that the primary mounting mechanism does not remain integrated with the drone after deployment; instead, it is designed as an integral part of the transport module itself and is released together with it. This approach distinguishes the system from conventional solutions in which the main attachment structure stays onboard the aerial platform, and it expands the flexibility of payload delivery and deployment scenarios. The payload assembly was designed in Autodesk Fusion 360 and manufactured in polyethylene terephthalate glycol-modified plastic using fused deposition modeling 3D printing on a Bambu Lab X1 Carbon printer. Structural analysis using finite element methods was performed to evaluate the displacement, stress, strain, and reaction forces of the module’s structure under operational loads. Advanced simulations suggest that the design satisfies mechanical protection requirements, additional proof will be gathered through flight testing. The system could be implemented to deploy unmanned ground vehicles from the air or for a drone to land on mobile ground vehicles, demonstrating its versatility in robotics and automation.

Morphological ambiguity constitutes a significant challenge for natural language processing in agglutinative languages, as a single word form might include many grammatical categories. The Kazakh language features productive suffixation, vowel harmony, and intricate morphophonological patterns, which considerably hinder automatic morphological analysis. This work presents a transformer-based methodology for morphological disambiguation in Kazakh texts, with the objective of identifying the appropriate morphological interpretation of word forms within context. A contextual language model tailored for Kazakh is refined for token-level morphological tagging utilizing a manually validated annotated corpus of news articles. The suggested method utilizes self-attention mechanisms to capture long-range contextual dependencies that are challenging to represent with conventional rule­based or recurrent neural techniques. The experimental assessment reveals that the transformer-based model attains superior accuracy and F1-score relative to rule-based morphological analyzers and BiLSTM-based benchmarks. The findings demonstrate that contextualized embeddings significantly enhance the resolution of morphological ambiguity, especially with homonymous suffixes and infrequent grammatical structures. The results validate the efficacy of transformer topologies for low-resource agglutinative languages and establish a feasible basis for incorporating morphology-aware models into comprehensive Kazakh natural language processing frameworks.

The article presents the results of the development and pilot testing of an adaptive learning system designed to develop and assess students’ critical thinking using elements of artificial intelligence. The scientific novelty of the study, in comparison with existing research on automated essay scoring and LLM-based assessment, lies in the multi-component assessment of critical thinking, the cascade mechanism of AI-based assessment, and the appeal mechanism. These features make it possible to more accurately identify students’ weak points in the development of critical thinking, ensure transparent assessment of students’ critical thinking, increase the system’s resilience to failures, and provide students with the opportunity to challenge the assessment in controversial situations. The system evaluates essays and problem-solving tasks according to six components of critical thinking: analysis, interpretation, evaluation, argumentation, logical coherence, and reflection. For each component, calculation formulas were proposed; these are computed using language models, semantic embeddings based on sentence-transformers, and a backup TF-IDF mechanism. The pilot testing was conducted on a sample of 20 students from Satbayev University. Three instructors, represented by the authors, participated in the expert assessment. The experimental data included 40 student works: 20 essays and 20 problem-solving tasks. Each work was independently assessed by three instructors as well as by the learning system. The reference score was determined as the average score of the three instructors. The correlation between the automated assessment and the averaged expert assessment was r = 0.84, while the mean deviation was 1.0 point on a ten-point scale. The results obtained indicate that the system can be used as a tool to support instructors; however, further research on a larger sample is required. Gamification elements, progress visualization, and the appeal mechanism increase students’ engagement and ensure a balance between automated and expert assessment.

This paper presents and experimentally validates an electronic voting model that stores results in a decentralized blockchain while using biometric voter verification and automatic detection of synthetically generated (deepfake) facial images. For identity verification, Vision Transformer and ResNet-50 architectures are fine-tuned with LowRank Adaptation (LoRA). Frequency-domain information from the Fast Fourier Transform (FFT) is also included as a descriptor for detecting synthetic images. Since blockchain records cannot be altered, the model introduces a controlled revoting process based on self-destructing ballots and timestamps recorded on the blockchain; the votes themselves are submitted through an Ethereum smart contract. In the experiments, the LoRA-adapted ViT configuration distinguishes real from synthetic facial images with an accuracy of 97.48%, exceeding the performance of the LoRA-adapted ResNet-50 baseline. However, because no separate FFT ablation study has been conducted, this work does not claim that the frequency descriptor makes an independent contribution. The proposed system is presented as part of Smart City digital infrastructure and lays the groundwork for future research on privacypreserving biometric protocols and blockchain platforms capable of scaling to electoral processes.

The increasing use of centralised Student Information Systems (SIS) as part of the digitalisation of higher education has led to these systems being used despite the risk of unauthorised alteration of student data due to reliance of data entry and administration on one administrative trust domain. This paper presents the design and experimental validation of a Distributed Ledger Technology (DLT) architecture to provide verifiable integrity of examination results. The proposed architecture is based on a combination of an SQL-based SIS and a Hyperledger Fabric Blockchain via a secure middleware layer. Examination records in the SIS are converted to a cryptographic hash commitment and stored on an immutable ledger. Experimental validation was carried out by deploying a Hyperledger Fabric Network in a containerized environment and testing it using Hyperledger Caliper at workload levels up to 120,000 transactions. Key performance metrics were throughput, commit latency, ability to detect violations of data integrity and storage characteristics of the ledger. The proposed architecture achieved throughput of 728 transactions/second and maximum throughput of 842 transactions/second with average commit latency less than 2.0 seconds. Tampering tests were performed to verify the architecture’s ability to detect unauthorised changes and it successfully detected all unauthorised changes with average detection time of 19.8ms. Analysis of storage requirements showed that the ledger grew linearly and would require less than 5.68 GB over ten years. The results demonstrate that the proposed architecture eliminates the single point of failure in centralized systems and provides scalable, cryptographically verifiable academic record integrity.

This article addresses the problem of improving face detection reliability in video streams used in driver state monitoring systems. Reliable localization of the driver’s face is considered a preliminary computer vision stage required for the subsequent analysis of eye state, gaze direction, facial behavior, and fatigue-related visual cues. The purpose of this study is to develop and experimentally evaluate an adaptive face detection method that maintains stable performance under variations in illumination, head pose, image noise, and partial face occlusion while operating with a fixed, externally calibrated camera. The proposed method combines initial face-region detection using the Viola–Jones algorithm with SVM-based verification, external camera calibration, and incremental parameter updating based on high-confidence detections used as pseudo-labels. The method was evaluated using 18,000 video frames obtained from 12 recordings involving seven volunteers under different shooting conditions. The experimental results showed that the integrated method provided higher face detection performance than the standalone Viola–Jones algorithm and the Viola–Jones method combined with a conventional SVM classifier. The proposed method achieved a Precision of 0.91 and a Recall of 0.90 while maintaining a processing speed of 22 frames per second, which is sufficient for real-time operation. The study evaluates face detection and does not directly classify driver fatigue, drowsiness, eye state, or facial expressions. The obtained results demonstrate that the proposed method can be used as a lightweight preliminary component of driver state monitoring systems operating on embedded hardware with limited computational resources.

This article addresses the development of a biometric personal identification system based on palm parameter analysis using machine learning technologies. The relevance of the study is determined by the need to ensure information security, access control, and reliable user identification. Since the palm as a biometric identifier possesses a high degree of individuality, its anatomical features–such as finger contours, skin texture, palm lines, and blood vessel patterns–are used as effective characteristics for personal recognition. In the course of the study, the ROI-LANet model was employed to automatically identify significant biometric regions in palm images. This model enables accurate localization of key palm points, image alignment, and extraction of regions of interest (ROI). User classification based on the extracted ROI was performed using the deep convolutional neural network ResNet18. The residual block-based architecture allows efficient recognition of complex biometric features and improves classification accuracy. Experimental results demonstrated that the proposed system achieves high accuracy and stability. During testing, the identification accuracy reached approximately 97%. In addition, the system showed robustness to variations in lighting conditions and image acquisition angles. The obtained results indicate a high potential for practical application of the developed biometric system in security systems, access control, and transaction verification.

The article presents the development of an integration model based on knowledge graphs and machine learning methods for analyzing relationships between educational programs, labor market requirements, and online courses (MOOCs). The relevance of the study is driven by rapid changes in the information technology sector, which lead to a mismatch between graduate competencies and employer expectations. To reduce this gap, a heterogeneous knowledge graph integrating data from university curricula, job vacancies, and online courses is proposed. The model implements a meta-path connecting skills, job requirements, and learning courses, enabling recommendations for educational and career development for different categories of users. Multilingual transformers were applied to process textual descriptions in Russian and English. The data sources include competency maps of educational programs, job postings from the hh.kz platform, and courses from Coursera, all integrated into the Neo4j graph database. The obtained results demonstrate the capability to identify missing competencies and automatically recommend relevant courses for their acquisition. The proposed model can be applied in curriculum design, development of learning trajectories, and decision-making support within educational environments.

This paper presents a modified SEIR epidemiological model integrated with machine learning techniques to improve COVID-19 transmission dynamics forecasting. The proposed approach extends the classical framework with time-dependent parameters capturing the effects of social distancing, vaccination coverage, and population mobility. The Grid Auto-Search methodology enables automatic parameter optimization and adaptive trainingperiod selection. XGBoost and LSTM algorithms are applied to forecast the time-varying transmission coefficient β(t). The dataset comprises 1,143 daily observations for France from three open sources (Johns Hopkins University, Our World in Data, Google Mobility). The model achieves R² up to 0.96 across 30–90-day horizons; crossvalidation confirms the absence of significant overfitting. The results support the practical applicability of this hybrid approach for healthcare decision-support systems. The research results are confirmed by two Qazpatent certificates of copyright registration.

Deep learning methods and transformer architectures have fundamentally reshaped the field of natural language processing; however, these advancements remain unevenly distributed. While high-resource languages like English benefit from large-scale benchmarks, robust text classification for low-resource languages, such as Kazakh, continues to pose a significant challenge. This paper presents a comparative analysis of the classical Word2Vec embedding, BiLSTM recurrent models, BERT and XLM-R transformers, the generative mT5 model, alongside classical and statistical baselines for few-shot Kazakh news text classification under limited annotation constraints. Experiments were conducted on the KazNews dataset, curated for few-shot classification from tengrinews.kz and inform.kz articles, as well as parallel HuffPost corpora. The datasets comprise five categories (sports, politics, business, travel, etc.). The experimental findings in 1-shot and 5-shot settings demonstrate that no single model consistently dominates across all datasets and regimes: Word2Vec achieved the best performance on the HuffPost corpora in the 5-shot regime (0.48 and 0.53); mT5-Seq2Seq outperformed others on KazNews under 5-shot (0.71); while BERT remained competitive on the long-text KazNews dataset in the 5-shot setup. The authors also observed a sharp performance gain for the mT5-Seq2Seq model in the 5-shot setup. Conversely, none of the evaluated models yielded acceptable quality metrics in the 1-shot regime.

The widespread use of Quick Response (QR) codes has increased exposure to QR-based phishing, malicious redirection, and malware distribution because users cannot readily inspect the destination before scanning. This study develops aQRace, a machine-learning-based QR code security application that analyses embedded Universal Resource Locators (URLs) before users access their destinations. The study combines URL-based feature extraction, comparative machine-learning evaluation, and operational system integration. URL characteristics covering structural, linguistic, technical, and advanced properties were extracted for classification. Four supervised machine-learning algorithms–XGBoost, Random Forest, Decision Tree, and Logistic Regression–were trained and evaluated using a balanced dataset of 14,000 URLs comprising 7,000 safe and 7,000 malicious samples. Model performance was compared using accuracy, precision, recall, and F1-score. XGBoost achieved the highest accuracy of 84% and precision of 0.80, whereas Random Forest achieved the highest recall of 0.95. Both models achieved an F1 Score of 0.84. XGBoost was selected for deployment based on its overall performance profile and integrated into a client-server application comprising a Flutter mobile frontend and Flask cloud backend. The application supports QR code image uploads and camera-based scanning, and performs URL processing, feature extraction, classification, and user warnings before destination access. Functional testing with representative safe and malicious QR codes confirmed the operation of the complete detection workflow. The findings demonstrate that conventional machine learning using URL-based features can support pre-access QR code security assessment while providing an operational alternative to increasingly complex deep learning approaches.

Precision agriculture requires efficient monitoring of large agricultural fields with the aim that crop anomalies can be identified and spatially located as early as possible to enable targeted and timely field management. The traditional in-field inspection is time-consuming, tedious and measures only a small number of points. Deep learning technologies are coupled with unmanned aerial vehicles (UAVs) to create the opportunity to automate highresolution agricultural imagery. This work examines a task specific U-Net approach to multi label segmentation of agricultural anomaly patterns, but not a novel network architecture, and rather emphasizes the impact of practical design decisions. The evaluated configuration has four-channel RGB-NIR (NRGB) input, uses sigmoid-based multilabel prediction, and uses a combined Binary Cross-Entropy (BCE) and Dice loss. The experiments are carried out on the Agriculture-Vision 2021 benchmark dataset with over 21,000 labeled aerial images of eight agriculture anomaly classes. To isolate the contributions of NIR channel and Dice loss component, a controlled ablation study is performed, and additional experiments compare the resulting configuration to an identical experiment protocol against the standard U-Net, FCN, SegNet and a modern segmentation baseline. The performance is measured by Intersection over Union (IoU) and mean Intersection over Union (mIoU) on each class and the overall mean, respectively. Results show the effectiveness of NIR information and Dice-based loss for anomaly segmentation in agriculture and the performance of a relatively simple U-Net-based configuration on the benchmark dataset. The results validate the potential of using task-specific U-Net structure for automated UAV monitoring in agriculture and demonstrate individual input and loss-function contribution to the task.

In this study, a porous, hierarchical flower-like zinc oxide (ZnO) structure was successfully synthesized via a hydrothermal method using urea and CTAB as shape-directing agents to optimize photocatalytic efficiency for environmental remediation. Structural characterization via scanning electron microscopy (SEM) and X-ray diffraction (XRD) confirmed the formation of a highly crystalline, pure hexagonal wurtzite phase assembled from ultrathin nanosheet building blocks. The photocatalytic performance of the synthesized ZnO was evaluated through the degradation of methylene blue (MB) dye under UV light illumination (365 nm). The catalyst achieved a (98.7 ± 0.5)% degradation efficiency within 40 minutes, following a pseudo-first-order kinetic model with a reaction rate constant of 0.109 min-1. Optimization studies identified a catalyst dosage of 0.2 g/L as near-optimal, while coexistence tests demonstrated that the system maintains high stability and excellent removal efficiency in the presence of various competing environmental anions (Cl- , NO3 - , HCO3 - , CO3 2-). Radical scavenger investigations revealed a clear inhibition hierarchy, establishing that the degradation mechanism is predominantly driven by hydroxyl radical-mediated oxidation alongside secondary contributions from photogenerated holes. These findings highlight the prepared porous ZnO architecture as an efficient, stable, and sustainable candidate for the degradation of persistent organic pollutants in municipal wastewater.

The paper presents the results of an experimental investigation of changes in the microhardness and mass of a preheated tungsten target, the primary candidate material for the first wall and divertor, subjected to a pulsed plasma flow of approximately 0.2 ms duration. The experiments were conducted as a series of 13 pulses, each with an energy density and power density of approximately 1.1 MJ/m² and 5.5 GW/m² , respectively. Microhardness was measured using an HVS-1000B hardness tester; the target mass before and after plasma exposure was measured using a CPA225D precision balance. It is shown that the microhardness of the tungsten target remains virtually unchanged even at the maximum irradiation dose (up to 39 pulses), and the observed change is within the measurement error, which indicates the structural stability of the surface layer. At the same time, an increase in the target mass (up to 31 mg) was recorded, probably caused by contamination and redeposition of the material of external structural components. Energy-dispersive spectroscopy (EDS) analysis revealed Cu, Zn, Ni, and C lines in dust particles redeposited on the target surface. The results indicate that these foreign components have a significant impact on surface morphology and should be considered when interpreting tungsten damage mechanisms.

This study presents a mathematical and empirical analysis of gradient boosting algorithms applied to thermogravimetric analysis (TGA) data for composites based on vinyl ether matrices and epoxides with mineral fillers. The gradient boosting algorithm is output step by step with a numerical example of TGA. XGBoost and CatBoost are then analyzed formally. 12 algorithms were compared on 478 data from the peer-reviewed literature using 5-repeated 5-fold cross-validation. Extra Trees and CatBoost achieved statistically equivalent top performance (mean R² = 0.7430 ± 0.0879 and 0.7313 ± 0.0793, respectively; Nadeau & Bengio corrected t = 0.483, p = 0.634), outperforming all linear baselines, SVR, and MLP. CatBoost is recommended for practical deployment: its ordered boosting produces unbiased gradient estimates on small datasets (n < 1000), and SHAP analysis of its gradient tree structure reveals physically interpretable feature attributions aligned with known TGA mechanisms. SHAP values identified TGA peak temperature (T_max) and matrix type as the dominant predictors. The proposed preprocessing pipeline including phr to wt% conversion, label encoding, and median imputation provides a reproducible strategy for heterogeneous TGA literature datasets.

The development of porous TiO₂ thin films is of significant interest for photovoltaic and photocatalytic applications owing to their ability to enhance light utilization. Herein, porous TiO₂ films were prepared using 350 nm polystyrene (PS) particles as sacrificial templates, and the effect of PS concentration on photocurrent generation was systematically evaluated. Morphological and optical analyses confirmed the successful formation of porous structures with improved light-harvesting characteristics. Typically, the optimized PS-modified TiO₂ film exhibited a photocurrent density of ~118 μA/cm² , corresponding to an enhancement of ~20.5% relative to bare TiO2 film. Integrating sphere measurements revealed increased total absorbance for the porous films, indicating that the improved photocurrent density generation originates from enhanced light trapping within the porous surface. This work highlights the importance of porosity engineering and demonstrates that PS-assisted templating is an effective approach for enhancing the performance of TiO2 -based photoactive thin films.

Hydrogen evolution reaction (HER) electrocatalysts based on earth-abundant materials are promising alternatives to noble metal catalysts. In this work, WS₂ thin films were deposited on titanium substrates by RF magnetron sputtering with deposition times ranging from 30 to 120 min to investigate the influence of film thickness on structural and electrochemical properties. In addition, Ti/CNWs/WS₂ composite electrodes were fabricated by combining vertically aligned carbon nanowalls (CNWs) with a WS₂ catalytic layer. Raman spectroscopy confirmed the successful formation of multilayer WS₂ films. Electrochemical measurements revealed that HER performance strongly depends on the WS₂ deposition time. Among the investigated samples, the Ti/WS₂ electrode prepared with a deposition time of 90 min exhibited the highest catalytic activity, delivering an overpotential of 294 mV at 10 mA cm^-2 and the lowest charge transfer resistance (Rct = 15.9 Ohm) within the Ti/WS₂ series. Following electrochemical activation, the overpotential further decreased to 144 mV. Electrochemical impedance spectroscopy (EIS) showed that the incorporation of CNWs significantly enhanced charge-transfer kinetics, reducing the Rct to 3.7 Ohm. However, despite the improved electron transport, the Ti/CNWs/WS₂ electrode did not outperform Ti/WS₂ in HER activity, indicating that catalytic performance depends not only on charge-transfer properties but also on the density and accessibility of active sites. These findings highlight the importance of optimizing WS₂ film thickness and electrode architecture for efficient hydrogen production.

This work investigates the effect of tantalum carbide and vanadium carbide additions on the phase composition, microstructure, and mechanical properties of AlN-based ceramic composites synthesized by mechanochemical solid-state processing followed by thermal treatment. The molar fraction of the carbide additives was varied within the range of x = 0.01-0.10. Scanning electron microscopy combined with energy-dispersive X-ray spectroscopy revealed a homogeneous distribution of the reinforcing phases after mechanochemical activation. X-ray diffraction analysis showed that TaC remained chemically stable during processing, leading to the formation of a two-phase AlN-TaC composite with reduced lattice distortions of the AlN matrix. In contrast, VC underwent partial oxidation during heat treatment, resulting in the formation of secondary V₂ O₅ and a different phase evolution. Mechanochemical processing produced a more homogeneous powder morphology and suppressed the formation of large carbide agglomerates, providing favorable conditions for subsequent microstructural development. Mechanical testing demonstrated that a pronounced improvement in hardness and flexural strength occurred at dopant contents above x = 0.03, with the greatest strengthening effect observed for the TaC-containing composites owing to the retention of a stable, uniformly dispersed high-hardness carbide phase. The enhanced mechanical performance is attributed to the combined effects of dispersion strengthening, inhibition of grain growth, grain-boundary strengthening, and crack-deflection mechanisms. These results establish clear relationships between processing, phase evolution, microstructure, and mechanical behavior, demonstrating the potential of mechanochemical processing for the development of high-performance AlN-based ceramic composites intended for demanding structural and hightemperature applications.

Amorphous carbon nitride (a-Cx Ny ) thin films were synthesized by plasma-enhanced chemical vapor deposition (PECVD) in a nitrogen–methane radiofrequency (RF) discharge plasma (13.56 MHz) using pulsed plasma modulation. The evolution of the DC self-bias voltage (Vdc) was monitored as an in situ diagnostic parameter to track nanoparticle nucleation, growth, and agglomeration during synthesis. It was found that the absolute value of Vdc decreases monotonically following plasma ignition, and that the onset of nanoparticle agglomeration can be reliably identified from a characteristic change in the Vdc(t) curve, consistent with the three-stage model of dustparticle formation in low-temperature plasma. The agglomeration onset time (tagl) was shown to depend strongly on methane flow rate, working pressure, and RF power: at baseline conditions (0.86 mbar, 20 W), increasing the CH4 flow rate from 2 to 7 sccm reduced tagl from 4.4 s to 200 ms, while pressure and power exhibited dominant and nonmonotonic effects, respectively. These findings guided the selection of plasma-on pulse durations for controlled film growth. Samples deposited under pulsed modulation were characterized by SEM, EDS, FTIR, and XRD, confirming the formation of continuous, nonstoichiometric carbon nitride films dominated by C–N and C=N bonding. The results demonstrate that self-bias voltage monitoring is an effective, accessible method for in situ control of a-Cx Ny nanoparticle growth kinetics and for tailoring film thickness, morphology, and chemical structure via pulsed RF plasma parameters.

This paper presents an automated pipeline for wavelength calibration of slit-based astronomical spectra, combining robust peak-matching techniques with a lightweight preprocessing step. The method is based on initial peak alignment using Dynamic Time Warping (DTW), followed by refinement of the dispersion solution with the Random Sample Consensus (RANSAC) algorithm. To ensure consistency between observed and reference spectra, a compact one-dimensional convolutional neural network (CNN) is employed at the preprocessing stage to identify the spectral range and corresponding diffraction grating. This step enables the selection of an appropriate reference spectrum but does not directly affect the calibration procedure. The DTW-based approach provides reliable initial correspondences between spectral peaks under nonlinear distortions and noise, while RANSAC filtering removes incorrect matches and yields a robust polynomial dispersion solution. The method is validated on real observational data from calibration lamps and demonstrates stable performance, high matching accuracy, and full automation capability. Compared to traditional manual or semi-automatic approaches, the proposed pipeline significantly reduces processing time while maintaining calibration precision. The developed solution is planned to be deployed as a service in the virtual observatory infrastructure and is applicable for processing numerous spectral data.

This paper presents the results of studies on Y₃Al_xGa_5-xO₁₂: Ce (YAGG:Ce) ceramic phosphors synthesized for the first time using a radiation method. The influence of the Al/Ga ratio on the optical properties of Ce³+doped Y₃Al_xGa_5-xO₁₂: Ce (YAGG:Ce) garnet phosphors, which are promising for LED light source applications, was investigated. Excitation and photoluminescence spectra of samples with different aluminum and gallium contents at a fixed concentration of Ce³+ activator ions were analyzed. It was shown that variations in composition lead to changes in the position and shape of the excitation and emission bands, variations in the photoluminescence band width, as well as redistribution of the emission intensity under excitation in the ultraviolet and blue spectral regions. It was found that a decrease in the Al/Ga ratio causes the excitation band to shift toward the longer wavelength region near 340-370 nm, which increases the excitation efficiency when near-ultraviolet and violet LEDs are used. The obtained results confirm the feasibility of using the radiation method for the synthesis of efficient YAGG:Ce phosphors.

This study investigates the influence of spin-coated phosphosilicate glass (PSG) on the crystallization and electrical properties of silicon (Si) thin films deposited by magnetron sputtering, with a focus on optimizing their structural and electrical properties for potential applications in photovoltaics and microelectronics. The results demonstrate that higher spin-coating speeds lead to thinner and more uniform PSG layers, which in turn promote more efficient phosphorus diffusion and enhance film crystallization. Annealing at 950 °C resulted in enhanced dopant activation, improved crystallinity, and reduced sheet resistance. These structural and phase transformations were systematically analyzed by X-ray diffraction (XRD), providing detailed insight into the evolution of the Si thin films during thermal processing. In particular, the XRD results reveal a clear transition from the amorphous to the polycrystalline state, especially for samples processed at optimized spin speeds and elevated annealing temperatures. Optical characterization further showed that thermal annealing modified the reflectance and transmittance spectra and slightly increased the refractive index and extinction coefficient of the sputtered Si films. Overall, the combination of environmentally friendly spin-on doping with scalable physical deposition techniques represents a promising approach for the development of advanced photovoltaic technologies, offering a viable pathway toward sustainable and efficient device fabrication.

This article proposes a unified exergetic methodology for comparable, reproducible assessment and subsequent standardization of plasma processes. Based on the Second Law of Thermodynamics, it integrates exergy efficiency (Яех), specific energy consumption (Q_sp), exergy ignition cost (EIC), and environmental performance index (EPI) into a formalized KPI-stack. Unlike First Law criteria, the approach accounts for energy quality, irreversibility, and the non-equilibrium nature of plasma, distinguishing chemical, physical, and ionization components and quantifying exergy destruction at individual process stages. The methodology covers plasma ignition, gasification, reforming, and processing of gaseous, liquid, and solid hydrocarbon feedstocks and waste. Measurement, uncertainty assessment, and quality control are aligned with cited metrological requirements; environmental comparisons may incorporate life-cycle assessment. An oil-sludge case study compares conventional combustion at 1100 K with plasma-supported combustion at 1500-2000 K at a fixed process thermal power of 2000 kW. Calculations show differential exergy efficiency increasing from 0.546 to 0.636-0.675 and specific exergy losses decreasing from 18.27 to 14.48 MJ/kg. Environmental performance improves through a higher EPI and lower NO? emissions. Beyond thermal disposal, plasma processing produces valuable synthesis gas with total exergy reaching 156.7 MJ. The unified KPI-stack supports technology comparison, operating-mode optimization, and integration into regulatory and technical documentation.

Despite a significant amount of research on the effect of nanoparticles on the stability of gas foams and the coefficient of reduced mobility (MRF), there are no uniform recommendations in the literature on the optimal range of nanoparticle concentrations, taking into account a set of reservoir parameters, including temperature, pressure, core permeability and salinity of reservoir fluids. In addition, the effect of nanostabilized foams on oil displacement processes and interaction with the oil phase has not been sufficiently studied, and on the scale of real deposits, the effectiveness of nanofoams (NAF) at various levels of reservoir heterogeneity has not been practically performed. The present study aims to address these gaps through a combination of laboratory experiments and numerical simulations. The paper presents the results of a comprehensive study of the physico-chemical characteristics of composite systems based on surfactants and silicon dioxide nanoparticles, aimed at increasing the stability of gas foams used in oil recovery enhancement processes. It has been shown that traditional surfactant foams exhibit limited stability at high temperatures, pressures, and mineralization, which significantly reduces their effectiveness in real formations. To solve this problem, the effect of SiO₂ nanoparticles on the interfacial properties of such systems, including anionic, cationic, and nonionic surfactants, has been studied. Based on experimental data, modifying coefficients for the foam half-life and MRF have been determined, which are integrated into the simulation model of oil displacement. Numerical modeling has shown a significant increase in reservoir coverage and delayed gas breakthrough when using nanocomposite foams, especially in conditions of high heterogeneity. The results obtained confirm the prospects of nanostabilized foams as effective MUN agents and substantiate their use to increase oil recovery in complex fields.

The “NARYN” geoelectric model, which runs along the 76° E meridian through the Central Tien Shan from the Kazakh Shield to the Tarim Platform, is one of the longest cross-sections in continental Asia. This paper provides a historical overview of geoelectric models based on magnetotelluric (MT) soundings from 1999-2000 and 2005. Two inversion models constructed from these data using different algorithms are presented: the discrete-smoothed NARYN-RLM model and the piecewise continuous NARYN-INV2D model. A local earthquake catalog for 1980­2022 is considered. Hypocenter distributions from four regional and global seismic catalogs are superimposed on a reference cross-section to re-examine the previously identified relationship between crustal conductivity and earthquake depth distribution. It is noted that weak seismicity is mainly concentrated above the upper boundary of the middle and lower crustal conductor, while the strongest events are grouped in resistive, consolidated blocks of the earth’s crust or at their contacts with conductive zones.

The effectiveness of hydraulic fracturing (HF) largely depends on how well fracture conductivity is preserved after the fracturing fluid interacts with the reservoir rock. This paper systematizes the permeability-damage mechanisms related to the water sensitivity of clay minerals interlayer smectite swelling, osmotic effects, deflocculation and fines migration and examines the non-target consumption pathways of clay stabilizers (KCl, quaternary ammonium compounds) and gel breakers (ammonium persulfate, hydrogen peroxide, enzymes) on rock minerals and co-additives. It is shown that the nominal reagent dosage, evaluated without a material balance, can simultaneously overestimate both the degree of clay stabilization and the extent of polymer network degradation. A minimum laboratory testing matrix is proposed, including oriented-sample XRD analysis, sorption isotherms, coreflood tests and retained fracture-conductivity tests. Using the Vostochny Moldabek field (Embamunaigaz JSC) as a case study, it is shown that the permeability decline observed when the reservoir is saturated with low-salinity water is caused by clay and mica swelling, whereas formation water does not cause such degradation. A formation­water-based fracturing technology is substantiated and implemented, using a polyacrylamide-based gel instead of guar gum, which is unsuitable at a water salinity above 130 g/L. Field results from eight wells show that production rates after fracturing on formation water with the synthetic polymer exceeded those of the standard technology by more than five times.

Permanent geological storage of anthropogenic CO₂ and the development of low-carbon energy resources, including natural hydrogen (H₂), have become two key components of carbon-neutrality strategies. Ultramafic rocks are increasingly recognized as promising geological media because they can support both mineral carbonation and natural H₂ generation. However, the relationship between these processes, and the extent to which similar reactions occur in mafic rocks, remain poorly understood. Mafic and ultramafic rocks from the Charsk Ophiolite (eastern Kazakhstan) and the Kempirsai Ultramafic Massif (western Kazakhstan) were investigated under experimental conditions to evaluate their potential for CO₂ mineralization and natural H₂ generation. Serpentine-rich ultramafic rocks displayed the highest carbonation potential. Dissolution of lizardite released Mg²⁺ ions that promoted secondary magnesite precipitation and efficient mineral trapping of CO₂. Carbonation was substantially weaker in magnesiochromite-rich ultramafic rocks. Natural H₂ generation followed a different pattern. The highest H₂ concentrations were measured in ultramafic (0.872 mol%) and mafic (0.721 mol%) rocks from the Charsk Ophiolite, whereas only trace amounts were detected in samples from the Kempirsai Massif. CO₂ mineralization and natural H₂ generation were controlled by different mineralogical factors. Serpentine-rich ultramafic rocks were most favorable for long-term CO₂ immobilization through mineral carbonation, whereas Fe-bearing mafic and ultramafic rocks with high reaction potential were more favorable for natural H₂ generation. Kazakhstan’s ophiolitic complexes therefore represent promising geological targets for both permanent CO₂ storage and natural hydrogen exploration.

Enhanced oil recovery (EOR) selection requires reservoir performance simulation that reproduces not only production trends but also the physical mechanisms controlling pressure, fluid behavior, multiphase flow, gas mobility, well constraints, and long-term reservoir response. This study presents a reservoir performance simulation and process-to-simulator quality-control workflow for CO₂-based EOR evaluation in anonymized carbonatecontaining Field X, Kazakhstan. The field-scale simulation uses a three-dimensional ECLIPSE E300 compositional model with 93 × 88 × 78 grid cells, 194,413 active cells, and a 40-year forecast period. The framework integrates compositional pressure-volume-temperature/equation-of-state (PVT/EOS) behavior, minimum miscibility pressure (MMP), three-phase relative permeability, hysteresis, vertical connectivity and gravity override, injection scheduling, bottomhole-pressure (BHP) control, wellbore thermal response, CO₂ tracking, and mass-balance verification. Each mechanism is linked to a simulator functional block, diagnostic output, sensitivity test, and acceptance criterion. A representative ECLIPSE E300 process-to-code sequence – RUNSPEC, GRID, PROPS, SOLUTION, SCHEDULE, SUMMARY, followed by MATLAB post-processing – provides traceability from reservoir physics to forecast interpretation. Sensitivity analysis identifies vertical connectivity, gas relative permeability, MMP uncertainty, fault transmissibility, hysteresis, and BHP constraints as the principal controls on simulated reservoir performance. Continuous CO₂ injection increases cumulative oil production from 17.96 to 25.37 Mt, yielding 7.41 Mt of incremental oil (+41.3%); at year 40, the CO₂ case maintains 238.6 t/d while the Base case declines to zero. The results demonstrate that reservoir performance simulation provides a rigorous engineering basis for forecasting reservoir behavior, quantifying model sensitivity, validating physical consistency, and supporting defensible EOR method selection.

The article presents the results of the industrial application of gel polymer treatment of injection wells to enhance the oil recovery of high-viscosity oil in one of the fields of the Mangystau region (Kazakhstan). The theoretical foundations of the formation of the spatial gel structure of polyacrylamide in interaction with trivalent chromium ions are considered. The results of treatment of four injection wells with a viscoelastic compound (VC) based on polyacrylamide grade FP-307 and chromium acetate with a total volume of 300 m3 each are presented. It has been shown that the injection of gel–polymer compositions makes it possible to redistribute filtration flows in the bottomhole zone, involve low-permeability layers in development, and increase the oil flow rate of responding offset production wells by an average of 0.3-0.6 t/day while reducing the produced-fluid water cut by 2–10 percentage points. The duration of the effect on individual wells exceeds 200 days. A comparative analysis was carried out with global polymer flooding projects, including Pelican Lake (Canada), Daqing (China) and Uzen (Kazakhstan) fields. The technology has shown high adaptability and the absence of critical injection complications associated with premature gelation in the equipment, which confirms the correctness of the selected rheological parameters of the composition for reservoir conditions. The results obtained confirm the expediency of using gel polymer technologies to recover hard-to-recover viscous oil reserves from heterogeneous reservoirs of viscous oils from heterogeneous reservoirs.

The purpose of the study is to develop an adaptive mathematical model for forecasting financial markets based on the Financial instruments exhibit the greatest sensitivity to trading volumes (β=0.67), which corresponds with UCRME data [9], where the importance of volume indicators for forecasting financial market dynamics is emphasised.principles of self-organisation. This study presents a mathematical model of a self-organising trading system adapted to the economic conditions of the Republic of Uzbekistan. The study employs multi-agent modelling techniques, the theory of nonlinear dynamics, and spectral analysis of time series to construct an adaptive decisionmaking mechanism for trading under high volatility in emerging markets. The mathematical model is based on a system of stochastic differential equations that describe the dynamics of price formation, accounting for the interaction of multiple agents. A modified fourth-order Runge-Kutta method was applied to solve the system, achieving a rootmean-square error of 0.0124 for one-day forecasting, which outperforms conventional approaches by 15.6%. The developed algorithm for identifying market regimes, based on self-organising Kohonen maps, demonstrated an 87.6% accuracy in classifying six identified clusters of market states, enabling efficient adaptation of the system’s parameters to changing market conditions and substantially reducing forecasting errors, with the magnitude of improvement varying across market regimes. Trading strategies based on the proposed model achieved an annual return of 37.2% with a Sharpe ratio of 1.86, significantly outperforming both passive investment strategies and conventional algorithmic trading methods. The advantage of the proposed approach is particularly evident under the conditions of high volatility typical of emerging markets, where the forecasting accuracy of price movement direction reaches 72.8% for a one-day horizon. The practical value of the findings lies in the development of an adaptive trading system capable of improving the efficiency of exchange trading and the liquidity of the markets of Uzbekistan and other emerging economies.

Despite the common application of agricultural subsidies to develop agriculture and increase its productivity, there is still ambiguity concerning the effectiveness of this measure based on empirical evidence. Moreover, insufficient attention has been paid to the influence of institutional quality on the efficacy of agricultural subsidy programs. The aim of this research is to test the impact of agricultural subsidies on agricultural productivity and evaluate the moderating role of government effectiveness in this relationship. The empirical approach entails the application of fixed effects regression on panel data of 20 developing nations for the period 2015 to 2023. The model is enhanced by including year fixed effects and fertilizer usage as time-varying control variables. Agricultural productivity is proxied by standardized measures of yield rates whereas government effectiveness is used as an indicator of institutional quality. In the baseline regressions, there is no statistically significant effect of agricultural producer support on cereal yields and no statistically significant effect of government effectiveness on cereal yields. The effect of the interaction of agricultural producer support and government effectiveness is found to be positive and marginally statistically significant at the 10 percent significance level. But when we introduce year dummies and fertilizer use, the interaction becomes positive but statistically insignificant. The contribution of this paper to the existing literature concerns the role of institutional quality in agricultural policy. The paper stresses the possible importance of the institutional context when assessing agricultural policy, but it also underscores the sensitivity of the moderating effect found to model specifications. Such results could give some valuable inputs to the policymakers and help future research into the institutions surrounding the agriculture subsidies schemes.

This study examines the main drivers and barriers influencing the adoption of artificial intelligence (AI) in project management, with a specific focus on emerging market conditions. The research is based on a systematic literature review conducted in accordance with the PRISMA 2020 guidelines. The final sample included 41 peer-reviewed articles and conference papers retrieved from the Scopus database for the period 2016–2025. The identified adoption factors were coded and grouped using the Technology–Organisation–Environment (TOE) framework. As a result, twenty-one driver categories and twenty barrier categories were identified. The findings show that AI adoption in project management is not limited to the availability of advanced tools. Organisations are often interested in AI because of its potential to improve project performance, yet practical implementation is constrained by weak data infrastructure, limited technical maturity, shortage of qualified specialists, and unclear regulatory conditions. This gap is particularly visible in emerging markets, where environmental factors such as digital infrastructure, regulation, standards, and institutional support often determine whether firm-level adoption is possible at all. The study argues that AI adoption strategies should therefore be adapted to the maturity of the local context. For emerging markets, this means that policy and organisational efforts should not begin only with firm-level training or software implementation, but also with the development of basic digital, regulatory, and institutional conditions that make sustainable AI adoption feasible.

This study examines how inequality in access to research and development (R&D) is represented within the global research landscape and why it remains weakly consolidated as an independent analytical category. Drawing on bibliographic records retrieved from Scopus and the Web of Science Core Collection for 2014 – 2025, the study applies author – keyword co – occurrence analysis to examine the thematic structure of research related to regional innovation, spatial inequality, economic geography, and research capacity. The integrated dataset contains 7,592 records. Using a minimum occurrence threshold of five, 106 author keywords were retained and organized into nine thematic clusters. The resulting network is dominated by established themes such as regional development, peripheral regions, core – periphery relations, regional innovation, and spatial inequality, while terminology explicitly referring to R&D access appears comparatively weak and fragmented. The findings therefore provide bibliometric evidence of limited thematic consolidation around R&D access rather than establishing it as an empirically validated dimension of inequality. Building on this pattern, the study proposes R&D access as a distinct analytical dimension of regional inequality and outlines three conceptual forms – material, cognitive, and institutional access-for future empirical investigation.