Research on Condition Assessment of Nuclear Power Systems Based on Fault Severity and Fault Harmfulness

Energy, Год журнала: 2024, Номер 311, С. 133396 - 133396

Опубликована: Окт. 9, 2024

Язык: Английский

---

A Systematic Review of Artificial Intelligence in Orthopaedic Disease Detection: A Taxonomy for Analysis and Trustworthiness Evaluation

International Journal of Computational Intelligence Systems, Год журнала: 2024, Номер 17(1)

Опубликована: Дек. 18, 2024

Язык: Английский

---

Technology Adoption in Healthcare – a Modified Tam Model & Empirical Analysis

Опубликована: Янв. 1, 2025

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Язык: Английский

---

Trustworthy and Explainable Federated System for Extracting Descriptive Rules in a Data Streaming Environment

Results in Engineering, Год журнала: 2025, Номер unknown, С. 104137 - 104137

Опубликована: Янв. 1, 2025

Язык: Английский

---

A GRA-based heterogeneous multi-attribute group decision-making method with attribute interactions

Computers & Industrial Engineering, Год журнала: 2025, Номер unknown, С. 110920 - 110920

Опубликована: Янв. 1, 2025

Язык: Английский

---

Trustworthiness Optimisation Process: A Methodology for Assessing and Enhancing Trust in AI Systems

Electronics, Год журнала: 2025, Номер 14(7), С. 1454 - 1454

Опубликована: Апрель 3, 2025

The emerging capabilities of artificial intelligence (AI) and the systems that employ them have reached a point where they are integrated into critical decision-making processes, making it paramount to change adjust how evaluated, monitored, governed. For this reason, trustworthy AI (TAI) has received increased attention lately, primarily aiming build trust between humans AI. Due far-reaching socio-technical consequences AI, organisations government bodies already started implementing frameworks legislation for enforcing TAI, such as European Union’s Act. Multiple approaches evolved around covering different aspects trustworthiness include fairness, bias, explainability, robustness, accuracy, more. Moreover, depending on models stage system lifecycle, several methods techniques can be used each characteristic assess potential risks mitigate them. Deriving from all above is need comprehensive tools solutions help stakeholders follow TAI guidelines adopt practically increase trustworthiness. In paper, we formulate propose Trustworthiness Optimisation Process (TOP), which operationalises brings together its procedural technical throughout lifecycle. It incorporates state-of-the-art enablers documentation cards, risk management, toolkits find given system. To showcase application proposed methodology, case study conducted, demonstrating fairness an increased.

Язык: Английский

---

An integrated consensus-oriented decision-making framework for exploring the barriers and applications of medical digital twins

Expert Systems with Applications, Год журнала: 2025, Номер 284, С. 127683 - 127683

Опубликована: Апрель 28, 2025

Язык: Английский

---

Assessing ML classification algorithms and NLP techniques for depression detection: An experimental case study

PLoS ONE, Год журнала: 2025, Номер 20(5), С. e0322299 - e0322299

Опубликована: Май 28, 2025

Context and background. Depression has affected millions of people worldwide become one the most common mental disorders. Early disorder detection can reduce costs for public health agencies prevent other major comorbidities. Additionally, shortage specialized personnel is very concerning since depression diagnosis highly dependent on expert professionals time-consuming. Research problems . Recent research evidenced that machine learning (ML) natural language processing (NLP) tools techniques have significantly benefited depression. However, there are still several challenges in assessment approaches which conditions such as post-traumatic stress (PTSD) present. These include assessing alternatives terms data cleaning pre-processing techniques, feature selection, appropriate ML classification algorithms. Purpose study This paper tackles an based a case compares different classifiers, specifically pre-processing, parameter setting, model choices. Methodology The experimental Distress Analysis Interview Corpus - Wizard-of-Oz (DAIC-WOZ) dataset, designed to support disorders depression, anxiety, PTSD. Major findings Besides alternative we were able build models with accuracy levels around 84% Random Forest XGBoost models, higher than results from comparable literature presented level 72% SVM model. Conclusions More comprehensive assessments algorithms NLP advance state art improved settings performance.

Язык: Английский

---

Evaluation of energy economic optimization models using multi-criteria decision-making approach

Expert Systems with Applications, Год журнала: 2024, Номер 255, С. 124842 - 124842

Опубликована: Июль 26, 2024

Язык: Английский

---

Early Prediction of Cardiac Arrest in the Intensive Care Unit using Explainable Machine Learning: Retrospective Study (Preprint)

Опубликована: Июнь 4, 2024

Cardiac arrest (CA) is the superior cause of death in patients intensive care unit (ICU). Many CA prediction models with high sensitivity have been developed to prevent advance, but there was a difficulty practical use due lack generalization verification. Furthermore, different subtypes ICU heterogeneity, those characteristics were not identified. We propose clinically interpretable ensemble approach for timely accurate within 24-hour regardless including patient populations and ICU. In addition, subject-independent evaluation performed emphasize performance model we analyzed results that can adopted by clinicians real-time. Patients retrospectively using data from Medical Information Mart Intensive Care-IV (MIMIC-IV) eICU Collaborative Research Database (eICU-CRD). To address problem underperformance, constructed our framework vital sign-based, multi-resolution statistical, Gini index-based feature sets 12-hour window learn unique itself. extracted three types features each database compare between groups at risk MIMIC-IV without eICU-CRD. After extracting features, TabNet cost-sensitive learning. 10-fold leave one subject out cross-validation cross-dataset method used check real-time performance. evaluated eICU-CRD cohort eICU-CRD, respectively. Finally, external validation ability. The decision mask proposed capture interpretability model. achieved conventional methods both obtained higher than baseline Interpretable facilitate clinician’s understanding as statistical test non-CA groups. Next, tested MIMIC-IV-trained ability, compared baselines. Our novel provides stable predictive power environments Most global information shows differences groups, demonstrating they are useful indicators clinical decisions. Therefore, system mature algorithm allows intervene early through be applied trials digital health.

Язык: Английский

---