Buyer persona expert system: An ontology-based knowledge framework for persona formulation

Deep reinforcement learning-based energy management strategy integrating physics information and expert system: efficient regenerative braking energy recovery in urban rail transit traction power system

Cross-scale attention network for automated carbon nanomaterial recognition in TEM images.

Mass transport by diffusion helps shape extracellular gradients of soluble signaling molecules in tumor microenvironments and other physiological settings. Microfluidic technologies are conducive to generating predictable chemical gradients. Yet, they often require specialized fluid-handling expertise and external pumping systems. We designed and implemented a simple microfluidic-based lateral diffusion assay (LDA) that enables reproducible and predictable biomolecular gradients without pumps and is devoid of any confounding pressure-driven flow. Using breast cancer cells that coexpress a kinase translocation reporter (KTR) for Akt, we demonstrate quantitative, real-time analysis of intracellular kinase signaling in response to diffusion-limited extracellular gradients of epidermal growth factor (EGF) in the LDA. We observed temporally and spatially staggered Akt activation and deactivation in KTR cells, with these signaling dynamics correlating to the rate of EGF delivery across zonal boundaries or EGF flux. We identified a threshold EGF concentration required for Akt activation in the median cell population and showed that this threshold concentration increases with cell density. Using mathematical modeling that incorporated empirically derived parameters, we accurately predicted individual cell Akt activation patterns among different EGF source concentrations and cell densities. Finally, we showed that both activation and deactivation patterns depend on the rate of the EGF concentration change, revealing the pivotal role of EGF flux in controlling signaling dynamics. Together, these findings establish the LDA as a powerful and accessible platform for dissecting the dynamics of extracellular gradients in controlling intracellular signaling.

Developing machine learning models for fluid milk spoilage classification.

This article presents an inverse reinforcement learning (RL) $H_{\infty } $ optimal control approach for a Takagi-Sugeno (T-S) fuzzy system (learner system) with disturbances. To reconstruct the expert system's cost function and imitate the expert system's behavior, a learner-expert framework and two inverse RL algorithms are proposed for the cases of the learner system's dynamics being known and unknown. The two developed learning algorithms are composed of three stages: an optimal control policy update stage, a gradient descent correction stage, and an inverse optimal control iteration stage. The first stage is to update the learner's optimal policies via game algebraic Riccati equations (GAREs), the second stage is to obtain the correction factor by observing the expert system's demonstrated trajectory, and the third stage is to adjust the state-penalty matrix. It is proven that the presented two algorithms are convergent, and the presented fuzzy inverse RL optimal control methodology can ensure the controlled system is asymptotically stable and achieve a Nash equilibrium solution. Finally, the presented fuzzy $H_{\infty } $ optimal control method is applied to an autonomous surface vehicle (ASV) system; the computer simulation and results illustrated the effectiveness of the developed methodology.

Direct compression is a generally preferred for tablet manufacturing method. However, it demands excipients with balanced flowability, compressibility, and packing properties. Microcrystalline cellulose (MCC) and calcium sulphate (CaSO₄) have balancing functionalities but individual limitations.To develop and optimize a novel MCC-CaSO₄ co-processed excipient (MCCASUL) for direct compression application using the SeDeM Expert System.To determine critical material attributes (CMA),12 SeDeM parameters were used to categorised MCC and CaSO₄. Functional indices, Parametric Profile Index (IPP), and Good Compression Index (IGC) were also calculated. To overcome the compressibility deficiency of CaSO₄ the required proportion of MCC was determined using SeDeM dilution potential equation. Co-dispersion method was used to prepare 9 MCCASUL batch with varying MCC:CaSO₄ ratios and further evaluated using SeDeM indices. Further characterization by FTIR, DSC, SEM, Heckel, and Kawakita analyses were performed for the optimised batch.MCC was more compressible (ѱc = 7.00) than CaSO₄ (ѱc = 4.67). The theoretical amount of correction for MCC was 14.16% among the batch preparation made, MCCASUL 4 exhibited the best results (IPP = 6.50, ѱf = 7.46, ѱc = 6.31). FTIR analysis proved the compatibility between the materials, DSC analysis revealed that modified thermal characteristics were stable, and SEM revealed improved morphology.With SeDeM expert System, co-processing yielded MCCASUL a highly versatile excipient that exhibited superior flowability, compressibility and stability and showed immense potential for direct compression tablet production.

The evolution of artificial intelligence has enabled the creation of hyper-realistic synthetic images and videos, significantly complicating the authentication of media evidence. Many algorithmic approaches to media authentication have proven ineffective in digital forensics, particularly due to the difficulty of expressing results from black-box algorithms. Compounding this challenge, practitioners face long examination times and an ever-increasing volume of media evidence in modern investigations. To address these challenges, the authors developed an artificial intelligence expert system incorporating several logic-based authentication tests related to media file structure, attribute similarity analysis, and internal file attributes. These tests were organized into distinct authentication pathways to produce concise, natural language conclusions comparable to opinions rendered by human examiners. The automated nature of an expert system may allow for demonstrably accurate results at scale without extensive human resources. This preliminary study compared the accuracy of the artificial intelligence expert system's opinions with those of trained human examiners from diverse digital forensic backgrounds. Both the expert system and the examiners were provided the same dataset, consisting of original, transmitted, edited, and synthetic videos. Both groups received file structure data, attribute similarity analyses, proprietary structural data, and metadata values. Using a 14-question survey, responses were collected and evaluated for accuracy, and error rates were calculated. Across 20 respondents, the human mean score was 69% as compared to the expert system score of 91%. These results can be used to evaluate the appropriateness of deploying artificial intelligence expert systems for forensic examinations, alongside human interpretations.

OralAbsPredict: A data-driven framework to predict human intestinal absorption (HIA) and human oral bioavailability (HOB) from chemical structures.

The management of safety in wastewater treatment plants (WWTPs) is faced with fundamental challenges, including sparse domain knowledge, dynamic evolution of safety protocols, and the necessity for highly reliable decision-making. While traditional risk assessment methods and expert systems provide essential support, they struggle to integrate multi-source heterogeneous knowledge to mitigate high-consequence, low-frequency(HCLF) risks. Existing general-purpose large language models (LLMs) demonstrate significant deficiencies in domain-specific knowledge, meanwhile, traditional fine-tuning methods are susceptible to catastrophic forgetting and knowledge conflicts during continual learning, rendering them unsuitable for direct application in this context. To address these challenges, this study proposes a progressive fine-tuning strategy to develop a domain-specific LLM tailored specifically for WWTP safety management. First, a domain-specific dataset is constructed through specialized dataset engineering. Subsequently, the proposed progressive fine-tuning strategy partitions domain knowledge into sequential stages, enabling the model to gradually learn and consolidate core knowledge at each stage before proceeding to the next. This orderly accumulation process ensures the deep integration of knowledge. The model is deployed and continuously optimized using vLLM, and direct preference optimization (DPO). The experimental results demonstrate that the progressive fine-tuning strategy effectively mitigates knowledge conflicts arising from multi-task fine-tuning. This approach not only ensures precise adherence to bottom-line safety protocols and enhances the model's depth of domain understanding in WWTP safety management, but also facilitates more coordinated capability allocation and knowledge integration across different professional tasks. By progressively refining the model, the proposed approach achieves superior task-specific performance even compared to models with substantially larger parameter scales, offering an effective and reproducible pathway for addressing analogous domain adaptation challenges.

Using the Scopus database, this study conducted a bibliometric analysis of the literature on expert systems (ES) in engineering education from 1983 to 2024. Data was retrieved from the Scopus database with a controlled search strategy on the TITLE-ABS-KEY field, resulting in 425 initial documents, filtered to 332 final documents after the PRISMA selection process. VOSviewer and Bibliometrix analysis was conducted to map collaboration networks, keyword trends, and research impact. Results showed an annual publication growth of 5.2 %, with conference papers (63.55 %) dominating journal articles (27.11 %). Keyword analysis indicated a shift from rule-based systems to artificial intelligence (AI)-based approaches, such as machine learning and deep learning. The United States (212 documents) and China (101 documents) were the main contributors, while Swansea University and the University of Fortaleza were the most productive institutions. Key challenges include the black-box nature of AI systems, the need for large-scale data, and resistance to technological change. This research highlights the potential of ES in personalizing learning while emphasizing the need for further research on AI transparency, global collaboration, and ethical frameworks for sustainable implementation.

In silico prediction of Ames mutagenicity for organosilicon compounds: Exploring and enhancing chemical space boundaries.

Optical biopsy enables minimally invasive, quantitative tissue assessment, yet clinically useful implementations require rapid and objective decision-making from compact sensors. We present a refractive-index (RI) driven classification framework based on reflection spectra acquired with an extrinsic fiber-optic Fabry-Pérot interferometric cavity (280 μm) over the biologically relevant RI range 1.33-1.42. Three proxy classes ("healthy", "HCC-like", "metastatic") were defined using literature-guided RI windows for liver tissue, and measurements were performed on certified reference liquids. As a physics-only reference, RI was estimated analytically from fringe periodicity in the wavenumber domain, achieving 0.70 accuracy and 0.48 macro-F1. To enhance discrimination, we engineered 62 spectral descriptors capturing fringe spacing (RI-related), fringe visibility, and spectral-shape cues, and trained tree-ensemble and SVM models together with an interpretable GA-optimized fuzzy expert system. On a held-out test set, tree ensembles reached macro-F1 = 1.00, while SVM and the fuzzy system achieved 0.96 and 0.97, respectively. Feature attribution identified RI as the dominant discriminative signal, with visibility-related metrics improving robustness near the HCC-like boundary. These results demonstrate that ML-augmented fiber-optic interferometry can deliver accurate and explainable diagnostic signatures, supporting the translational potential of RI-based optical biopsy.

The aim of this work is to continue the development and updating of the information system for out-of-furnace cast iron treatment "POCH" with indicators of complex cast iron treatment. Modern steel production is characterized by an increase in the production of steels with a low content of non-metallic inclusions. Out-of-furnace pig iron refining is widely used to provide BOF processors with pig iron of a given quality, which significantly reduces the cost of producing high-quality steel products. Updating the database with information on technologies for not only pig iron desulphurization, but also complex impurity removal will allow us to obtain descriptive models for these technologies and become the basis for algorithmic support of the expert system's analytical block for the selected pig iron refining technology. The Out-of-Furnace Iron Processing Database "POCH" is a full-text form of documentary database organization and makes it possible to provide a thematic search for a particular problem or a certain issue, which is regulated by a corresponding query provided with a flexible interface. The information support subsystem was developed, in particular, a software module for the automated procedure of replenishing machine passports formatted according to a special template from the User's technological information archive in Excel was developed. The analytical unit was supplemented with a subsystem for analyzing the processes of interatomic interaction in the metal-slag system using the concept of directed chemical bonding, generating complex indicators, and assessing the role of individual components of the mathematical model. The dependences of the efficiency of impurity removal (S, Si, P) as a function of the properties of the injected slag, slag basicity, component content in the slag, and treatment intensity were analyzed. The architecture of the model system for expert evaluation of the technology of complex cast iron refining was developed.

Rapid and reliable detection of carbapenemase-producing Enterobacterales (CPE) is essential for infection control, epidemiological surveillance, and optimized antimicrobial therapy. As the prevalence and diversity of carbapenemases increase, diagnostic tools capable not only of detecting CPE but also predicting carbapenemase classes are increasingly needed to guide treatment decisions. In Europe, the emergence of OXA-48-like variants such as OXA-244 and OXA-484, which exhibit low hydrolytic activity to carbapenems and temocillin, complicates routine phenotypic detection and may lead to diagnostic failure. This study evaluates an automated solution integrating the VITEK2 Advanced Expert System and BIOART expert rules for CPE detection and classification in a representative French epidemiological context. We demonstrate that this integrated approach provides accurate and rapid screening, with good performance for carbapenemase class prediction, particularly for NDM and KPC producers. The implementation of a dedicated rule significantly improves detection of difficult-to-identify OXA-244/OXA-484 producers, addressing an important diagnostic gap.

ABSTRACT Employing active learning is a viable approach to reducing the human effort required to complete geographic entity recognition tasks using deep learning methods. Nevertheless, current sampling strategies often only capture one sample characteristic (e.g. information content) and disregard the temporal dynamics of the sample feature across different iterations. Moreover, active learning methods alone can only alleviate workload at the sample selection level. To address these issues, we propose a novel active learning framework that integrates temporal features into sampling and leverages a large language model (LLM) to enhance annotation efficiency and the precision of model prediction. Specifically, we propose a mixed sampling strategy based on temporal and semantic similarity, comprising three innovative indicators: temporal instability, dynamic variance entropy, and semantic similarity. We propose an approach to utilizing the LLM Qwen2.5-7B-Instruct for machine annotation and an expert system for annotation correction, further reducing the workload of geographic entity recognition tasks. Experiments demonstrate that our proposed sampling strategies outperform random sampling, with “temporal instability” exhibiting optimal performance and robustness. Furthermore, combining machine annotation with expert systems can save approximately 27% of annotation time. Our work provides a practical solution for designing sampling strategies with temporal features and fully leveraging LLM for automatic annotation to reduce the annotation cost of geographic entity recognition tasks.

PurposeThe aim of this study was to evaluate a new algorithm‐based approach in subjective refraction by comparing the refractions obtained by a professional expert with those obtained using the SiviewExam Expert system on a large population.MethodsTwo subjective refraction methods were compared prospectively on patients visiting Rothschild Foundation between October 2022 and May 2023: (i) A manual refraction performed by an expert using a standardized method and (ii) an automated one with the SiviewExam Expert system. Outcomes of the two subjective refractions were compared using a power vector analysis method. Power vectors are a geometric representation of spherocylindrical refractive errors in 3 fundamental dioptric components: M (M = S + C/2), J0 (J0 = (−C/2) cos (2α) at an axis of α = 0 = 180°), and J45 (J45 = (−C/2) sin (2α) at an axis of α = 45°). At the end of the examination, the SiviewExam Expert system provided a report that has been analyzed as well.ResultsA total of 107 patients, with a mean age of 35.7 ± 11.9 years (ranging from 19 to 69 years) were included in the study. The mean SE was −2.10 ± 3.37 D (ranging from −11 D to +7.13 D). The mean differences have shown that there were no bias between the measurements for M (0.01 D, −0.04–0.07 D) and J45 (0.009 D, −0.01–0.03 D) and a clinically negligible bias (0.05 D, −0.08–−0.02 D) in the estimation of J0. The limits of agreement were −0.54 D (lower limit: −0.63–−0.44) to 0.56 D (upper limit: 0.47–0.65), −0.22 D (lower limit: −0.26 to −0.18) to 0.24 D (upper limit: 0.20–0.27), –0.34 D (lower limit: –0.39 to –0.29) to 0.25 D (upper limit: 0.20 to 0.30) for M, J45, and J0, respectively, and were significantly lower than the expected inter‐examiner variability of 0.7 D. Finally, the report provided by the SiviewExam Expert system was found to be accurate according to the expert in 100% of the cases.ConclusionOur outcomes show that in a large adult population of over 100 patients with a wide range of ametropias, there were no statistically significant differences in the results for M, J0, and J45 between the two refraction methods, indicating that we could reasonably clinically consider the SiviewExam Expert system examination comparable with the one conducted by an expert.

Diagnosis omission in discharge diagnosis lists is common in electronic medical records (EMRs), leading to inaccurate documentation, incorrect Diagnosis Related Group (DRG) assignments, and reduced reimbursements from overlooked Complications and Comorbidities (CC) or Major Complications and Comorbidities (MCC). To address this, we propose a data and knowledge cross-level fusion-driven learning framework for automated identification of missed diagnoses. Evaluated on real-world EMRs from six hospitals across various provinces in China, our model outperforms expert system method, BERT-based method, and multiple LLM-based baseline methods, demonstrating superior F1 scores. Results show 37.8% of EMRs predicted to have missed diagnoses, with 9.0% experiencing altered DRG groupings, subsequently affecting 3.2% of insurance reimbursement. To minimize alert fatigue, we adopted a hybrid approach combining our model with expert system, boosting precision by 6.7-13.4%. We also designed two human-machine coupling modes to demonstrate the utility of our methods in the real world.

Smartphone addiction among students is a problem that interferes with their concentration while studying, social interactions, and their academic motivation. This study analyzed the level of smartphone addiction among students of SMKS Immanuel Medan using the Naive Bayes classification algorithm and the C4.5 decision tree. This study adopts a comparative quantitative approach using the phases of Knowledge Discovery in Databases (KDD), including data collection, data cleaning, data selection, data transformation, data mining, and evaluation. The research data was collected by distributing questionnaires to 100 students at SMKS Immanuel Medan. The study variables included age, gender, duration of smartphone use, purpose of smartphone use, dominant type of social media, and the level of smartphone addiction as target variables. The classification was carried out using RapidMiner software with a 70:30 training and testing data split. Model evaluation was carried out using a confusion matrix with the parameters of accuracy, precision, recall, and F1 score. The results show that the C4.5 decision tree algorithm gives better results than the Naive Bayes algorithm. The C4.5 algorithm achieved 90% accuracy, 88.9% precision, 80% recall, and 84% F1 score, while the Naive Bayes algorithm achieved 80% accuracy, 80% precision, 66.7% recall, and 73% F1 score. This research contributed to the development of a simple web-based expert system that helps schools assess the level of smartphone addiction among students quickly, objectively, and systematically, so that it can be used as a decision-making tool to monitor smartphone use in the education sector.

Providing a novel expert system embedded with STIRPAT and MSMGO models to explore the driving factors of carbon emissions