Parkinson’s disease(PD)is a debilitating neurological disorder affecting over 10 million people worldwide.PD classification models using voice signals as input are common in the literature.It is believed that using d...Parkinson’s disease(PD)is a debilitating neurological disorder affecting over 10 million people worldwide.PD classification models using voice signals as input are common in the literature.It is believed that using deep learning algorithms further enhances performance;nevertheless,it is challenging due to the nature of small-scale and imbalanced PD datasets.This paper proposed a convolutional neural network-based deep support vector machine(CNN-DSVM)to automate the feature extraction process using CNN and extend the conventional SVM to a DSVM for better classification performance in small-scale PD datasets.A customized kernel function reduces the impact of biased classification towards the majority class(healthy candidates in our consideration).An improved generative adversarial network(IGAN)was designed to generate additional training data to enhance the model’s performance.For performance evaluation,the proposed algorithm achieves a sensitivity of 97.6%and a specificity of 97.3%.The performance comparison is evaluated from five perspectives,including comparisons with different data generation algorithms,feature extraction techniques,kernel functions,and existing works.Results reveal the effectiveness of the IGAN algorithm,which improves the sensitivity and specificity by 4.05%–4.72%and 4.96%–5.86%,respectively;and the effectiveness of the CNN-DSVM algorithm,which improves the sensitivity by 1.24%–57.4%and specificity by 1.04%–163%and reduces biased detection towards the majority class.The ablation experiments confirm the effectiveness of individual components.Two future research directions have also been suggested.展开更多
Accurate purchase prediction in e-commerce critically depends on the quality of behavioral features.This paper proposes a layered and interpretable feature engineering framework that organizes user signals into three ...Accurate purchase prediction in e-commerce critically depends on the quality of behavioral features.This paper proposes a layered and interpretable feature engineering framework that organizes user signals into three layers:Basic,Conversion&Stability(efficiency and volatility across actions),and Advanced Interactions&Activity(crossbehavior synergies and intensity).Using real Taobao(Alibaba’s primary e-commerce platform)logs(57,976 records for 10,203 users;25 November–03 December 2017),we conducted a hierarchical,layer-wise evaluation that holds data splits and hyperparameters fixed while varying only the feature set to quantify each layer’s marginal contribution.Across logistic regression(LR),decision tree,random forest,XGBoost,and CatBoost models with stratified 5-fold cross-validation,the performance improvedmonotonically fromBasic to Conversion&Stability to Advanced features.With LR,F1 increased from 0.613(Basic)to 0.962(Advanced);boosted models achieved high discrimination(0.995 AUC Score)and an F1 score up to 0.983.Calibration and precision–recall analyses indicated strong ranking quality and acknowledged potential dataset and period biases given the short(9-day)window.By making feature contributions measurable and reproducible,the framework complements model-centric advances and offers a transparent blueprint for production-grade behavioralmodeling.The code and processed artifacts are publicly available,and future work will extend the validation to longer,seasonal datasets and hybrid approaches that combine automated feature learning with domain-driven design.展开更多
Detecting faces under occlusion remains a significant challenge in computer vision due to variations caused by masks,sunglasses,and other obstructions.Addressing this issue is crucial for applications such as surveill...Detecting faces under occlusion remains a significant challenge in computer vision due to variations caused by masks,sunglasses,and other obstructions.Addressing this issue is crucial for applications such as surveillance,biometric authentication,and human-computer interaction.This paper provides a comprehensive review of face detection techniques developed to handle occluded faces.Studies are categorized into four main approaches:feature-based,machine learning-based,deep learning-based,and hybrid methods.We analyzed state-of-the-art studies within each category,examining their methodologies,strengths,and limitations based on widely used benchmark datasets,highlighting their adaptability to partial and severe occlusions.The review also identifies key challenges,including dataset diversity,model generalization,and computational efficiency.Our findings reveal that deep learning methods dominate recent studies,benefiting from their ability to extract hierarchical features and handle complex occlusion patterns.More recently,researchers have increasingly explored Transformer-based architectures,such as Vision Transformer(ViT)and Swin Transformer,to further improve detection robustness under challenging occlusion scenarios.In addition,hybrid approaches,which aim to combine traditional andmodern techniques,are emerging as a promising direction for improving robustness.This review provides valuable insights for researchers aiming to develop more robust face detection systems and for practitioners seeking to deploy reliable solutions in real-world,occlusionprone environments.Further improvements and the proposal of broader datasets are required to developmore scalable,robust,and efficient models that can handle complex occlusions in real-world scenarios.展开更多
The aim of this article is to explore potential directions for the development of artificial intelligence(AI).It points out that,while current AI can handle the statistical properties of complex systems,it has difficu...The aim of this article is to explore potential directions for the development of artificial intelligence(AI).It points out that,while current AI can handle the statistical properties of complex systems,it has difficulty effectively processing and fully representing their spatiotemporal complexity patterns.The article also discusses a potential path of AI development in the engineering domain.Based on the existing understanding of the principles of multilevel com-plexity,this article suggests that consistency among the logical structures of datasets,AI models,model-building software,and hardware will be an important AI development direction and is worthy of careful consideration.展开更多
When dealing with imbalanced datasets,the traditional support vectormachine(SVM)tends to produce a classification hyperplane that is biased towards the majority class,which exhibits poor robustness.This paper proposes...When dealing with imbalanced datasets,the traditional support vectormachine(SVM)tends to produce a classification hyperplane that is biased towards the majority class,which exhibits poor robustness.This paper proposes a high-performance classification algorithm specifically designed for imbalanced datasets.The proposed method first uses a biased second-order cone programming support vectormachine(B-SOCP-SVM)to identify the support vectors(SVs)and non-support vectors(NSVs)in the imbalanced data.Then,it applies the synthetic minority over-sampling technique(SV-SMOTE)to oversample the support vectors of the minority class and uses the random under-sampling technique(NSV-RUS)multiple times to undersample the non-support vectors of the majority class.Combining the above-obtained minority class data set withmultiple majority class datasets can obtainmultiple new balanced data sets.Finally,SOCP-SVM is used to classify each data set,and the final result is obtained through the integrated algorithm.Experimental results demonstrate that the proposed method performs excellently on imbalanced datasets.展开更多
Inferring phylogenetic trees from molecular sequences is a cornerstone of evolutionary biology.Many standard phylogenetic methods(such as maximum-likelihood[ML])rely on explicit models of sequence evolution and thus o...Inferring phylogenetic trees from molecular sequences is a cornerstone of evolutionary biology.Many standard phylogenetic methods(such as maximum-likelihood[ML])rely on explicit models of sequence evolution and thus often suffer from model misspecification or inadequacy.The on-rising deep learning(DL)techniques offer a powerful alternative.Deep learning employs multi-layered artificial neural networks to progressively transform input data into more abstract and complex representations.DL methods can autonomously uncover meaningful patterns from data,thereby bypassing potential biases introduced by predefined features(Franklin,2005;Murphy,2012).Recent efforts have aimed to apply deep neural networks(DNNs)to phylogenetics,with a growing number of applications in tree reconstruction(Suvorov et al.,2020;Zou et al.,2020;Nesterenko et al.,2022;Smith and Hahn,2023;Wang et al.,2023),substitution model selection(Abadi et al.,2020;Burgstaller-Muehlbacher et al.,2023),and diversification rate inference(Voznica et al.,2022;Lajaaiti et al.,2023;Lambert et al.,2023).In phylogenetic tree reconstruction,PhyDL(Zou et al.,2020)and Tree_learning(Suvorov et al.,2020)are two notable DNN-based programs designed to infer unrooted quartet trees directly from alignments of four amino acid(AA)and DNA sequences,respectively.展开更多
Climate change significantly affects environment,ecosystems,communities,and economies.These impacts often result in quick and gradual changes in water resources,environmental conditions,and weather patterns.A geograph...Climate change significantly affects environment,ecosystems,communities,and economies.These impacts often result in quick and gradual changes in water resources,environmental conditions,and weather patterns.A geographical study was conducted in Arizona State,USA,to examine monthly precipi-tation concentration rates over time.This analysis used a high-resolution 0.50×0.50 grid for monthly precip-itation data from 1961 to 2022,Provided by the Climatic Research Unit.The study aimed to analyze climatic changes affected the first and last five years of each decade,as well as the entire decade,during the specified period.GIS was used to meet the objectives of this study.Arizona experienced 51–568 mm,67–560 mm,63–622 mm,and 52–590 mm of rainfall in the sixth,seventh,eighth,and ninth decades of the second millennium,respectively.Both the first and second five year periods of each decade showed accept-able rainfall amounts despite fluctuations.However,rainfall decreased in the first and second decades of the third millennium.and in the first two years of the third decade.Rainfall amounts dropped to 42–472 mm,55–469 mm,and 74–498 mm,respectively,indicating a downward trend in precipitation.The central part of the state received the highest rainfall,while the eastern and western regions(spanning north to south)had significantly less.Over the decades of the third millennium,the average annual rainfall every five years was relatively low,showing a declining trend due to severe climate changes,generally ranging between 35 mm and 498 mm.The central regions consistently received more rainfall than the eastern and western outskirts.Arizona is currently experiencing a decrease in rainfall due to climate change,a situation that could deterio-rate further.This highlights the need to optimize the use of existing rainfall and explore alternative water sources.展开更多
Standardized datasets are foundational to healthcare informatization by enhancing data quality and unleashing the value of data elements.Using bibliometrics and content analysis,this study examines China's healthc...Standardized datasets are foundational to healthcare informatization by enhancing data quality and unleashing the value of data elements.Using bibliometrics and content analysis,this study examines China's healthcare dataset standards from 2011 to 2025.It analyzes their evolution across types,applications,institutions,and themes,highlighting key achievements including substantial growth in quantity,optimized typology,expansion into innovative application scenarios such as health decision support,and broadened institutional involvement.The study also identifies critical challenges,including imbalanced development,insufficient quality control,and a lack of essential metadata—such as authoritative data element mappings and privacy annotations—which hampers the delivery of intelligent services.To address these challenges,the study proposes a multi-faceted strategy focused on optimizing the standard system's architecture,enhancing quality and implementation,and advancing both data governance—through authoritative tracing and privacy protection—and intelligent service provision.These strategies aim to promote the application of dataset standards,thereby fostering and securing the development of new productive forces in healthcare.展开更多
Face recognition has emerged as one of the most prominent applications of image analysis and under-standing,gaining considerable attention in recent years.This growing interest is driven by two key factors:its extensi...Face recognition has emerged as one of the most prominent applications of image analysis and under-standing,gaining considerable attention in recent years.This growing interest is driven by two key factors:its extensive applications in law enforcement and the commercial domain,and the rapid advancement of practical technologies.Despite the significant advancements,modern recognition algorithms still struggle in real-world conditions such as varying lighting conditions,occlusion,and diverse facial postures.In such scenarios,human perception is still well above the capabilities of present technology.Using the systematic mapping study,this paper presents an in-depth review of face detection algorithms and face recognition algorithms,presenting a detailed survey of advancements made between 2015 and 2024.We analyze key methodologies,highlighting their strengths and restrictions in the application context.Additionally,we examine various datasets used for face detection/recognition datasets focusing on the task-specific applications,size,diversity,and complexity.By analyzing these algorithms and datasets,this survey works as a valuable resource for researchers,identifying the research gap in the field of face detection and recognition and outlining potential directions for future research.展开更多
This study investigated the impacts of random negative training datasets(NTDs)on the uncertainty of machine learning models for geologic hazard susceptibility assessment of the Loess Plateau,northern Shaanxi Province,...This study investigated the impacts of random negative training datasets(NTDs)on the uncertainty of machine learning models for geologic hazard susceptibility assessment of the Loess Plateau,northern Shaanxi Province,China.Based on randomly generated 40 NTDs,the study developed models for the geologic hazard susceptibility assessment using the random forest algorithm and evaluated their performances using the area under the receiver operating characteristic curve(AUC).Specifically,the means and standard deviations of the AUC values from all models were then utilized to assess the overall spatial correlation between the conditioning factors and the susceptibility assessment,as well as the uncertainty introduced by the NTDs.A risk and return methodology was thus employed to quantify and mitigate the uncertainty,with log odds ratios used to characterize the susceptibility assessment levels.The risk and return values were calculated based on the standard deviations and means of the log odds ratios of various locations.After the mean log odds ratios were converted into probability values,the final susceptibility map was plotted,which accounts for the uncertainty induced by random NTDs.The results indicate that the AUC values of the models ranged from 0.810 to 0.963,with an average of 0.852 and a standard deviation of 0.035,indicating encouraging prediction effects and certain uncertainty.The risk and return analysis reveals that low-risk and high-return areas suggest lower standard deviations and higher means across multiple model-derived assessments.Overall,this study introduces a new framework for quantifying the uncertainty of multiple training and evaluation models,aimed at improving their robustness and reliability.Additionally,by identifying low-risk and high-return areas,resource allocation for geologic hazard prevention and control can be optimized,thus ensuring that limited resources are directed toward the most effective prevention and control measures.展开更多
Model evaluation using benchmark datasets is an important method to measure the capability of large language models(LLMs)in specific domains,and it is mainly used to assess the knowledge and reasoning abilities of LLM...Model evaluation using benchmark datasets is an important method to measure the capability of large language models(LLMs)in specific domains,and it is mainly used to assess the knowledge and reasoning abilities of LLMs.Therefore,in order to better assess the capability of LLMs in the agricultural domain,Agri-Eval was proposed as a benchmark for assessing the knowledge and reasoning ability of LLMs in agriculture.The assessment dataset used in Agri-Eval covered seven major disciplines in the agricultural domain:crop science,horticulture,plant protection,animal husbandry,forest science,aquaculture science,and grass science,and contained a total of 2283 questions.Among domestic general-purpose LLMs,DeepSeek R1 performed best with an accuracy rate of 75.49%.In the realm of international general-purpose LLMs,Gemini 2.0 pro exp 0205 standed out as the top performer,achieving an accuracy rate of 74.28%.As an LLMs in agriculture vertical,Shennong V2.0 outperformed all the LLMs in China,and the answer accuracy rate of agricultural knowledge exceeded that of all the existing general-purpose LLMs.The launch of Agri-Eval helped the LLM developers to comprehensively evaluate the model's capability in the field of agriculture through a variety of tasks and tests to promote the development of the LLMs in the field of agriculture.展开更多
Small datasets are often challenging due to their limited sample size.This research introduces a novel solution to these problems:average linkage virtual sample generation(ALVSG).ALVSG leverages the underlying data st...Small datasets are often challenging due to their limited sample size.This research introduces a novel solution to these problems:average linkage virtual sample generation(ALVSG).ALVSG leverages the underlying data structure to create virtual samples,which can be used to augment the original dataset.The ALVSG process consists of two steps.First,an average-linkage clustering technique is applied to the dataset to create a dendrogram.The dendrogram represents the hierarchical structure of the dataset,with each merging operation regarded as a linkage.Next,the linkages are combined into an average-based dataset,which serves as a new representation of the dataset.The second step in the ALVSG process involves generating virtual samples using the average-based dataset.The research project generates a set of 100 virtual samples by uniformly distributing them within the provided boundary.These virtual samples are then added to the original dataset,creating a more extensive dataset with improved generalization performance.The efficacy of the ALVSG approach is validated through resampling experiments and t-tests conducted on two small real-world datasets.The experiments are conducted on three forecasting models:the support vector machine for regression(SVR),the deep learning model(DL),and XGBoost.The results show that the ALVSG approach outperforms the baseline methods in terms of mean square error(MSE),root mean square error(RMSE),and mean absolute error(MAE).展开更多
With the increasing growth of online news,fake electronic news detection has become one of the most important paradigms of modern research.Traditional electronic news detection techniques are generally based on contex...With the increasing growth of online news,fake electronic news detection has become one of the most important paradigms of modern research.Traditional electronic news detection techniques are generally based on contextual understanding,sequential dependencies,and/or data imbalance.This makes distinction between genuine and fabricated news a challenging task.To address this problem,we propose a novel hybrid architecture,T5-SA-LSTM,which synergistically integrates the T5 Transformer for semantically rich contextual embedding with the Self-Attentionenhanced(SA)Long Short-Term Memory(LSTM).The LSTM is trained using the Adam optimizer,which provides faster and more stable convergence compared to the Stochastic Gradient Descend(SGD)and Root Mean Square Propagation(RMSProp).The WELFake and FakeNewsPrediction datasets are used,which consist of labeled news articles having fake and real news samples.Tokenization and Synthetic Minority Over-sampling Technique(SMOTE)methods are used for data preprocessing to ensure linguistic normalization and class imbalance.The incorporation of the Self-Attention(SA)mechanism enables the model to highlight critical words and phrases,thereby enhancing predictive accuracy.The proposed model is evaluated using accuracy,precision,recall(sensitivity),and F1-score as performance metrics.The model achieved 99%accuracy on the WELFake dataset and 96.5%accuracy on the FakeNewsPrediction dataset.It outperformed the competitive schemes such as T5-SA-LSTM(RMSProp),T5-SA-LSTM(SGD)and some other models.展开更多
In modern industrial production,foreign object detection in complex environments is crucial to ensure product quality and production safety.Detection systems based on deep-learning image processing algorithms often fa...In modern industrial production,foreign object detection in complex environments is crucial to ensure product quality and production safety.Detection systems based on deep-learning image processing algorithms often face challenges with handling high-resolution images and achieving accurate detection against complex backgrounds.To address these issues,this study employs the PatchCore unsupervised anomaly detection algorithm combined with data augmentation techniques to enhance the system’s generalization capability across varying lighting conditions,viewing angles,and object scales.The proposed method is evaluated in a complex industrial detection scenario involving the bogie of an electric multiple unit(EMU).A dataset consisting of complex backgrounds,diverse lighting conditions,and multiple viewing angles is constructed to validate the performance of the detection system in real industrial environments.Experimental results show that the proposed model achieves an average area under the receiver operating characteristic curve(AUROC)of 0.92 and an average F1 score of 0.85.Combined with data augmentation,the proposed model exhibits improvements in AUROC by 0.06 and F1 score by 0.03,demonstrating enhanced accuracy and robustness for foreign object detection in complex industrial settings.In addition,the effects of key factors on detection performance are systematically analyzed,providing practical guidance for parameter selection in real industrial applications.展开更多
Most predictive maintenance studies have emphasized accuracy but provide very little focus on Interpretability or deployment readiness.This study improves on prior methods by developing a small yet robust system that ...Most predictive maintenance studies have emphasized accuracy but provide very little focus on Interpretability or deployment readiness.This study improves on prior methods by developing a small yet robust system that can predict when turbofan engines will fail.It uses the NASA CMAPSS dataset,which has over 200,000 engine cycles from260 engines.The process begins with systematic preprocessing,which includes imputation,outlier removal,scaling,and labelling of the remaining useful life.Dimensionality is reduced using a hybrid selection method that combines variance filtering,recursive elimination,and gradient-boosted importance scores,yielding a stable set of 10 informative sensors.To mitigate class imbalance,minority cases are oversampled,and class-weighted losses are applied during training.Benchmarking is carried out with logistic regression,gradient boosting,and a recurrent design that integrates gated recurrent units with long short-term memory networks.The Long Short-Term Memory–Gated Recurrent Unit(LSTM–GRU)hybrid achieved the strongest performance with an F1 score of 0.92,precision of 0.93,recall of 0.91,ReceiverOperating Characteristic–AreaUnder the Curve(ROC-AUC)of 0.97,andminority recall of 0.75.Interpretability testing using permutation importance and Shapley values indicates that sensors 13,15,and 11 are the most important indicators of engine wear.The proposed system combines imbalance handling,feature reduction,and Interpretability into a practical design suitable for real industrial settings.展开更多
In the competitive retail industry of the digital era,data-driven insights into gender-specific customer behavior are essential.They support the optimization of store performance,layout design,product placement,and ta...In the competitive retail industry of the digital era,data-driven insights into gender-specific customer behavior are essential.They support the optimization of store performance,layout design,product placement,and targeted marketing.However,existing computer vision solutions often rely on facial recognition to gather such insights,raising significant privacy and ethical concerns.To address these issues,this paper presents a privacypreserving customer analytics system through two key strategies.First,we deploy a deep learning framework using YOLOv9s,trained on the RCA-TVGender dataset.Cameras are positioned perpendicular to observation areas to reduce facial visibility while maintaining accurate gender classification.Second,we apply AES-128 encryption to customer position data,ensuring secure access and regulatory compliance.Our system achieved overall performance,with 81.5%mAP@50,77.7%precision,and 75.7%recall.Moreover,a 90-min observational study confirmed the system’s ability to generate privacy-protected heatmaps revealing distinct behavioral patterns between male and female customers.For instance,women spent more time in certain areas and showed interest in different products.These results confirm the system’s effectiveness in enabling personalized layout and marketing strategies without compromising privacy.展开更多
Predictive maintenance(PdM)is vital for ensuring the reliability,safety,and cost efficiency of heavyduty vehicle fleets.However,real-world sensor data are often highly imbalanced,noisy,and temporally irregular,posing ...Predictive maintenance(PdM)is vital for ensuring the reliability,safety,and cost efficiency of heavyduty vehicle fleets.However,real-world sensor data are often highly imbalanced,noisy,and temporally irregular,posing significant challenges to model robustness and deployment.Using multivariate time-series data from Scania trucks,this study proposes a novel PdM framework that integrates efficient feature summarization with cost-sensitive hierarchical classification.First,the proposed last_k_summary method transforms recent operational records into compact statistical and trend-based descriptors while preserving missingness,allowing LightGBM to leverage its inherent split rules without ad-hoc imputation.Then,a two-stage LightGBM framework is developed for fault detection and severity classification:Stage A performs safety-prioritized fault screening(normal vs.fault)with a false-negativeweighted objective,and Stage B refines the detected faults into four severity levels through a cascaded hierarchy of binary classifiers.Under the official cost matrix of the IDA Industrial Challenge,the framework achieves total misclassification costs of 36,113(validation)and 36,314(test),outperforming XGBoost and Bi-LSTM by 3.8%-13.5%while maintaining high recall for the safety-critical class(0.83 validation,0.77 test).These results demonstrate that the proposed approach not only improves predictive accuracy but also provides a practical and deployable PdM solution that reduces maintenance cost,enhances fleet safety,and supports data-driven decision-making in industrial environments.展开更多
It remains difficult to automate the creation and validation of Unified Modeling Language(UML)dia-grams due to unstructured requirements,limited automated pipelines,and the lack of reliable evaluation methods.This stu...It remains difficult to automate the creation and validation of Unified Modeling Language(UML)dia-grams due to unstructured requirements,limited automated pipelines,and the lack of reliable evaluation methods.This study introduces a cohesive architecture that amalgamates requirement development,UML synthesis,and multimodal validation.First,LLaMA-3.2-1B-Instruct was utilized to generate user-focused requirements.Then,DeepSeek-R1-Distill-Qwen-32B applies its reasoning skills to transform these requirements into PlantUML code.Using this dual-LLM pipeline,we constructed a synthetic dataset of 11,997 UML diagrams spanning six major diagram families.Rendering analysis showed that 89.5%of the generated diagrams compile correctly,while invalid cases were detected automatically.To assess quality,we employed a multimodal scoring method that combines Qwen2.5-VL-3B,LLaMA-3.2-11B-Vision-Instruct and Aya-Vision-8B,with weights based on MMMU performance.A study with 94 experts revealed strong alignment between automatic and manual evaluations,yielding a Pearson correlation of r=0.82 and a Fleiss’Kappa of 0.78.This indicates a high degree of concordance between automated metrics and human judgment.Overall,the results demonstrated that our scoring system is effective and that the proposed generation pipeline produces UML diagrams that are both syntactically correct and semantically coherent.More broadly,the system provides a scalable and reproducible foundation for future work in AI-driven software modeling and multimodal verification.展开更多
Skin diseases affect millions worldwide.Early detection is key to preventing disfigurement,lifelong disability,or death.Dermoscopic images acquired in primary-care settings show high intra-class visual similarity and ...Skin diseases affect millions worldwide.Early detection is key to preventing disfigurement,lifelong disability,or death.Dermoscopic images acquired in primary-care settings show high intra-class visual similarity and severe class imbalance,and occasional imaging artifacts can create ambiguity for state-of-the-art convolutional neural networks(CNNs).We frame skin lesion recognition as graph-based reasoning and,to ensure fair evaluation and avoid data leakage,adopt a strict lesion-level partitioning strategy.Each image is first over-segmented using SLIC(Simple Linear Iterative Clustering)to produce perceptually homogeneous superpixels.These superpixels form the nodes of a region-adjacency graph whose edges encode spatial continuity.Node attributes are 1280-dimensional embeddings extracted with a lightweight yet expressive EfficientNet-B0 backbone,providing strong representational power at modest computational cost.The resulting graphs are processed by a five-layer Graph Attention Network(GAT)that learns to weight inter-node relationships dynamically and aggregates multi-hop context before classifying lesions into seven classes with a log-softmax output.Extensive experiments on the DermaMNIST benchmark show the proposed pipeline achieves 88.35%accuracy and 98.04%AUC,outperforming contemporary CNNs,AutoML approaches,and alternative graph neural networks.An ablation study indicates EfficientNet-B0 produces superior node descriptors compared with ResNet-18 and DenseNet,and that roughly five GAT layers strike a good balance between being too shallow and over-deep while avoiding oversmoothing.The method requires no data augmentation or external metadata,making it a drop-in upgrade for clinical computer-aided diagnosis systems.展开更多
基金The work described in this paper was fully supported by a grant from Hong Kong Metropolitan University(RIF/2021/05).
文摘Parkinson’s disease(PD)is a debilitating neurological disorder affecting over 10 million people worldwide.PD classification models using voice signals as input are common in the literature.It is believed that using deep learning algorithms further enhances performance;nevertheless,it is challenging due to the nature of small-scale and imbalanced PD datasets.This paper proposed a convolutional neural network-based deep support vector machine(CNN-DSVM)to automate the feature extraction process using CNN and extend the conventional SVM to a DSVM for better classification performance in small-scale PD datasets.A customized kernel function reduces the impact of biased classification towards the majority class(healthy candidates in our consideration).An improved generative adversarial network(IGAN)was designed to generate additional training data to enhance the model’s performance.For performance evaluation,the proposed algorithm achieves a sensitivity of 97.6%and a specificity of 97.3%.The performance comparison is evaluated from five perspectives,including comparisons with different data generation algorithms,feature extraction techniques,kernel functions,and existing works.Results reveal the effectiveness of the IGAN algorithm,which improves the sensitivity and specificity by 4.05%–4.72%and 4.96%–5.86%,respectively;and the effectiveness of the CNN-DSVM algorithm,which improves the sensitivity by 1.24%–57.4%and specificity by 1.04%–163%and reduces biased detection towards the majority class.The ablation experiments confirm the effectiveness of individual components.Two future research directions have also been suggested.
基金supported by the research fund of Hanyang University(HY-202500000001616).
文摘Accurate purchase prediction in e-commerce critically depends on the quality of behavioral features.This paper proposes a layered and interpretable feature engineering framework that organizes user signals into three layers:Basic,Conversion&Stability(efficiency and volatility across actions),and Advanced Interactions&Activity(crossbehavior synergies and intensity).Using real Taobao(Alibaba’s primary e-commerce platform)logs(57,976 records for 10,203 users;25 November–03 December 2017),we conducted a hierarchical,layer-wise evaluation that holds data splits and hyperparameters fixed while varying only the feature set to quantify each layer’s marginal contribution.Across logistic regression(LR),decision tree,random forest,XGBoost,and CatBoost models with stratified 5-fold cross-validation,the performance improvedmonotonically fromBasic to Conversion&Stability to Advanced features.With LR,F1 increased from 0.613(Basic)to 0.962(Advanced);boosted models achieved high discrimination(0.995 AUC Score)and an F1 score up to 0.983.Calibration and precision–recall analyses indicated strong ranking quality and acknowledged potential dataset and period biases given the short(9-day)window.By making feature contributions measurable and reproducible,the framework complements model-centric advances and offers a transparent blueprint for production-grade behavioralmodeling.The code and processed artifacts are publicly available,and future work will extend the validation to longer,seasonal datasets and hybrid approaches that combine automated feature learning with domain-driven design.
基金funded by A’Sharqiyah University,Sultanate of Oman,under Research Project grant number(BFP/RGP/ICT/22/490).
文摘Detecting faces under occlusion remains a significant challenge in computer vision due to variations caused by masks,sunglasses,and other obstructions.Addressing this issue is crucial for applications such as surveillance,biometric authentication,and human-computer interaction.This paper provides a comprehensive review of face detection techniques developed to handle occluded faces.Studies are categorized into four main approaches:feature-based,machine learning-based,deep learning-based,and hybrid methods.We analyzed state-of-the-art studies within each category,examining their methodologies,strengths,and limitations based on widely used benchmark datasets,highlighting their adaptability to partial and severe occlusions.The review also identifies key challenges,including dataset diversity,model generalization,and computational efficiency.Our findings reveal that deep learning methods dominate recent studies,benefiting from their ability to extract hierarchical features and handle complex occlusion patterns.More recently,researchers have increasingly explored Transformer-based architectures,such as Vision Transformer(ViT)and Swin Transformer,to further improve detection robustness under challenging occlusion scenarios.In addition,hybrid approaches,which aim to combine traditional andmodern techniques,are emerging as a promising direction for improving robustness.This review provides valuable insights for researchers aiming to develop more robust face detection systems and for practitioners seeking to deploy reliable solutions in real-world,occlusionprone environments.Further improvements and the proposal of broader datasets are required to developmore scalable,robust,and efficient models that can handle complex occlusions in real-world scenarios.
文摘The aim of this article is to explore potential directions for the development of artificial intelligence(AI).It points out that,while current AI can handle the statistical properties of complex systems,it has difficulty effectively processing and fully representing their spatiotemporal complexity patterns.The article also discusses a potential path of AI development in the engineering domain.Based on the existing understanding of the principles of multilevel com-plexity,this article suggests that consistency among the logical structures of datasets,AI models,model-building software,and hardware will be an important AI development direction and is worthy of careful consideration.
基金supported by the Natural Science Basic Research Program of Shaanxi(Program No.2024JC-YBMS-026).
文摘When dealing with imbalanced datasets,the traditional support vectormachine(SVM)tends to produce a classification hyperplane that is biased towards the majority class,which exhibits poor robustness.This paper proposes a high-performance classification algorithm specifically designed for imbalanced datasets.The proposed method first uses a biased second-order cone programming support vectormachine(B-SOCP-SVM)to identify the support vectors(SVs)and non-support vectors(NSVs)in the imbalanced data.Then,it applies the synthetic minority over-sampling technique(SV-SMOTE)to oversample the support vectors of the minority class and uses the random under-sampling technique(NSV-RUS)multiple times to undersample the non-support vectors of the majority class.Combining the above-obtained minority class data set withmultiple majority class datasets can obtainmultiple new balanced data sets.Finally,SOCP-SVM is used to classify each data set,and the final result is obtained through the integrated algorithm.Experimental results demonstrate that the proposed method performs excellently on imbalanced datasets.
基金supported by the National Key R&D Program of China(2022YFD1401600)the National Science Foundation for Distinguished Young Scholars of Zhejang Province,China(LR23C140001)supported by the Key Area Research and Development Program of Guangdong Province,China(2018B020205003 and 2020B0202090001).
文摘Inferring phylogenetic trees from molecular sequences is a cornerstone of evolutionary biology.Many standard phylogenetic methods(such as maximum-likelihood[ML])rely on explicit models of sequence evolution and thus often suffer from model misspecification or inadequacy.The on-rising deep learning(DL)techniques offer a powerful alternative.Deep learning employs multi-layered artificial neural networks to progressively transform input data into more abstract and complex representations.DL methods can autonomously uncover meaningful patterns from data,thereby bypassing potential biases introduced by predefined features(Franklin,2005;Murphy,2012).Recent efforts have aimed to apply deep neural networks(DNNs)to phylogenetics,with a growing number of applications in tree reconstruction(Suvorov et al.,2020;Zou et al.,2020;Nesterenko et al.,2022;Smith and Hahn,2023;Wang et al.,2023),substitution model selection(Abadi et al.,2020;Burgstaller-Muehlbacher et al.,2023),and diversification rate inference(Voznica et al.,2022;Lajaaiti et al.,2023;Lambert et al.,2023).In phylogenetic tree reconstruction,PhyDL(Zou et al.,2020)and Tree_learning(Suvorov et al.,2020)are two notable DNN-based programs designed to infer unrooted quartet trees directly from alignments of four amino acid(AA)and DNA sequences,respectively.
文摘Climate change significantly affects environment,ecosystems,communities,and economies.These impacts often result in quick and gradual changes in water resources,environmental conditions,and weather patterns.A geographical study was conducted in Arizona State,USA,to examine monthly precipi-tation concentration rates over time.This analysis used a high-resolution 0.50×0.50 grid for monthly precip-itation data from 1961 to 2022,Provided by the Climatic Research Unit.The study aimed to analyze climatic changes affected the first and last five years of each decade,as well as the entire decade,during the specified period.GIS was used to meet the objectives of this study.Arizona experienced 51–568 mm,67–560 mm,63–622 mm,and 52–590 mm of rainfall in the sixth,seventh,eighth,and ninth decades of the second millennium,respectively.Both the first and second five year periods of each decade showed accept-able rainfall amounts despite fluctuations.However,rainfall decreased in the first and second decades of the third millennium.and in the first two years of the third decade.Rainfall amounts dropped to 42–472 mm,55–469 mm,and 74–498 mm,respectively,indicating a downward trend in precipitation.The central part of the state received the highest rainfall,while the eastern and western regions(spanning north to south)had significantly less.Over the decades of the third millennium,the average annual rainfall every five years was relatively low,showing a declining trend due to severe climate changes,generally ranging between 35 mm and 498 mm.The central regions consistently received more rainfall than the eastern and western outskirts.Arizona is currently experiencing a decrease in rainfall due to climate change,a situation that could deterio-rate further.This highlights the need to optimize the use of existing rainfall and explore alternative water sources.
文摘Standardized datasets are foundational to healthcare informatization by enhancing data quality and unleashing the value of data elements.Using bibliometrics and content analysis,this study examines China's healthcare dataset standards from 2011 to 2025.It analyzes their evolution across types,applications,institutions,and themes,highlighting key achievements including substantial growth in quantity,optimized typology,expansion into innovative application scenarios such as health decision support,and broadened institutional involvement.The study also identifies critical challenges,including imbalanced development,insufficient quality control,and a lack of essential metadata—such as authoritative data element mappings and privacy annotations—which hampers the delivery of intelligent services.To address these challenges,the study proposes a multi-faceted strategy focused on optimizing the standard system's architecture,enhancing quality and implementation,and advancing both data governance—through authoritative tracing and privacy protection—and intelligent service provision.These strategies aim to promote the application of dataset standards,thereby fostering and securing the development of new productive forces in healthcare.
文摘Face recognition has emerged as one of the most prominent applications of image analysis and under-standing,gaining considerable attention in recent years.This growing interest is driven by two key factors:its extensive applications in law enforcement and the commercial domain,and the rapid advancement of practical technologies.Despite the significant advancements,modern recognition algorithms still struggle in real-world conditions such as varying lighting conditions,occlusion,and diverse facial postures.In such scenarios,human perception is still well above the capabilities of present technology.Using the systematic mapping study,this paper presents an in-depth review of face detection algorithms and face recognition algorithms,presenting a detailed survey of advancements made between 2015 and 2024.We analyze key methodologies,highlighting their strengths and restrictions in the application context.Additionally,we examine various datasets used for face detection/recognition datasets focusing on the task-specific applications,size,diversity,and complexity.By analyzing these algorithms and datasets,this survey works as a valuable resource for researchers,identifying the research gap in the field of face detection and recognition and outlining potential directions for future research.
基金supported by a project entitled Loess Plateau Region-Watershed-Slope Geological Hazard Multi-Scale Collaborative Intelligent Early Warning System of the National Key R&D Program of China(2022YFC3003404)a project of the Shaanxi Youth Science and Technology Star(2021KJXX-87)public welfare geological survey projects of Shaanxi Institute of Geologic Survey(20180301,201918,202103,and 202413).
文摘This study investigated the impacts of random negative training datasets(NTDs)on the uncertainty of machine learning models for geologic hazard susceptibility assessment of the Loess Plateau,northern Shaanxi Province,China.Based on randomly generated 40 NTDs,the study developed models for the geologic hazard susceptibility assessment using the random forest algorithm and evaluated their performances using the area under the receiver operating characteristic curve(AUC).Specifically,the means and standard deviations of the AUC values from all models were then utilized to assess the overall spatial correlation between the conditioning factors and the susceptibility assessment,as well as the uncertainty introduced by the NTDs.A risk and return methodology was thus employed to quantify and mitigate the uncertainty,with log odds ratios used to characterize the susceptibility assessment levels.The risk and return values were calculated based on the standard deviations and means of the log odds ratios of various locations.After the mean log odds ratios were converted into probability values,the final susceptibility map was plotted,which accounts for the uncertainty induced by random NTDs.The results indicate that the AUC values of the models ranged from 0.810 to 0.963,with an average of 0.852 and a standard deviation of 0.035,indicating encouraging prediction effects and certain uncertainty.The risk and return analysis reveals that low-risk and high-return areas suggest lower standard deviations and higher means across multiple model-derived assessments.Overall,this study introduces a new framework for quantifying the uncertainty of multiple training and evaluation models,aimed at improving their robustness and reliability.Additionally,by identifying low-risk and high-return areas,resource allocation for geologic hazard prevention and control can be optimized,thus ensuring that limited resources are directed toward the most effective prevention and control measures.
文摘Model evaluation using benchmark datasets is an important method to measure the capability of large language models(LLMs)in specific domains,and it is mainly used to assess the knowledge and reasoning abilities of LLMs.Therefore,in order to better assess the capability of LLMs in the agricultural domain,Agri-Eval was proposed as a benchmark for assessing the knowledge and reasoning ability of LLMs in agriculture.The assessment dataset used in Agri-Eval covered seven major disciplines in the agricultural domain:crop science,horticulture,plant protection,animal husbandry,forest science,aquaculture science,and grass science,and contained a total of 2283 questions.Among domestic general-purpose LLMs,DeepSeek R1 performed best with an accuracy rate of 75.49%.In the realm of international general-purpose LLMs,Gemini 2.0 pro exp 0205 standed out as the top performer,achieving an accuracy rate of 74.28%.As an LLMs in agriculture vertical,Shennong V2.0 outperformed all the LLMs in China,and the answer accuracy rate of agricultural knowledge exceeded that of all the existing general-purpose LLMs.The launch of Agri-Eval helped the LLM developers to comprehensively evaluate the model's capability in the field of agriculture through a variety of tasks and tests to promote the development of the LLMs in the field of agriculture.
基金funding support from the National Science and Technology Council(NSTC),under Grant No.114-2410-H-011-026-MY3.
文摘Small datasets are often challenging due to their limited sample size.This research introduces a novel solution to these problems:average linkage virtual sample generation(ALVSG).ALVSG leverages the underlying data structure to create virtual samples,which can be used to augment the original dataset.The ALVSG process consists of two steps.First,an average-linkage clustering technique is applied to the dataset to create a dendrogram.The dendrogram represents the hierarchical structure of the dataset,with each merging operation regarded as a linkage.Next,the linkages are combined into an average-based dataset,which serves as a new representation of the dataset.The second step in the ALVSG process involves generating virtual samples using the average-based dataset.The research project generates a set of 100 virtual samples by uniformly distributing them within the provided boundary.These virtual samples are then added to the original dataset,creating a more extensive dataset with improved generalization performance.The efficacy of the ALVSG approach is validated through resampling experiments and t-tests conducted on two small real-world datasets.The experiments are conducted on three forecasting models:the support vector machine for regression(SVR),the deep learning model(DL),and XGBoost.The results show that the ALVSG approach outperforms the baseline methods in terms of mean square error(MSE),root mean square error(RMSE),and mean absolute error(MAE).
基金supported by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R195)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘With the increasing growth of online news,fake electronic news detection has become one of the most important paradigms of modern research.Traditional electronic news detection techniques are generally based on contextual understanding,sequential dependencies,and/or data imbalance.This makes distinction between genuine and fabricated news a challenging task.To address this problem,we propose a novel hybrid architecture,T5-SA-LSTM,which synergistically integrates the T5 Transformer for semantically rich contextual embedding with the Self-Attentionenhanced(SA)Long Short-Term Memory(LSTM).The LSTM is trained using the Adam optimizer,which provides faster and more stable convergence compared to the Stochastic Gradient Descend(SGD)and Root Mean Square Propagation(RMSProp).The WELFake and FakeNewsPrediction datasets are used,which consist of labeled news articles having fake and real news samples.Tokenization and Synthetic Minority Over-sampling Technique(SMOTE)methods are used for data preprocessing to ensure linguistic normalization and class imbalance.The incorporation of the Self-Attention(SA)mechanism enables the model to highlight critical words and phrases,thereby enhancing predictive accuracy.The proposed model is evaluated using accuracy,precision,recall(sensitivity),and F1-score as performance metrics.The model achieved 99%accuracy on the WELFake dataset and 96.5%accuracy on the FakeNewsPrediction dataset.It outperformed the competitive schemes such as T5-SA-LSTM(RMSProp),T5-SA-LSTM(SGD)and some other models.
文摘In modern industrial production,foreign object detection in complex environments is crucial to ensure product quality and production safety.Detection systems based on deep-learning image processing algorithms often face challenges with handling high-resolution images and achieving accurate detection against complex backgrounds.To address these issues,this study employs the PatchCore unsupervised anomaly detection algorithm combined with data augmentation techniques to enhance the system’s generalization capability across varying lighting conditions,viewing angles,and object scales.The proposed method is evaluated in a complex industrial detection scenario involving the bogie of an electric multiple unit(EMU).A dataset consisting of complex backgrounds,diverse lighting conditions,and multiple viewing angles is constructed to validate the performance of the detection system in real industrial environments.Experimental results show that the proposed model achieves an average area under the receiver operating characteristic curve(AUROC)of 0.92 and an average F1 score of 0.85.Combined with data augmentation,the proposed model exhibits improvements in AUROC by 0.06 and F1 score by 0.03,demonstrating enhanced accuracy and robustness for foreign object detection in complex industrial settings.In addition,the effects of key factors on detection performance are systematically analyzed,providing practical guidance for parameter selection in real industrial applications.
基金supported by the Deanship of Scientific Research,Vice Presidency for Graduate Studies and Scientific Research,King Faisal University,Saudi Arabia Grant No.KFU253765.
文摘Most predictive maintenance studies have emphasized accuracy but provide very little focus on Interpretability or deployment readiness.This study improves on prior methods by developing a small yet robust system that can predict when turbofan engines will fail.It uses the NASA CMAPSS dataset,which has over 200,000 engine cycles from260 engines.The process begins with systematic preprocessing,which includes imputation,outlier removal,scaling,and labelling of the remaining useful life.Dimensionality is reduced using a hybrid selection method that combines variance filtering,recursive elimination,and gradient-boosted importance scores,yielding a stable set of 10 informative sensors.To mitigate class imbalance,minority cases are oversampled,and class-weighted losses are applied during training.Benchmarking is carried out with logistic regression,gradient boosting,and a recurrent design that integrates gated recurrent units with long short-term memory networks.The Long Short-Term Memory–Gated Recurrent Unit(LSTM–GRU)hybrid achieved the strongest performance with an F1 score of 0.92,precision of 0.93,recall of 0.91,ReceiverOperating Characteristic–AreaUnder the Curve(ROC-AUC)of 0.97,andminority recall of 0.75.Interpretability testing using permutation importance and Shapley values indicates that sensors 13,15,and 11 are the most important indicators of engine wear.The proposed system combines imbalance handling,feature reduction,and Interpretability into a practical design suitable for real industrial settings.
文摘In the competitive retail industry of the digital era,data-driven insights into gender-specific customer behavior are essential.They support the optimization of store performance,layout design,product placement,and targeted marketing.However,existing computer vision solutions often rely on facial recognition to gather such insights,raising significant privacy and ethical concerns.To address these issues,this paper presents a privacypreserving customer analytics system through two key strategies.First,we deploy a deep learning framework using YOLOv9s,trained on the RCA-TVGender dataset.Cameras are positioned perpendicular to observation areas to reduce facial visibility while maintaining accurate gender classification.Second,we apply AES-128 encryption to customer position data,ensuring secure access and regulatory compliance.Our system achieved overall performance,with 81.5%mAP@50,77.7%precision,and 75.7%recall.Moreover,a 90-min observational study confirmed the system’s ability to generate privacy-protected heatmaps revealing distinct behavioral patterns between male and female customers.For instance,women spent more time in certain areas and showed interest in different products.These results confirm the system’s effectiveness in enabling personalized layout and marketing strategies without compromising privacy.
基金supported by the GRRC program of Gyeonggi province[GRRC KGU 2023-B01,Research on Intelligent Industrial Data Analytics].
文摘Predictive maintenance(PdM)is vital for ensuring the reliability,safety,and cost efficiency of heavyduty vehicle fleets.However,real-world sensor data are often highly imbalanced,noisy,and temporally irregular,posing significant challenges to model robustness and deployment.Using multivariate time-series data from Scania trucks,this study proposes a novel PdM framework that integrates efficient feature summarization with cost-sensitive hierarchical classification.First,the proposed last_k_summary method transforms recent operational records into compact statistical and trend-based descriptors while preserving missingness,allowing LightGBM to leverage its inherent split rules without ad-hoc imputation.Then,a two-stage LightGBM framework is developed for fault detection and severity classification:Stage A performs safety-prioritized fault screening(normal vs.fault)with a false-negativeweighted objective,and Stage B refines the detected faults into four severity levels through a cascaded hierarchy of binary classifiers.Under the official cost matrix of the IDA Industrial Challenge,the framework achieves total misclassification costs of 36,113(validation)and 36,314(test),outperforming XGBoost and Bi-LSTM by 3.8%-13.5%while maintaining high recall for the safety-critical class(0.83 validation,0.77 test).These results demonstrate that the proposed approach not only improves predictive accuracy but also provides a practical and deployable PdM solution that reduces maintenance cost,enhances fleet safety,and supports data-driven decision-making in industrial environments.
基金supported by the DH2025-TN07-07 project conducted at the Thai Nguyen University of Information and Communication Technology,Thai Nguyen,Vietnam,with additional support from the AI in Software Engineering Lab.
文摘It remains difficult to automate the creation and validation of Unified Modeling Language(UML)dia-grams due to unstructured requirements,limited automated pipelines,and the lack of reliable evaluation methods.This study introduces a cohesive architecture that amalgamates requirement development,UML synthesis,and multimodal validation.First,LLaMA-3.2-1B-Instruct was utilized to generate user-focused requirements.Then,DeepSeek-R1-Distill-Qwen-32B applies its reasoning skills to transform these requirements into PlantUML code.Using this dual-LLM pipeline,we constructed a synthetic dataset of 11,997 UML diagrams spanning six major diagram families.Rendering analysis showed that 89.5%of the generated diagrams compile correctly,while invalid cases were detected automatically.To assess quality,we employed a multimodal scoring method that combines Qwen2.5-VL-3B,LLaMA-3.2-11B-Vision-Instruct and Aya-Vision-8B,with weights based on MMMU performance.A study with 94 experts revealed strong alignment between automatic and manual evaluations,yielding a Pearson correlation of r=0.82 and a Fleiss’Kappa of 0.78.This indicates a high degree of concordance between automated metrics and human judgment.Overall,the results demonstrated that our scoring system is effective and that the proposed generation pipeline produces UML diagrams that are both syntactically correct and semantically coherent.More broadly,the system provides a scalable and reproducible foundation for future work in AI-driven software modeling and multimodal verification.
基金funded by the Deanship of Graduate Studies and Scientific Research at Jouf University under grant No.(DGSSR-2025-02-01296).
文摘Skin diseases affect millions worldwide.Early detection is key to preventing disfigurement,lifelong disability,or death.Dermoscopic images acquired in primary-care settings show high intra-class visual similarity and severe class imbalance,and occasional imaging artifacts can create ambiguity for state-of-the-art convolutional neural networks(CNNs).We frame skin lesion recognition as graph-based reasoning and,to ensure fair evaluation and avoid data leakage,adopt a strict lesion-level partitioning strategy.Each image is first over-segmented using SLIC(Simple Linear Iterative Clustering)to produce perceptually homogeneous superpixels.These superpixels form the nodes of a region-adjacency graph whose edges encode spatial continuity.Node attributes are 1280-dimensional embeddings extracted with a lightweight yet expressive EfficientNet-B0 backbone,providing strong representational power at modest computational cost.The resulting graphs are processed by a five-layer Graph Attention Network(GAT)that learns to weight inter-node relationships dynamically and aggregates multi-hop context before classifying lesions into seven classes with a log-softmax output.Extensive experiments on the DermaMNIST benchmark show the proposed pipeline achieves 88.35%accuracy and 98.04%AUC,outperforming contemporary CNNs,AutoML approaches,and alternative graph neural networks.An ablation study indicates EfficientNet-B0 produces superior node descriptors compared with ResNet-18 and DenseNet,and that roughly five GAT layers strike a good balance between being too shallow and over-deep while avoiding oversmoothing.The method requires no data augmentation or external metadata,making it a drop-in upgrade for clinical computer-aided diagnosis systems.
