Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Zhanar Lamasheva has a total of 138 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    FROM AN ANNOTATED CORPUS TO AN ONTOLOGY: A FORMAL MODEL OF MULTI-LEVEL SEMANTIC ANNOTATION OF FAKE NEWS IN THE KAZAKH–RUSSIAN MEDIA SPACE (KazFakeCorpus / KazFakeOnto)

    by Anargul Ajmuratovna Nekessova, Elena Anatolevna Sidorova, Zhanar Bejbutovna Lamasheva, Madina Aralbaevna Sambetbayeva, Aigerim Sembekovna Yarimbetova, Mira Zhorabekovna Kaldarova, Aksaule Abzalқyzy Nazymkhan

    The paper addresses the formal representation of knowledge about disinformation in a low-resource bilingual media environment. Automatic fake news detection is traditionally formulated as a binary REAL/FAKE classification task, which does not capture the internal structure of an unreliable message: the type of fake content, the disinformation technique employed, the author’s communicative intent, and the characteristics of the source and evidence base. Two interrelated resources are presented. The first is KazFakeCorpus, a bilingual corpus of 4,276 texts in Kazakh and Russian, balanced by class within each language; the REAL class was drawn from official materials of the Gov.kz portal and the FAKE class from their controlled transformations. The texts were annotated in Label Studio by two independent experts following a multi-level semantic scheme; Krippendorff’s Alpha for the principal levels ranged from 0.79 to 0.88. The second resource is KazFakeOnto, an ontology that recasts the annotation scheme from a flat relational representation into a formal OWL 2 DL model: 38 classes, 26 object properties, 22 data properties, 112 individuals and 1,155 asserted triples. Eleven defined classes are not asserted manually but inferred by a reasoner from property values; a property chain links a news item to its annotator through an annotation object. Consistency was verified with HermiT, and the query layer was evaluated on fifteen SPARQL and thirteen DL queries. The ontological layer is shown to provide a reproducible specification of the annotation scheme, the automatic inference of semantic categories, and a unified formal environment for formulating corpus-linguistic questions and analysing machine-learning model errors. The limitations of the approach and directions for scaling the ontology to the full corpus and to naturally occurring disinformation are discussed.

  • Open Access

    A Multi-Model Approach to Academic Performance Prediction Using Student–Semester Educational Data

    by Tole Bi Yermek, Banu Yergesh, Amanzhol Yelemessov, Madina Yelemessova, Gulmira Bekmanova, Assel Omarbekova, Aizhan Nazyrova, Zhanar Lamasheva

    Identifying students at academic risk early calls for forecasting that respects the temporal structure of educational data and stays readable for advising staff. In this work, linear, gradient-boosted, and deep tabular models are compared under one protocol on two tasks: next-semester performance prediction and binary risk classification. The data are aggregated student–semester records from a university Learning Management System, covering 732,965 academic records on 20,868 students over 14 academic years at a Central Asian institution. All models are evaluated under a strict temporal train–validation–test split, so each is asked to forecast a year it has never seen. The feature set adds two trajectory variables (avg_mark_lag1, delta_avg) for the direction of a student’s performance and two within-semester indicators (std_mark, min_mark) for its stability. On the 2022 test set, Ridge Regression reached R² = 0.552 and Logistic Regression PR-AUC = 0.432 (ROC-AUC = 0.904). The gradient-boosted ensembles (LightGBM, CatBoost, XGBoost) did not clearly beat the regularized linear baselines under temporal validation, and the deep tabular architectures (FT-Transformer, TabNet, AutoInt, MLP-embeddings, Deep & Cross), run under the identical protocol, did not outperform either family. A Top-K intervention rule recovers 33% of all at-risk students within the top 5% of ranked cases (lift ≈ 4.15×). That is a decision rule an advising office can work with under a fixed caseload. Received: 15 May 2026 | Revised: 11 August 2026 | Accepted: 30 August 2026 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in GitHub at https://github.com/MMR000/Smart_EDU_Score_Predict. Author Contribution Statement Tole Bi Yermek: Conceptualization, Methodology, Software, Formal analysis, Writing – original draft. Banu Yergesh: Methodology, Validation, Formal analysis. Amanzhol Yelemessov: Methodology, Software, Data curation, Visualization. Madina Yelemessova: Validation, Investigation, Data curation, Visualization. Gulmira Bekmanova: Investigation, Writing – review & editing. Assel Omarbekova: Resources, Project administration, Funding acquisition. Aizhan Nazyrova: Conceptualization, Writing – original draft, Supervision. Zhanar Lamasheva: Writing – review & editing, Supervision.

  • Open Access

    Comparative Evaluation of Machine Learning Models for Early Prediction of Student Academic Performance Using the Open University Learning Analytics Dataset

    by Assel Omarbekova, Aizhan Nazyrova, Gulmira Bekmanova, Zhanar Lamasheva, Rakhila Turebayeva, Zhainagul Abzal

    Early identification of students at risk of academic failure is a central problem in learning analytics, yet the multi-table, behaviorally heterogeneous, and class-imbalanced nature of educational data complicates reliable prediction. This study develops and comparatively evaluates four supervised machine learning models—logistic regression, support vector machine (SVM) with a radial basis function kernel, random forest, and extreme gradient boosting (XGBoost)—for multi-class prediction of final student outcome on the Open University Learning Analytics Dataset (OULAD), comprising 32,593 anonymized student records and more than ten million virtual-learning-environment interaction logs. A reproducible, leakage-controlled feature-engineering pipeline is proposed: final examination records are excluded, and virtual-learning-environment activity is restricted to the first 120 days of each presentation, yielding 85 demographics, assessment, and behavioral features. All models are trained on an identical stratified 80/20 split and assessed within a unified evaluation framework using accuracy, weighted and macro F1-score, balanced accuracy, per-class recall, and stratified cross-validation. XGBoost achieved the strongest overall performance (accuracy 0.7302, macro F1 0.6658, balanced accuracy 0.6555, and cross-validated accuracy 0.7314), followed closely by random forest and logistic regression, while SVM, despite the weakest aggregate scores, produced the highest recall for the practically critical Fail class (0.4026). The analysis demonstrates that aggregate accuracy substantially overstates the usefulness of these models for risk detection and that the fail class remains the principal bottleneck. A deployed Streamlit prototype illustrates operationalization. The contribution lies in the leakage-aware early-window pipeline and the imbalance-sensitive, algorithm-aware comparison that together expose where predictive value is gained and lost.


  • Open Access

    Integration of an ontological model of CDIO competencies into the design of educational programs: transforming educational projects into knowledge content

    by Yenglik Kadyr, Bulat Kubekov, Aislu Kassekeyeva, Rakhila Turebayeva, Aizhan Nazyrova, Zhanar Lamasheva, Ainur Shekerbek

    The Conceive–Design–Implement–Operate (CDIO) Syllabus expresses graduate competencies in natural language developed through project-based learning, yet the link between declared competencies and delivered content is maintained manually. This study proposes the CDIO Competency–Knowledge Ontology (CDIO–CKO): a decidable OWL 2 Description Logic (DL) formalization of the Syllabus connected to programs, projects, courses, learning outcomes, and knowledge-content units, paired with a parameterized transformation that converts educational projects into competency-linked units via decomposition, extraction, semantic annotation, and aggregation. Applied to one 240-ECTS mechatronics and robotics program (42 courses, 18 projects, 156 competencies), the methodology produced 1,287 individuals and 9,540 triples classifying in 2.8 s. On a 240-unit held-out test set against a three-expert gold standard, automated mapping reached precision 0.91, recall 0.86, and F1 0.88 (95% CI 0.84–0.91; system–standard κ = 0.79, expert pre-adjudication κ = 0.74). Traceable coverage rose from 0.59 to 0.84, and 23 prerequisite inconsistencies missed by routine manual review were surfaced (23 of 26 issues in a dedicated expert audit: precision 1.00, recall 0.88). Against five comparator families, the method achieved the highest F1 among the five evaluated baselines while uniquely supporting consistency checking, proficiency propagation, and an auditable evidence trail. Results are a single-program proof-of-concept; broader validation remains future work.

  • Open Access

    ONTOLOGY-GROUNDED NEURO-SYMBOLIC INTELLIGENT TUTORING AND KNOWLEDGE-ASSESSMENT SYSTEM SUPPORTING SMART DIGITAL EDUCATION

    by Gulbarshyn Nurgazinova, Assel Omarbekova, Aizhan Nazyrova, Zhanar Lamasheva

    Intelligent tutoring systems (ITS) for elementary algebra face a persistent tension between the interpretability of symbolic, knowledge-engineered models and the predictive power of purely neural learner models, while recent large-language-model (LLM) assistants add fluency at the cost of factual and pedagogical reliability. This paper describes an algebra tutoring system organised around a formal ontology. The ontology is an OWL knowledge base: it lists the algebraic concepts, records which concepts are prerequisites for which, and states axioms for the misconceptions students commonly hold. Both the reasoning and the assessment components read from it. Reasoning is handled by a rule-based computer-algebra solver paired with a neural knowledge-tracing model. The prerequisite relations enter the training objective as a regularisation term: the tracer is penalised when it claims a student has mastered a concept while that student still fails the prerequisites for it. Assessment uses an item-response model with ontology-weighted item parameters. Because the weights are tied to concepts, the model returns a separate mastery estimate for each concept, which is what makes the output legible to a teacher. A retrieval-augmented, ontology-constrained LLM pedagogical assistant generates hints and explanations whose vocabulary and inferences are restricted to entailments of the knowledge base, sharply reducing hallucination. The study ran for eight weeks with 428 learners. On held-out interactions the knowledge-tracing model reached an AUC of 0.851 and an RMSE of 0.369, ahead of both the Bayesian and the deep knowledge-tracing baselines. Learning gains separated the conditions more sharply than the tracing metrics did: normalized gain was (g) = 0.58 for the full system, 0.41 for a generic LLM tutor, and 0.24 for the textbook condition. Expert raters also preferred the ontology-grounded hints when asked to judge faithfulness. What this shows is narrower than it may look. A single cohort over eight weeks cannot establish that ontological grounding is what produced the gains, since the grounded system also differs from the LLM tutor in its solver and its hint templates. It does show that the combination is workable at the scale of a real course, and that grounding the hints in an explicit knowledge base does not cost accuracy in exchange for transparency.

  • Open Access

    Monitoring and comparative analysis of containerized environments

    by Assel Omarbekova, Lazzat Kussepova, Altanbek Zulkhazhav, Mamyr Altaibek, Gulmira Bekmanova, Aizhan Nazyrova, Zhanar Lamasheva

    This paper systematically examines monitoring tools for containerized applications, microservices, and DevOps environments, and provides an experimental evaluation across two deployment scenarios. The study was conducted across two infrastructures: on-premises (Docker Swarm, Kubernetes) and cloud-based (Google Kubernetes Engine). A test application was deployed, and load and stress tests were performed using Apache JMeter and k6 with virtual user counts ranging from 100 to 1,000. Metrics including CPU, memory, I/O, and network utilization were collected using Prometheus, Grafana, cAdvisor, and Docker stats. Based on exploration and evaluation, 69 unique monitoring-related tools were identified and grouped into four non-mutually exclusive categories: container-based, cloud-based, microservices-oriented, and DevOps-focused (18, 25, 13, and 38 assignments, respectively; 94 assignments in total because some tools belong to multiple categories). A classification is presented according to visualization capabilities, supported metrics, and quality attributes, including performance, security, interoperability, and usability. Experimental results demonstrate that both monitoring stacks captured variations in processor utilization under load at the resolution configured by their default sampling intervals (15–30 s for Prometheus + cAdvisor and 60 s for Google Cloud Monitoring), with the 1,000-virtual-user stress bursts visible as sharp CPU peaks in both stacks. The contribution of this work is a practical guideline for selecting monitoring tools, developed on the basis of a reference experimental environment.

  • Open Access

    Learning Analytics of Students’ Interaction with ChatGPT in Programming Education: A Process-Oriented Analysis

    by Assel Omarbekova, Marek Milosz, Gulmira Bekmanova, Yershin Rakymkan, Aizhan Nazyrova, Zhanar Lamasheva

    Generative AI tools are now widely used in undergraduate programming, yet most evidence about how students use them comes from self-report rather than from observed behaviour. This study examined the sequential structure of students’ ChatGPT (GPT- 4o, OpenAI)-supported programming work and the cognitive complexity of their queries. Screen recordings of 363 second-year students completing an individual Python 3.12 data-visualisation assignment were coded into 3985 activity episodes and analysed using descriptive statistics, lag-1 sequential analysis, and cognitive network analysis; because recordings capture actions rather than cognition, the coded categories are treated as behavioural indicators interpreted within, rather than as measurements of, the Self-Regulated Learning framework. Programming activities accounted for 63% of coded actions and ChatGPT interactions for 20%. Behaviour was organised around a troubleshooting cycle, the strongest association being between submitting error messages and reviewing ChatGPT responses (PCM → RF, Yule’s Q = 0.84, a descriptive association measure, rather than a transition probability, whose stability across students was not tested). ChatGPT use was concentrated in activities indicative of monitoring and control and was largely absent from planning and reflection. High-achieving students produced a higher proportion of deep-level submissions (32.2% vs. 19.2%) and a lower proportion of surface-level submissions (23.2% vs. 38.4%) than low-achieving students; because submissions are nested within students, this difference is reported as a property of the observed distribution rather than as an inferential finding. Comparisons computed at the level of the student were tested inferentially and reached significance with small effect sizes; comparisons computed at the level of coded actions or query submissions are reported throughout as observed properties of the aggregate distributions rather than as inferentially established differences. These episode-level findings support instructional scaffolding that structures query formulation and reflection.


  • Open Access

    Machine Learning-Based Portfolio Optimization Using Cross-Sectional Signals: Evidence from the NSE Market

    by Yerassyl Iskakov, Ainur Shekerbek, Aizhan Nazyrova, Zhanar Lamasheva, Rayimbek Durbenov

    This paper presents a two-stage Machine Learning (ML) framework for signal-based portfolio optimization in the National Stock Exchange (NSE) of India. Using a dataset of 20 major equities from 2020 to 2025, the study evaluates eleven predictive models, including regularized linear regressions, tree-based methods, and hybrid Deep Learning (DL) architectures such as Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) and Long Short-Term Memory–Multilayer Perceptron (LSTM-MLP). In the first stage, a 60-day sliding window is used to generate one-day-ahead return forecasts, which are then stabilized through a 10-period smoothing mechanism to reduce short-term prediction noise. In the second stage, these forecasts are transformed into cross-sectional investment signals and incorporated into a portfolio optimization framework. The results indicate that hybrid DL models achieve better predictive performance than traditional approaches in capturing non-linear market dynamics. Overall, the proposed framework provides a practical approach for converting model forecasts into portfolio decisions and may contribute to improved risk-adjusted performance.

  • Open Access

    Intelligent Inclusive Navigation System for a University Digital Ecosystem

    by Aibol Tileukhan, Gulmira Bekmanova, Valentina Franzoni, Alibek Barlybayev, Lena Zhetkenbay, Altynbek Sharipbay, Zhanar Lamasheva, Assel Omarbekova, Aizhan Nazyrova

    Indoor navigation remains challenging for students with visual impairments because GPS is unavailable indoors and building layouts are often complex. This paper presents a wearable marker-assisted navigation system integrating QR code localization, SSD MobileNet V3 obstacle detection, TFmini-S LiDAR ranging, A*-based dynamic route planning, and audio feedback on a Raspberry Pi 5. The main contribution is an analytical framework relating marker spacing to predicted localization uncertainty and defining a latency budget for obstacle warnings. A confidence-weighted sensor-fusion method is developed analytically but was not implemented in the evaluated prototype, in which the QR code, camera, and LiDAR channels operated independently. The proposed fusion method and the simulated multi-floor planning extension require further experimental validation. Controlled tests produced a mean positioning error below 1.2 m, a LiDAR ranging MAE of 8.3 cm, and an object-detection throughput of 6–9 FPS. A pilot field evaluation covered nine routes totalling 901 m across two buildings and included one participant with self-reported vision loss of approximately 95%. All route trials were completed, although some required researcher assistance. The system remains a proof of concept and has not yet been evaluated against a baseline or with a sufficiently large target-user sample.

  • An AI-Driven Digital Transformation Skills Framework: A Systematic Literature Review and Machine Learning Analysis

    by Assel Omarbekova, Gulmira Bekmanova, Latifa Sautbaeva, Alma Zakirova, Aizhan Nazyrova, Zhanar Lamasheva