Applied Intelligence最新文献

Addressing out-of-distribution states for offline reinforcement learning with interference robustness 具有干扰鲁棒性的离线强化学习的分布外状态寻址

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-07 DOI: 10.1007/s10489-026-07558-7

Wenzhang Liu, Xiaoyang Quan, Lu Ren, Linchen Yang, Changyin Sun

{"title":"Addressing out-of-distribution states for offline reinforcement learning with interference robustness","authors":"Wenzhang Liu, Xiaoyang Quan, Lu Ren, Linchen Yang, Changyin Sun","doi":"10.1007/s10489-026-07558-7","DOIUrl":"10.1007/s10489-026-07558-7","url":null,"abstract":"<div><p>Offline reinforcement learning (RL) suffers from the out-of-distribution (OOD) problem, which significantly limits its deployment in real-world scenarios. A critical yet under-explored aspect of this challenge is OOD states: once an agent enters an unfamiliar state, it may fall into a trajectory of successive OOD states, resulting in severe performance degradation. We propose an effective and simple solution: Addressing out-of-distribution States for offline reinforcement learning with interference robustness (ASIR). At its core, ASIR introduces a transition-constrained policy, enforced through a regularization term derived from an inverse dynamics model (IDM). The IDM pre-trained on the offline dataset, captures the underlying dynamics of valid state transitions. By encouraging the policy to prefer actions that lead to familiar and predictable next states, ASIR keeps the agent within the data manifold and away from OOD regions. Theoretically, our method not only corrects for OOD states but also applies to OOD actions. ASIR also demonstrates robust performance against disturbances. In experiments, our method achieves state-of-the-art results on the D4RL and NeoRL benchmark tasks. Furthermore, to evaluate robustness under real-world disturbances, we inject Gaussian noise into the actions to push the agent toward OOD states. As the duration of the perturbation increases, ASIR exhibits significantly smaller performance degradation and lower variance, outperforming existing benchmark methods in challenging scenarios.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338637","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Deep feature extraction for long-term delinquency transition prediction under short observation windows 短观测窗下长期犯罪过渡预测的深度特征提取

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-06 DOI: 10.1007/s10489-026-07587-2

Yoona Chung, Daehan Kim, Hosoo Shin, Eunchan Kim

{"title":"Deep feature extraction for long-term delinquency transition prediction under short observation windows","authors":"Yoona Chung, Daehan Kim, Hosoo Shin, Eunchan Kim","doi":"10.1007/s10489-026-07587-2","DOIUrl":"10.1007/s10489-026-07587-2","url":null,"abstract":"<div><p>Early identification of borrowers at risk of transitioning into long-term delinquency is a critical challenge in credit risk management. In real-world financial settings, risk prediction is often constrained by short observation windows and highly aggregated tabular credit data, limiting the applicability of conventional time-series models and handcrafted feature-based approaches. To address these limitations, this study proposes a deep feature extraction framework that employs representation learning as a complementary enhancement to traditional classifiers rather than as a standalone predictive model. The proposed framework integrates self-attention-based global interaction learning with TabNet-inspired soft feature selection to construct latent representations that effectively capture delinquency-related risk signals from low-information credit data. Extensive experiments on a real-world multi-debtor credit dataset demonstrate the potential of the proposed feature extraction method to enhance long-term delinquency transition predictions. Beyond predictive accuracy, analyses of feature-space distances and model explainability reveal enhanced class separability and provide interpretable insights into the underlying drivers of delinquency transitions. Overall, the results suggest that the proposed deep representation learning offers a practical, interpretable, and effective solution for early delinquency transition detection in realistic tabular credit data environments with limited temporal information.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338564","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An oversampling framework for mitigating representation bias 一种缓解表征偏差的过采样框架

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-06 DOI: 10.1007/s10489-026-07551-0

Annalisa Deiana, Abdulhakim A. Qahtan

{"title":"An oversampling framework for mitigating representation bias","authors":"Annalisa Deiana, Abdulhakim A. Qahtan","doi":"10.1007/s10489-026-07551-0","DOIUrl":"10.1007/s10489-026-07551-0","url":null,"abstract":"<div><p>The use of Machine Learning (ML) algorithms in decision-making processes has increased significantly in recent years, providing alternatives to biased human decisions. However, ML algorithms can also exhibit bias, leading to discrimination against individuals or groups that are identified by sensitive attributes such as gender or race. This bias often arises from the imbalanced representation of demographic groups in the training data, particularly in the distribution of favorable class labels. Mitigating representation bias during the training phase is crucial to ensure fair application of ML models in decision-making processes. Synthetic data generation has been considered an effective technique for mitigating representation bias. Existing techniques fail to accurately identify the regions in the data space where adding more synthetic examples is effective. In this paper, we present a preprocessing framework for generating synthetic examples for the minority groups to create a balanced and fair dataset for training fair ML models. The proposed framework identifies subgroups with lower ratios in getting the favorable class label and identifies the boundaries of the group examples with favorable class labels. Subsequently, SMOTE-NC is used for generating synthetic examples around the group boundary until the group reaches a target imbalance ratio. Synthetic data generation is done carefully to minimize the effect on the performance of the ML models. The performance and fairness of the proposed framework are evaluated using standard performance and fairness measures. The results indicate that the proposed framework successfully reduces representation bias and enhances fairness across demographic groups, while maintaining competitive predictive performance relative to state-of-the-art bias mitigation techniques.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338621","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Balanced synthetic data generation with variational AutoEncoders: a trade-off analysis of balance, fairness, utility and privacy 与变分自动编码器平衡合成数据生成：平衡，公平，效用和隐私的权衡分析

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-06 DOI: 10.1007/s10489-026-07532-3

Marco Rondina, Antonio Vetrò, Luca Gilli, Shalini Kurapati, Juan Carlos De Martin

{"title":"Balanced synthetic data generation with variational AutoEncoders: a trade-off analysis of balance, fairness, utility and privacy","authors":"Marco Rondina, Antonio Vetrò, Luca Gilli, Shalini Kurapati, Juan Carlos De Martin","doi":"10.1007/s10489-026-07532-3","DOIUrl":"10.1007/s10489-026-07532-3","url":null,"abstract":"<div><p>The growing adoption of ML/AI systems in high-stakes contexts has raised concerns about the quality, fairness and privacy of training data. Although synthetic data offers a promising way to address these concerns, the imbalance of sensitive categorical features still represents a source of bias. This paper presents two alternative approaches for a Variational AutoEncoder (VAE) training loss function that penalise deviations from a balanced distribution of sensitive feature classes. These are an additive balance component (<span>(mathcal{B}mathcal{L}_{term})</span>) and a Simpson-index-based reweighting component (<span>(mathcal{B}mathcal{L}_{weight})</span>). In comparison with other approaches in literature, which alter the architecture of the VAE, our method is a lightweight and tunable mechanism to steer the generation towards balance. An empirical evaluation was conducted on 1,260 synthetic datasets generated from three widely used benchmark datasets (Adult, South German Credit, and COMPAS), across two batch-epoch configurations, seven loss function configurations and thirty train/evaluation splits. The results show that <span>(mathcal{B}mathcal{L}_{term})</span> reduces sensitive feature imbalance and improves statistical parity, while preserving utility and privacy at levels comparable to the baseline. The <span>(mathcal{B}mathcal{L}_{weight})</span> formulation, although capable of reducing imbalance, exhibits a weaker impact on fairness and may degrade utility and privacy at high magnitudes. The proposed contribution offers practitioners a simple, tunable mechanism for improving the balance and fairness of synthetic tabular data generation without compromising utility or privacy.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338620","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Predicting optimal conditions to reach target particle morphologies in emulsion polymerization: a reverse engineering approach 预测乳液聚合中达到目标粒子形态的最佳条件：一种逆向工程方法

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-05 DOI: 10.1007/s10489-026-07563-w

Leo Vattoly, Nicholas Ballard, Usue Mori Carrascal, Jose A. Lozano

{"title":"Predicting optimal conditions to reach target particle morphologies in emulsion polymerization: a reverse engineering approach","authors":"Leo Vattoly, Nicholas Ballard, Usue Mori Carrascal, Jose A. Lozano","doi":"10.1007/s10489-026-07563-w","DOIUrl":"10.1007/s10489-026-07563-w","url":null,"abstract":"<div><p>Particle morphology is a critical property of emulsion polymers that dictates their application properties. Achieving the desired morphology during the emulsion polymerization process requires precise control over the reaction parameters, such as the temperature profile and monomer feed rates. Traditional trial-and-error methods for experimentally optimizing these conditions are resource-intensive in terms of time and wasted chemical components. This study presents a Reverse Engineering (RE) methodology designed to predict the optimal values for the temperature and monomer feeding profiles to achieve specified particle morphologies. Our model consists of a deep learning-based surrogate model that can efficiently predict complex particle morphologies from monomer and temperature profiles. An optimization strategy was used to guide the surrogate toward configurations that closely match the targeted morphology distributions while optimizing reactor efficiency. The approach was validated by comparing the error between the distributions generated by the mathematical model for seeded semi-batch emulsion polymerization and those produced using the predicted optimal parameter combinations. The results demonstrated the effectiveness of the RE methodology in predicting the reaction conditions that lead to a given target particle morphology.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10489-026-07563-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A novel federated unlearning framework with feature-heterogeneity-aware aggregation for privacy-preserving UAV anomaly detection 基于特征异构感知聚合的无人机异常检测联合学习框架

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-05 DOI: 10.1007/s10489-026-07570-x

Mohamad Khairi Ishak, Kingkar Prosad Ghosh, Ms. Bithi, Jihane Ben Silamane, Md. Alamgir Hossain

{"title":"A novel federated unlearning framework with feature-heterogeneity-aware aggregation for privacy-preserving UAV anomaly detection","authors":"Mohamad Khairi Ishak, Kingkar Prosad Ghosh, Ms. Bithi, Jihane Ben Silamane, Md. Alamgir Hossain","doi":"10.1007/s10489-026-07570-x","DOIUrl":"10.1007/s10489-026-07570-x","url":null,"abstract":"<div><p>Unmanned Aerial Vehicles (UAVs) increasingly collaborate across Flying Ad-Hoc Networks (FANETs) to detect distributed network anomalies. However, deploying collaborative intrusion detection under emerging privacy regulations (e.g., GDPR “Right to be Forgotten”) faces two fundamental bottlenecks: (i) structural feature heterogeneity arising from diverse onboard sensor suites, which causes standard Federated Learning (FedAvg) to fail; and (ii) the severe computational burden of retraining models from scratch when a drone leaves or requests data retraction. To overcome these challenges, this paper proposes FedUnlearn-HetFL, a novel federated unlearning framework tailored for feature-heterogeneous UAV networks. The framework integrates localized attention-gated autoencoders that project asymmetric client feature dimensions (40, 45, 50, and 62 features) into a unified 64-dimensional latent space, coupled with a shared 1D-CNN backbone. When an unlearning request is initiated, FedUnlearn-HetFL executes targeted unlearning via selective layer-frozen gradient ascent on the classification head, resets local input projections, and applies rapid stabilization fine-tuning across retained nodes. Evaluated on the ISOT Drone benchmark under Dirichlet non-IID partitioning, the unlearned model collapses target node accuracy from 99.16% to 15.98% (ROC-AUC drops from 0.9999 to 0.4842), while retained clients maintain over 99.2% accuracy. Furthermore, black-box membership inference audits confirm empirical data erasure. Crucially, FedUnlearn-HetFL achieves a 1.70<span>(times )</span> execution speedup and reduces communication rounds by 95.0% compared to retraining from scratch, proving its suitability for bandwidth- and energy-constrained UAV swarms.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338516","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Topology-Aware point cloud generation via graph-induced simplicial diffusion and multi-scale GraphSAGE decoding 基于图诱导简单扩散和多尺度GraphSAGE解码的拓扑感知点云生成

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-05 DOI: 10.1007/s10489-026-07569-4

Zhongrong Zhang, Yunyun Zhang, Nengzhi Jin, Kang Lin

{"title":"Topology-Aware point cloud generation via graph-induced simplicial diffusion and multi-scale GraphSAGE decoding","authors":"Zhongrong Zhang, Yunyun Zhang, Nengzhi Jin, Kang Lin","doi":"10.1007/s10489-026-07569-4","DOIUrl":"10.1007/s10489-026-07569-4","url":null,"abstract":"<div><p>Generative models for point clouds can reproduce local geometric details, but they often struggle to maintain global structural coherence around thin parts, boundaries, and sparsely sampled regions. We propose a topology-aware point-cloud generation framework that combines PointAttention-based global tokenization, lightweight Transformer refinement, hierarchical GraphSAGE decoding, and a Graph-Induced Simplicial Layer (SCLayer) for topology-aware coordinate diffusion. The encoder compresses unordered point features into a compact permutation-invariant global token, while the decoder progressively upsamples point clouds through multi-scale graph aggregation. The proposed SCLayer is built on a <span>(varvec{k})</span>NN-graph-induced 1D simplicial structure and combines graph-induced Laplacian diffusion with a low-frequency harmonic projection. This design promotes structure-aware coordinate refinement and helps reduce fragmented thin structures, boundary discontinuities, and uneven point distributions without requiring explicit topological supervision. On ShapeNet, the proposed method achieves a COV-CD of 53.15% on Chair and an MMD-EMD of 0.787 on Airplane, outperforming the strongest baselines in the corresponding settings. Furthermore, it achieves a macro-averaged COV-CD of 51.29% over 16 ShapeNet categories, indicating favorable aggregate performance across a broader set of object categories. Ablation studies further show that SCLayer complements GraphSAGE decoding by improving structural continuity and distributional fidelity with explicitly measured computational overhead.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338148","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Cooling optimization of data centers utilizing on and off-policy deep reinforcement learning algorithms - two Danish data centers case studies 利用策略上和离线深度强化学习算法的数据中心冷却优化——两个丹麦数据中心案例研究

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-05 DOI: 10.1007/s10489-026-07439-z

Khurshed Ali, Konstantin Filoniko, Dominik Franjo Dominković

{"title":"Cooling optimization of data centers utilizing on and off-policy deep reinforcement learning algorithms - two Danish data centers case studies","authors":"Khurshed Ali, Konstantin Filoniko, Dominik Franjo Dominković","doi":"10.1007/s10489-026-07439-z","DOIUrl":"10.1007/s10489-026-07439-z","url":null,"abstract":"<div><p>Data center cooling accounts for a significant fraction of operational energy consumption and represents a major challenge for sustainable digital infrastructure. This paper presents a deep reinforcement learning (DRL) framework for cooling optimization using high-fidelity digital twins developed in Modelica and deployed through Functional Mock-up Units (FMUs). Two representative Danish data centers are investigated: one employing Aquifer Thermal Energy Storage (ATES) and another utilizing Cold Storage (CS). Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC), On and Off-Policy DRL algorithms, controllers are trained to minimize cooling costs while maintaining thermal stability and storage balance. In addition, reward redistribution mechanisms are introduced to improve long-term operational performance. Experimental results demonstrate that the proposed framework successfully adapts to varying weather conditions, electricity prices, and cooling demands while achieving energy-efficient operation. The study highlights the potential of combining physics-based digital twins with DRL for intelligent and scalable data center cooling management.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

RASA-Drive: from average robustness to long-tail safety diagnosis in scoring-based end-to-end autonomous driving RASA-Drive：基于评分的端到端自动驾驶从平均鲁棒性到长尾安全诊断

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-05 DOI: 10.1007/s10489-026-07556-9

Junjie Guo, Guoqing Geng, Liqin Sun

{"title":"RASA-Drive: from average robustness to long-tail safety diagnosis in scoring-based end-to-end autonomous driving","authors":"Junjie Guo, Guoqing Geng, Liqin Sun","doi":"10.1007/s10489-026-07556-9","DOIUrl":"10.1007/s10489-026-07556-9","url":null,"abstract":"<div><p>Average corruption scores do not identify which planning tokens are most vulnerable, and scorer confidence may be insufficient for temporal misalignment. We study this gap using the official frozen SparseDriveV2 planner on NAVSIM v2 full-navtest. We define a paired token-level failure atlas for camera-stream perturbations and evaluate whether candidate-score geometry can rank high-clean-to-collapse events without changing the selected trajectory. The protocol covers camera dropout, frame delay, occlusion, and blur/exposure/noise at three severities, with explicit score thresholds, deterministic manifests, and group-isolated evaluation. The resulting atlas shows concentrated safety-relevant performance collapse in a small number of perturbation cells. A logistic monitor using frozen scorer geometry improves risk ranking over a known-corruption metadata prior on token-, scene-, and log-group analyses, but does not improve held-out Frame delay: geometry-only ROC-AUC is 0.493 (95% CI 0.466–0.522), geometry plus metadata reaches 0.572 (0.548–0.597), and the metadata prior reaches 0.615 (0.601–0.630). A separate same-family SparseDrive NAVSIM v1 audit shows reproducible geometry signal for occlusion but weak camera-dropout transfer and a persistent Frame delay boundary. This negative result identifies a boundary of scorer-only monitoring: static candidate geometry does not directly measure temporal consistency. RASA-Drive therefore contributes a reproducible long-tail diagnosis and finite-budget risk-ranking protocol for frozen scoring-based planners, with modality-specific rather than universal transfer. It does not claim calibrated failure probabilities, closed-loop fallback benefit, or cross-architecture generalization.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338141","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A first step toward structured educational recommendations: knowledge graph embeddings in partially ordered graphs 结构化教育推荐的第一步：在部分有序图中嵌入知识图

IF 3.5 2区计算机科学

Applied Intelligence Pub Date : 2026-11-05 DOI: 10.1007/s10489-026-07576-5

Alex Martínez-Martínez, Raul Montoliu, Inmaculada Remolar

{"title":"A first step toward structured educational recommendations: knowledge graph embeddings in partially ordered graphs","authors":"Alex Martínez-Martínez, Raul Montoliu, Inmaculada Remolar","doi":"10.1007/s10489-026-07576-5","DOIUrl":"10.1007/s10489-026-07576-5","url":null,"abstract":"<div><p>The increasing complexity and scale of online educational resources call for intelli-gent recommendation systems capable of guiding learners through pedagogically coherent and adaptive learning paths. Traditional approaches, often centered on user-item interactions, tend to overlook the rich structural information encoded in prerequisite relationships and sequential dependencies inherent to educational content. This study takes a foundational step by exploring the applicability of classical Knowledge Graph Embedding methods to Partially Ordered Graphs, a formalism that naturally captures prerequisite constraints in educational con-tent. We evaluate four well-known Knowledge Graph Embedding models, TransE, TransH, TransR, and RotatE, on three educational graphs with varying struc-tural patterns. Our findings show that TransH and TransE are better suited for linear course structures, whereas RotatE performs more effectively in modular and interconnected graphs. These results suggest that the structural character-istics of a course influence the representational capacity of embedding models. This study aims to lay the groundwork and establish a baseline and conceptual foundation for future research into the representation of pedagogical structure using Partially Ordered Graphs, supporting the development of knowledge-based educational applications and expert systems for personalized learning.</p></div>","PeriodicalId":8041,"journal":{"name":"Applied Intelligence","volume":"56 16","pages":""},"PeriodicalIF":3.5,"publicationDate":"2026-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10489-026-07576-5.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149338560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0