To solve problems of poor security guarantee and insufficient training efficiency in the conventional reinforcement learning methods for decision-making,this study proposes a hybrid framework to combine deep reinforce...To solve problems of poor security guarantee and insufficient training efficiency in the conventional reinforcement learning methods for decision-making,this study proposes a hybrid framework to combine deep reinforcement learning with rule-based decision-making methods.A risk assessment model for lane-change maneuvers considering uncertain predictions of surrounding vehicles is established as a safety filter to improve learning efficiency while correcting dangerous actions for safety enhancement.On this basis,a Risk-fused DDQN is constructed utilizing the model-based risk assessment and supervision mechanism.The proposed reinforcement learning algorithm sets up a separate experience buffer for dangerous trials and punishes such actions,which is shown to improve the sampling efficiency and training outcomes.Compared with conventional DDQN methods,the proposed algorithm improves the convergence value of cumulated reward by 7.6%and 2.2%in the two constructed scenarios in the simulation study and reduces the number of training episodes by 52.2%and 66.8%respectively.The success rate of lane change is improved by 57.3%while the time headway is increased at least by 16.5%in real vehicle tests,which confirms the higher training efficiency,scenario adaptability,and security of the proposed Risk-fused DDQN.展开更多
The potential applications of multimodal physiological signals in healthcare,pain monitoring,and clinical decision support systems have garnered significant attention in biomedical research.Subjective self-reporting i...The potential applications of multimodal physiological signals in healthcare,pain monitoring,and clinical decision support systems have garnered significant attention in biomedical research.Subjective self-reporting is the foundation of conventional pain assessment methods,which may be unreliable.Deep learning is a promising alternative to resolve this limitation through automated pain classification.This paper proposes an ensemble deep-learning framework for pain assessment.The framework makes use of features collected from electromyography(EMG),skin conductance level(SCL),and electrocardiography(ECG)signals.We integrate Convolutional Neural Networks(CNN),Long Short-Term Memory Networks(LSTM),Bidirectional Gated Recurrent Units(BiGRU),and Deep Neural Networks(DNN)models.We then aggregate their predictions using a weighted averaging ensemble technique to increase the classification’s robustness.To improve computing efficiency and remove redundant features,we use Particle Swarm Optimization(PSO)for feature selection.This enables us to reduce the features’dimensionality without sacrificing the classification’s accuracy.With improved accuracy,precision,recall,and F1-score across all pain levels,the experimental results show that the suggested ensemble model performs better than individual deep learning classifiers.In our experiments,the suggested model achieved over 98%accuracy,suggesting promising automated pain assessment performance.However,due to differences in validation protocols,comparisons with previous studies are still limited.Combining deep learning and feature selection techniques significantly improves model generalization,reducing overfitting and enhancing classification performance.The evaluation was conducted using the BioVid Heat Pain Dataset,confirming the model’s effectiveness in distinguishing between different pain intensity levels.展开更多
The assessment of beach quality is an important prerequisite for beach development and serves as the foundation for coastal zone management and sustainable development.This topic has attracted widespread attention,and...The assessment of beach quality is an important prerequisite for beach development and serves as the foundation for coastal zone management and sustainable development.This topic has attracted widespread attention,and various evaluation systems have been established.Given that beach quality assessment(BQA)involves multidimensional and nonlinear indicators,machine learning methods are well-suited to handling complex data relationships.However,current research utilizing machine learning for BQA often faces challenges such as limited evaluation indicators and difficulties in obtaining relevant data.in this study,a machine learning-based model for beach quality evaluation is proposed to address the limitations of existing evaluation frameworks,particular-ly under conditions of data scarcity.Simulated data were generated,and the analytic hierarchy process was integrated to extract fea-tures from 21 beach evaluation factors.A comparative analysis was conducted using the following four machine learning models:de-cision tree,random forest,XGBoost,and MLP.Results indicate that XGBoost(mean squared error(MSE)=0.1825,weighted F1=0.7513)and MLP(Pearson coefficient=0.6053)outperform traditional models.Furthermore,an ensemble learning model combining XGBoost and MLP was developed,substantially improving predictive performance(reducing MSE to 0.0753,increasing the Pearson coefficient to 0.8002,and achieving an F1 score of 0.783).Validation using real data from Yangkou Beach demonstrated that the model maintained an accuracy of 58%even when 5–10 evaluation factors had randomly missing values.展开更多
The primary objective of this study is to measure fluoride levels in groundwater samples using machine learning approaches alongside traditional and fuzzy logic models based health risk assessment in the hard rock Arj...The primary objective of this study is to measure fluoride levels in groundwater samples using machine learning approaches alongside traditional and fuzzy logic models based health risk assessment in the hard rock Arjunanadi River basin,South India.Fluoride levels in the study area vary between 0.1 and 3.10 mg/L,with 32 samples exceeding the World Health Organization(WHO)standard of 1.5 mg/L.Hydrogeochemical analyses(Durov and Gibbs)clearly show that the overall water chemistry is primarily influenced by simple dissolution,mixing,and rock-water interactions,indicating that geogenic sources are the predominant contributors to fluoride in the study area.Around 446.5 km^(2)is considered at risk.In predictive analysis,five Machine Learning(ML)models were used,with the AdaBoost model performing better than the other models,achieving 96%accuracy and 4%error rate.The Traditional Health Risk Assessment(THRA)results indicate that 65%of samples pose highly susceptible for dental fluorosis,while 12%of samples pose highly susceptible for skeletal fluorosis in young age groups.The Fuzzy Inference System(FIS)model effectively manages ambiguity and linguistic factors,which are crucial when addressing health risks linked to groundwater fluoride contamination.In this model,input variables include fluoride concentration,individual age,and ingestion rate,while output variables consist of dental caries risk,dental fluorosis,and skeletal fluorosis.The overall results indicate that increased ingestion rates and prolonged exposure to contaminated water make adults and the elderly people vulnerable to dental and skeletal fluorosis,along with very young and young age groups.This study is an essential resource for local authorities,healthcare officials,and communities,aiding in the mitigation of health risks associated with groundwater contamination and enhancing quality of life through improved water management and health risk assessment,aligning with Sustainable Development Goals(SDGs)3 and 6,thereby contributing to a cleaner and healthier society.展开更多
Liquefaction is one of the prominent factors leading to damage to soil and structures.In this study,the rela-tionship between liquefaction potential and soil parameters is determined by applying feature importance met...Liquefaction is one of the prominent factors leading to damage to soil and structures.In this study,the rela-tionship between liquefaction potential and soil parameters is determined by applying feature importance methods to Random Forest(RF),Logistic Regression(LR),Multilayer Perceptron(MLP),Support Vector Machine(SVM)and eXtreme Gradient Boosting(XGBoost)algorithms.Feature importance methods consist of permuta-tion and Shapley Additive exPlanations(SHAP)importances along with the used model’s built-in feature importance method if it exists.These suggested approaches incorporate an extensive dataset of geotechnical parameters,historical liquefaction events,and soil properties.The feature set comprises 18 parameters that are gathered from 161 field cases.Algorithms are used to determine the optimum performance feature set.Compared to other approaches,the study assesses how well these algorithms predict soil liquefaction potential.Early findings show that the algorithms perform well,demonstrating their capacity to identify non-linear connections and improve prediction accuracy.Among the feature set,σ,v(psf),MSF,CSRσ,v,FC%,Vs*,40f t(f ps)and N1,60,CS are the ones that have the highest deterministic power on the result.The study’s contribution is that,in the absence of extensive data for liquefaction assessment,the proposed method estimates the liquefaction potential using five parameters with promising accuracy.展开更多
Due to their high mechanical compliance and excellent biocompatibility,conductive hydrogels exhibit significant potential for applications in flexible electronics.However,as the demand for high sensitivity,superior me...Due to their high mechanical compliance and excellent biocompatibility,conductive hydrogels exhibit significant potential for applications in flexible electronics.However,as the demand for high sensitivity,superior mechanical properties,and strong adhesion performance continues to grow,many conventional fabrication methods remain complex and costly.Herein,we propose a simple and efficient strategy to construct an entangled network hydrogel through a liquid-metal-induced cross-linking reaction,hydrogel demonstrates outstanding properties,including exceptional stretchability(1643%),high tensile strength(366.54 kPa),toughness(350.2 kJ m^(−3)),and relatively low mechanical hysteresis.The hydrogel exhibits long-term stable reusable adhesion(104 kPa),enabling conformal and stable adhesion to human skin.This capability allows it to effectively capture high-quality epidermal electrophysiological signals with high signal-to-noise ratio(25.2 dB)and low impedance(310 ohms).Furthermore,by integrating advanced machine learning algorithms,achieving an attention classification accuracy of 91.38%,which will significantly impact fields like education,healthcare,and artificial intelligence.展开更多
Objective:This study aimed to investigate the learning expectations of undergraduate nursing students regarding the flipped Health Assessment course.Methods:This descriptive,qualitative study was conducted at a medica...Objective:This study aimed to investigate the learning expectations of undergraduate nursing students regarding the flipped Health Assessment course.Methods:This descriptive,qualitative study was conducted at a medical university in Fuzhou,Fujian Province,China.An interview outline was designed based on the core dimensions of Expectation Confirmation Theory(expectation sources,expectation content,and expectation importance).Thirty second-year undergraduate nursing students who had completed first-year basic medical courses and were about to take the flipped Health Assessment course were interviewed between June and July 2022.Interview data were analyzed using qualitative content analysis.Results:Five major themes and thirteen subthemes were identified.Theme 1 was expectation sourcesperceived learning difficulties from past experiences,which included four subthemes:insufficient autonomous learning ability,confusion regarding learning methods,insufficient engagement in learning,and low professional identity.Theme 2 included knowledge and information expectations,which comprised three subthemes:knowledge to improve professional competence,knowledge to enhance academic competitiveness,and knowledge to boost self-efficacy.Theme 3 comprised logical expectations and included two subthemes:flexible teaching methods and efficient instructional tools.Theme 4,pleasure expectations,included two subthemes:vivid teaching styles and diversified teaching evaluations.Theme 5 comprised professional value expectations and included two subthemes:teachers'responsible professional attitudes and gentle emotional support.Conclusion:Students'learning expectations originate from perceived learning difficulties,such as insufficient learning autonomy and engagement,confusion about learning methods,and a lack of professional identity.They showed various learning expectations for the flipped Health Assessment course,including knowledge and information,logic,pleasure,and professional value expectations.展开更多
Most blind image quality assessment(BIQA)methods require a large amount of time to collect human opinion scores as training labels,which limits their usability in practice.Thus,we present an opinion-unaware BIQA metho...Most blind image quality assessment(BIQA)methods require a large amount of time to collect human opinion scores as training labels,which limits their usability in practice.Thus,we present an opinion-unaware BIQA method based on deep reinforcement learning which is trained without subjective scores,named DRL-IQA.Inspired by the human visual perception process,our model is formulated as a quality reinforced agent,which consists of the dynamic distortion generation part and the quality perception part.By considering the image distortion degradation process as a sequential decision-making process,the dynamic distortion generation part can develop a strategy to add as many different distortions as possible to an image,which enriches the distortion space to alleviate overfitting.A reward function calculated from quality degradation after adding distortion is utilized to continuously optimize the strategy.Furthermore,the quality perception part can extract rich quality features from the quality degradation process without using subjective scores,and accurately predict the state values that represent the image quality.Experimental results reveal that our method achieves competitive quality prediction performance compared to other state-of-the-art BIQA methods.展开更多
Background Non-invasive computed tomography angiography(CTA)-based fractional flow reserve(CT-FFR)could become a gatekeeper to invasive coronary angiography.Deep learning(DL)-based CT-FFR has shown promise when compar...Background Non-invasive computed tomography angiography(CTA)-based fractional flow reserve(CT-FFR)could become a gatekeeper to invasive coronary angiography.Deep learning(DL)-based CT-FFR has shown promise when compared to invasive FFR.To evaluate the performance of a DL-based CT-FFR technique,DeepVessel FFR(DVFFR).Methods This retrospective study was designed for iScheMia Assessment based on a Retrospective,single-center Trial of CTFFR(SMART).Patients suspected of stable coronary artery disease(CAD)and undergoing both CTA and invasive FFR examinations were consecutively selected from the Beijing Anzhen Hospital between January 1,2016 to December 30,2018.FFR obtained during invasive coronary angiography was used as the reference standard.DVFFR was calculated blindly using a DL-based CTFFR approach that utilized the complete tree structure of the coronary arteries.Results Three hundred and thirty nine patients(60.5±10.0 years and 209 men)and 414 vessels with direct invasive FFR were included in the analysis.At per-vessel level,sensitivity,specificity,accuracy,positive predictive value(PPV)and negative predictive value(NPV)of DVFFR were 94.7%,88.6%,90.8%,82.7%,and 96.7%,respectively.The area under the receiver operating characteristics curve(AUC)was 0.95 for DVFFR and 0.56 for CTA-based assessment with a significant difference(P<0.0001).At patient level,sensitivity,specificity,accuracy,PPV and NPV of DVFFR were 93.8%,88.0%,90.3%,83.0%,and 95.8%,respectively.The computation for DVFFR was fast with the average time of 22.5±1.9 s.Conclusions The results demonstrate that DVFFR was able to evaluate lesion hemodynamic significance accurately and effectively with improved diagnostic performance over CTA alone.Coronary artery disease(CAD)is a critical disease in which coronary artery luminal narrowing may result in myocardial ischemia.Early and effective assessment of myocardial ischemia is essential for optimal treatment planning so as to improve the quality of life and reduce medical costs.展开更多
Cardiovascular disease(CVD)has gradually become one of the main causes of harm to the life and health of residents.Exploring the influencing factors and risk assessment methods of CVD has become a general trend.In thi...Cardiovascular disease(CVD)has gradually become one of the main causes of harm to the life and health of residents.Exploring the influencing factors and risk assessment methods of CVD has become a general trend.In this paper,a machine learning-based decision-making mechanism for risk assessment of CVD is designed.In this mechanism,the logistics regression analysismethod and factor analysismodel are used to select age,obesity degree,blood pressure,blood fat,blood sugar,smoking status,drinking status,and exercise status as the main pathogenic factors of CVD,and an index systemof risk assessment for CVD is established.Then,a two-stage model combining K-means cluster analysis and random forest(RF)is proposed to evaluate and predict the risk of CVD,and the predicted results are compared with the methods of Bayesian discrimination,K-means cluster analysis and RF.The results show that thepredictioneffect of theproposedtwo-stagemodel is better than that of the comparedmethods.Moreover,several suggestions for the government,the medical industry and the public are provided based on the research results.展开更多
Machine learning is currently one of the research hotspots in the field of landslide prediction.To clarify and evaluate the differences in characteristics and prediction effects of different machine learning models,Co...Machine learning is currently one of the research hotspots in the field of landslide prediction.To clarify and evaluate the differences in characteristics and prediction effects of different machine learning models,Conghua District,which is the most prone to landslide disasters in Guangzhou,was selected for landslide susceptibility evaluation.The evaluation factors were selected by using correlation analysis and variance expansion factor method.Applying four machine learning methods namely Logistic Regression(LR),Random Forest(RF),Support Vector Machines(SVM),and Extreme Gradient Boosting(XGB),landslide models were constructed.Comparative analysis and evaluation of the model were conducted through statistical indices and receiver operating characteristic(ROC)curves.The results showed that LR,RF,SVM,and XGB models have good predictive performance for landslide susceptibility,with the area under curve(AUC)values of 0.752,0.965,0.996,and 0.998,respectively.XGB model had the highest predictive ability,followed by RF model,SVM model,and LR model.The frequency ratio(FR)accuracy of LR,RF,SVM,and XGB models was 0.775,0.842,0.759,and 0.822,respectively.RF and XGB models were superior to LR and SVM models,indicating that the integrated algorithm has better predictive ability than a single classification algorithm in regional landslide classification problems.展开更多
The rapid advancement of gas sensitive properties in metal oxides is crucial for detecting hazardous gases in industrial and coal mining environments.However,the conventional experimental trial and error approach pose...The rapid advancement of gas sensitive properties in metal oxides is crucial for detecting hazardous gases in industrial and coal mining environments.However,the conventional experimental trial and error approach poses significant challenges and resource consumption for the high throughput screening of gas sensitive materials.Consequently,this paper introduced a novel screening approach that integrates first principles with machine learning(ML)to rapidly predict the gas sensitivity of materials.Initially,a comprehensive database of multi-physical parameters was established by modeling various adsorption sites on the surface of WO3,which serves as a representative material.Since density functional theory(DFT)is one of the first principles,DFT calculations were conducted to derive essential multi-physical parameters,including bandgap,density of states(DOS),Fermi level,adsorption energy,and structural modifications resulting from adsorption.The collected data was subsequently utilized to develop a cor-relation model linking the multi-physical parameters to gas sensitive performance using intelligent algo-rithms.The model’s performance was assessed through receiver operating characteristic(ROC)curves,confusion matrices,and other evaluation metrics,ultimately achieving a prediction accuracy of 90%for identifying key features influencing gas adsorption performance.This proposed strategy for predicting the gas sensitive characteristics of materials holds significant potential for application in identifying addi-tional gas sensitive properties across various materials.展开更多
As the global demand for renewable energy grows,solar energy is gaining attention as a clean,sustainable energy source.Accurate assessment of solar energy resources is crucial for the siting and design of photovoltaic...As the global demand for renewable energy grows,solar energy is gaining attention as a clean,sustainable energy source.Accurate assessment of solar energy resources is crucial for the siting and design of photovoltaic power plants.This study proposes an integrated deep learning-based photovoltaic resource assessment method.Ensemble learning and deep learning methods are fused for photovoltaic resource assessment for the first time.The proposed method combines the random forest,gated recurrent unit,and long short-term memory to effectively improve the accuracy and reliability of photovoltaic resource assessment.The proposed method has strong adaptability and high accuracy even in the photovoltaic resource assessment of complex terrain and landscape.The experimental results show that the proposed method outperforms the comparison algorithm in all evaluation indexes,indicating that the proposed method has higher accuracy and reliability in photovoltaic resource assessment with improved generalization performance traditional single algorithm.展开更多
Zambia like any other country in most African regions is still grappling with the dynamics of harnessing technology for the betterment of Higher Education. The onset of the Covid 19 pandemic brought a test for the pre...Zambia like any other country in most African regions is still grappling with the dynamics of harnessing technology for the betterment of Higher Education. The onset of the Covid 19 pandemic brought a test for the preparedness of the Zambian Higher Education Institutions (HEIs) in harnessing technology for pedagogical activities. As countries worldwide switched to electronic learning during the pandemic, the same could not be said for Zambian HEIs. Zambian HEIs struggled to conduct pedagogical activities on learning management platforms. This study investigated the factors affecting the implementation and assessment of learning Management systems in Zambia’s HEIs. With its focus on assessing: 1) system features, 2) compliance with regulatory standards, 3) quality of service and 4) technology acceptance as the four key assessment areas of an LMS, this article proposed a model for assessing learning management systems in Zambian HEIs. To test the proposed model, a software tool was also developed.展开更多
Background:Frailty in older adults is linked to increased risks and lower quality of life.Pre-frailty,a condition preceding frailty,is intervenable,but its determinants and assessment are challenging.This study aims t...Background:Frailty in older adults is linked to increased risks and lower quality of life.Pre-frailty,a condition preceding frailty,is intervenable,but its determinants and assessment are challenging.This study aims to develop and validate an explainable machine learning model for pre-frailty risk assessment among community-dwelling older adults.Methods:The study included 3141 adults aged 60 or above from the China Health and Retirement Longitudinal Study.Pre-frailty was characterized by one or two criteria from the physical frailty phenotype scale.We extracted 80 distinct features across seven dimensions to evaluate pre-frailty risk.A model was constructed using recursive feature elimination and a stacking-CatBoost distillation module on 80%of the sample and validated on a separate 20%holdout data set.Results:The study used data from 2508 community-dwelling older adults(mean age,67.24 years[range,60–96];1215[48.44%]females)to develop a pre-frailty risk assessment model.We selected 57 predictive features and built a distilled CatBoost model,which achieved the highest discrimination(AUROC:0.7560[95%CI:0.7169,0.7928])on the 20%holdout data set.The living city,BMI,and peak expiratory flow(PEF)were the three most significant contributors to pre-frailty risk.Physical and environmental factors were the top 2 impactful feature dimensions.Conclusions:An accurate and interpretable pre-frailty risk assessment framework using state-of-the-art machine learning techniques and explanation methods has been developed.Our framework incorporates a wide range of features and determinants,allowing for a comprehensive and nuanced understanding of pre-frailty risk.展开更多
BACKGROUND Preoperative risk assessments are vital for identifying patients at high risk of postoperative mortality.However,traditional scoring systems can be time consuming.We hypothesized that the use of machine lea...BACKGROUND Preoperative risk assessments are vital for identifying patients at high risk of postoperative mortality.However,traditional scoring systems can be time consuming.We hypothesized that the use of machine learning models would enable rapid and accurate risk assessments to be performed.AIM To assess the potential of machine learning algorithms to develop predictive models of mortality risk after abdominal surgery.METHODS This retrospective study included 230 individuals who underwent abdominal surgery at the Seventh People’s Hospital of Shanghai University of Traditional Chinese Medicine between January 2023 and December 2023.Demographic and surgery-related data were collected and used to develop nomogram,decision-tree,random-forest,gradient-boosting,support vector machine,and naïve Bayesian models to predict 30-day mortality risk after abdominal surgery.Models were assessed using receiver operating characteristic curves and compared using the DeLong test.RESULTS Of the 230 included patients,52 died and 178 survived.Models were developed using the training cohort(n=161)and assessed using the validation cohort(n=68).The areas under the receiver operating characteristic curves for the nomogram,decision-tree,random-forest,gradient-boosting tree,support vector machine,and naïve Bayesian models were 0.908[95%confidence interval(CI):0.824-0.992],0.874(95%CI:0.785-0.963),0.928(95%CI:0.869-0.987),0.907(95%CI:0.837-0.976),0.983(95%CI:0.959-1.000),and 0.807(95%CI:0.702-0.911),respectively.CONCLUSION Nomogram,random-forest,gradient-boosting tree,and support vector machine models all demonstrate strong performances for the prediction of postoperative mortality and can be selected based on the clinical circumstances.展开更多
Hurricanes are one of the most destructive natural disasters that can cause catastrophic losses to both communities and infrastructure.Assessment of hurricane risk furnishes a spatial depiction of the interplay among ...Hurricanes are one of the most destructive natural disasters that can cause catastrophic losses to both communities and infrastructure.Assessment of hurricane risk furnishes a spatial depiction of the interplay among hazard,vulnerability,exposure,and mitigation capacity,crucial for understanding and managing the risks hurricanes pose to communities.These assessments aid in gauging the efficacy of existing hurricane mitigation strategies and gauging their resilience across diverse climate change scenarios.A systematic review was conducted,encompassing 94 articles,to scrutinize the structure,data inputs,assumptions,methodologies,perils modelled,and key predictors of hurricane risk.This review identified key research gaps essential for enhancing future risk assessments.The complex interaction between hurricane perils may be disastrous and underestimated in the majority of risk assessments which focus on a single peril,commonly storm surge and flood,resulting in inadequacies in disaster resilience planning.Most risk assessments were based on hurricane frequency rather than hurricane damage,which is more insightful for policymakers.Furthermore,considering secondary indirect impacts stemming from hurricanes,including real estate market and business interruption,could enrich economic impact assessments.Hurricane mitigation measures were the most under-utilised category of predictors leveraged in only 5%of studies.The top six predictive factors for hurricane risk were land use,slope,precipitation,elevation,population density,and soil texture/drainage.Another notable research gap identified was the potential of machine learning techniques in risk assessments,offering advantages over traditional MCDM and numerical models due to their ability to capture complex nonlinear relationships and adaptability to different study regions.Existing machine learning based risk assessments leverage random forest models(42%of studies)followed by neural network models(19%of studies),with further research required to investigate diverse machine learning algorithms such as ensemble models.A further research gap is model validation,in particular assessing transferability to a new study region.Additionally,harnessing simulated data and refining projections related to demographic and built environment dynamics can bolster the sophistication of climate change scenario assessments.By addressing these research gaps,hurricane risk assessments can furnish invaluable insights for national policymakers,facilitating the development of robust hurricane mitigation strategies and the construction of hurricane-resilient communities.To the authors’knowledge,this represents the first literature review specifically dedicated to quantitative hurricane risk assessments,encompassing a comparison of Multi-criteria Decision Making(MCDM),numerical models,and machine learning models.Ultimately,advancements in hurricane risk assessments and modelling stand poised to mitigate potential losses to communities and infrastructure both in the immediate and long-term future.展开更多
We propose Mixed-Precision Multibranch Network(M+MNet)to compensate for the neglect of background information in image aesthetics assessment(IAA)while providing strategies for overcoming the dilemma between training c...We propose Mixed-Precision Multibranch Network(M+MNet)to compensate for the neglect of background information in image aesthetics assessment(IAA)while providing strategies for overcoming the dilemma between training costs and performance.First,two exponentially weighted pooling methods are used to selectively boost the extraction of background and salient information during downsampling.Second,we propose Corner Grid,an unsupervised data augmentation method that leverages the diffusive characteristics of convolution to force the network to seek more relevant background information.Third,we perform mixed-precision training by switching the precision format,thus significantly reducing the time and memory consumption of data representation and transmission.Most of our methods specifically designed for IAA tasks have demonstrated generalizability to other IAA works.For performance verification,we develop a large-scale benchmark(the most comprehensive thus far)by comparing 17 methods with M+MNet on two representative datasets:the Aesthetic Visual Analysis(AVA)dataset and FLICKR-Aesthetic Evaluation Subset(FLICKR-AES).M+MNet achieves state-of-the-art performance on all tasks.展开更多
The 2019 Typhoon Lekima triggered extensive landslides in Zhejiang Province.To explore the impact of typhoon paths on the distribution of landslide susceptibility,this study proposes a spatiotemporal zoning assessment...The 2019 Typhoon Lekima triggered extensive landslides in Zhejiang Province.To explore the impact of typhoon paths on the distribution of landslide susceptibility,this study proposes a spatiotemporal zoning assessment framework based on typhoon paths and inner rainbands.According to the typhoon landing path and its rainfall impact range,the study area is divided into the typhoon event period(TEP)and the annual non-typhoon period(ANP).The model uses 14 environmental factors,with the only difference between TEP and ANP being the rainfall index:TEP uses 48-hour rainfall during the typhoon,while ANP uses multi-year average annual rainfall.Modeling and comparative analysis were conducted using six machine learning models including random forest(RF)and support vector machine(SVM).The results show that the distribution pattern of high-risk landslide areas during TEP is significantly correlated with typhoon intensity:when the intensity is level 12,high-risk areas are radially distributed;at levels 10-11,they tend to concentrate asymmetrically along the coast;and when the intensity drops to below level 9,the overall susceptibility decreases significantly.During ANP,the distribution of landslides is relatively uniform with no obvious spatial concentration.Analysis on the factor contribution rate indicates that the rainfall weight in TEP is as high as 32.1%,making it the dominant factor;in ANP,the rainfall weight drops to 13.6%while the influence of factors such as slope and topographic wetness index increases,revealing differences in landslide formation mechanisms between the two periods.This study demonstrates that the spatiotemporal zoning method based on typhoon paths can effectively characterize the spatial susceptibility patterns of landslides and improve disaster identification capabilities under extreme weather conditions.The finally generated annual susceptibility zoning map divides the study area into four types of risk regions,providing a reference for dynamic monitoring and differentiated risk management of landslides in typhoon-prone areas.展开更多
基金Supported by National Key Research and Development Program of China(Grant No.2022YFE0117100)National Science Foundation of China(Grant No.52102468,52325212)Fundamental Research Funds for the Central Universities。
文摘To solve problems of poor security guarantee and insufficient training efficiency in the conventional reinforcement learning methods for decision-making,this study proposes a hybrid framework to combine deep reinforcement learning with rule-based decision-making methods.A risk assessment model for lane-change maneuvers considering uncertain predictions of surrounding vehicles is established as a safety filter to improve learning efficiency while correcting dangerous actions for safety enhancement.On this basis,a Risk-fused DDQN is constructed utilizing the model-based risk assessment and supervision mechanism.The proposed reinforcement learning algorithm sets up a separate experience buffer for dangerous trials and punishes such actions,which is shown to improve the sampling efficiency and training outcomes.Compared with conventional DDQN methods,the proposed algorithm improves the convergence value of cumulated reward by 7.6%and 2.2%in the two constructed scenarios in the simulation study and reduces the number of training episodes by 52.2%and 66.8%respectively.The success rate of lane change is improved by 57.3%while the time headway is increased at least by 16.5%in real vehicle tests,which confirms the higher training efficiency,scenario adaptability,and security of the proposed Risk-fused DDQN.
基金funded by the Deanship of Graduate Studies and Scientific Research at Jouf University under grant No.(DGSSR-2023-02-02341).
文摘The potential applications of multimodal physiological signals in healthcare,pain monitoring,and clinical decision support systems have garnered significant attention in biomedical research.Subjective self-reporting is the foundation of conventional pain assessment methods,which may be unreliable.Deep learning is a promising alternative to resolve this limitation through automated pain classification.This paper proposes an ensemble deep-learning framework for pain assessment.The framework makes use of features collected from electromyography(EMG),skin conductance level(SCL),and electrocardiography(ECG)signals.We integrate Convolutional Neural Networks(CNN),Long Short-Term Memory Networks(LSTM),Bidirectional Gated Recurrent Units(BiGRU),and Deep Neural Networks(DNN)models.We then aggregate their predictions using a weighted averaging ensemble technique to increase the classification’s robustness.To improve computing efficiency and remove redundant features,we use Particle Swarm Optimization(PSO)for feature selection.This enables us to reduce the features’dimensionality without sacrificing the classification’s accuracy.With improved accuracy,precision,recall,and F1-score across all pain levels,the experimental results show that the suggested ensemble model performs better than individual deep learning classifiers.In our experiments,the suggested model achieved over 98%accuracy,suggesting promising automated pain assessment performance.However,due to differences in validation protocols,comparisons with previous studies are still limited.Combining deep learning and feature selection techniques significantly improves model generalization,reducing overfitting and enhancing classification performance.The evaluation was conducted using the BioVid Heat Pain Dataset,confirming the model’s effectiveness in distinguishing between different pain intensity levels.
基金supported by the National Natural Science Foundation of China(Nos.82202299,62203060,62403492).
文摘The assessment of beach quality is an important prerequisite for beach development and serves as the foundation for coastal zone management and sustainable development.This topic has attracted widespread attention,and various evaluation systems have been established.Given that beach quality assessment(BQA)involves multidimensional and nonlinear indicators,machine learning methods are well-suited to handling complex data relationships.However,current research utilizing machine learning for BQA often faces challenges such as limited evaluation indicators and difficulties in obtaining relevant data.in this study,a machine learning-based model for beach quality evaluation is proposed to address the limitations of existing evaluation frameworks,particular-ly under conditions of data scarcity.Simulated data were generated,and the analytic hierarchy process was integrated to extract fea-tures from 21 beach evaluation factors.A comparative analysis was conducted using the following four machine learning models:de-cision tree,random forest,XGBoost,and MLP.Results indicate that XGBoost(mean squared error(MSE)=0.1825,weighted F1=0.7513)and MLP(Pearson coefficient=0.6053)outperform traditional models.Furthermore,an ensemble learning model combining XGBoost and MLP was developed,substantially improving predictive performance(reducing MSE to 0.0753,increasing the Pearson coefficient to 0.8002,and achieving an F1 score of 0.783).Validation using real data from Yangkou Beach demonstrated that the model maintained an accuracy of 58%even when 5–10 evaluation factors had randomly missing values.
基金the Anusandhan National Research Foundation(ANRF),New Delhi[Erstwhile,Science and Engineering Research Board(SERB)]Department of Science and Technology(DST)(Government of India)(File No.:CRG/2022/002618 Dated:22.08.2023)for providing the grant and support to carry out this work effectively.
文摘The primary objective of this study is to measure fluoride levels in groundwater samples using machine learning approaches alongside traditional and fuzzy logic models based health risk assessment in the hard rock Arjunanadi River basin,South India.Fluoride levels in the study area vary between 0.1 and 3.10 mg/L,with 32 samples exceeding the World Health Organization(WHO)standard of 1.5 mg/L.Hydrogeochemical analyses(Durov and Gibbs)clearly show that the overall water chemistry is primarily influenced by simple dissolution,mixing,and rock-water interactions,indicating that geogenic sources are the predominant contributors to fluoride in the study area.Around 446.5 km^(2)is considered at risk.In predictive analysis,five Machine Learning(ML)models were used,with the AdaBoost model performing better than the other models,achieving 96%accuracy and 4%error rate.The Traditional Health Risk Assessment(THRA)results indicate that 65%of samples pose highly susceptible for dental fluorosis,while 12%of samples pose highly susceptible for skeletal fluorosis in young age groups.The Fuzzy Inference System(FIS)model effectively manages ambiguity and linguistic factors,which are crucial when addressing health risks linked to groundwater fluoride contamination.In this model,input variables include fluoride concentration,individual age,and ingestion rate,while output variables consist of dental caries risk,dental fluorosis,and skeletal fluorosis.The overall results indicate that increased ingestion rates and prolonged exposure to contaminated water make adults and the elderly people vulnerable to dental and skeletal fluorosis,along with very young and young age groups.This study is an essential resource for local authorities,healthcare officials,and communities,aiding in the mitigation of health risks associated with groundwater contamination and enhancing quality of life through improved water management and health risk assessment,aligning with Sustainable Development Goals(SDGs)3 and 6,thereby contributing to a cleaner and healthier society.
文摘Liquefaction is one of the prominent factors leading to damage to soil and structures.In this study,the rela-tionship between liquefaction potential and soil parameters is determined by applying feature importance methods to Random Forest(RF),Logistic Regression(LR),Multilayer Perceptron(MLP),Support Vector Machine(SVM)and eXtreme Gradient Boosting(XGBoost)algorithms.Feature importance methods consist of permuta-tion and Shapley Additive exPlanations(SHAP)importances along with the used model’s built-in feature importance method if it exists.These suggested approaches incorporate an extensive dataset of geotechnical parameters,historical liquefaction events,and soil properties.The feature set comprises 18 parameters that are gathered from 161 field cases.Algorithms are used to determine the optimum performance feature set.Compared to other approaches,the study assesses how well these algorithms predict soil liquefaction potential.Early findings show that the algorithms perform well,demonstrating their capacity to identify non-linear connections and improve prediction accuracy.Among the feature set,σ,v(psf),MSF,CSRσ,v,FC%,Vs*,40f t(f ps)and N1,60,CS are the ones that have the highest deterministic power on the result.The study’s contribution is that,in the absence of extensive data for liquefaction assessment,the proposed method estimates the liquefaction potential using five parameters with promising accuracy.
基金supported by the National Key Research&Development Program of China(grant no.2022YFC3500503)the National Natural Science Foundation of China(grant nos.62227807,12374171,12004034,62402041)+2 种基金the Beijing Institute of Technology Research Fund Program for Young Scholars,Chinathe Fundamental Research Funds for the Central Universities(grant nos.2024CX06060)Beijing Youth Talent Lifting Project.
文摘Due to their high mechanical compliance and excellent biocompatibility,conductive hydrogels exhibit significant potential for applications in flexible electronics.However,as the demand for high sensitivity,superior mechanical properties,and strong adhesion performance continues to grow,many conventional fabrication methods remain complex and costly.Herein,we propose a simple and efficient strategy to construct an entangled network hydrogel through a liquid-metal-induced cross-linking reaction,hydrogel demonstrates outstanding properties,including exceptional stretchability(1643%),high tensile strength(366.54 kPa),toughness(350.2 kJ m^(−3)),and relatively low mechanical hysteresis.The hydrogel exhibits long-term stable reusable adhesion(104 kPa),enabling conformal and stable adhesion to human skin.This capability allows it to effectively capture high-quality epidermal electrophysiological signals with high signal-to-noise ratio(25.2 dB)and low impedance(310 ohms).Furthermore,by integrating advanced machine learning algorithms,achieving an attention classification accuracy of 91.38%,which will significantly impact fields like education,healthcare,and artificial intelligence.
基金This research received grant from 2022 Undergraduate Education and Teaching Research Project of Fujian Medical University(22028).
文摘Objective:This study aimed to investigate the learning expectations of undergraduate nursing students regarding the flipped Health Assessment course.Methods:This descriptive,qualitative study was conducted at a medical university in Fuzhou,Fujian Province,China.An interview outline was designed based on the core dimensions of Expectation Confirmation Theory(expectation sources,expectation content,and expectation importance).Thirty second-year undergraduate nursing students who had completed first-year basic medical courses and were about to take the flipped Health Assessment course were interviewed between June and July 2022.Interview data were analyzed using qualitative content analysis.Results:Five major themes and thirteen subthemes were identified.Theme 1 was expectation sourcesperceived learning difficulties from past experiences,which included four subthemes:insufficient autonomous learning ability,confusion regarding learning methods,insufficient engagement in learning,and low professional identity.Theme 2 included knowledge and information expectations,which comprised three subthemes:knowledge to improve professional competence,knowledge to enhance academic competitiveness,and knowledge to boost self-efficacy.Theme 3 comprised logical expectations and included two subthemes:flexible teaching methods and efficient instructional tools.Theme 4,pleasure expectations,included two subthemes:vivid teaching styles and diversified teaching evaluations.Theme 5 comprised professional value expectations and included two subthemes:teachers'responsible professional attitudes and gentle emotional support.Conclusion:Students'learning expectations originate from perceived learning difficulties,such as insufficient learning autonomy and engagement,confusion about learning methods,and a lack of professional identity.They showed various learning expectations for the flipped Health Assessment course,including knowledge and information,logic,pleasure,and professional value expectations.
基金supported by the Fundamental Research Funds for the Central Universities.
文摘Most blind image quality assessment(BIQA)methods require a large amount of time to collect human opinion scores as training labels,which limits their usability in practice.Thus,we present an opinion-unaware BIQA method based on deep reinforcement learning which is trained without subjective scores,named DRL-IQA.Inspired by the human visual perception process,our model is formulated as a quality reinforced agent,which consists of the dynamic distortion generation part and the quality perception part.By considering the image distortion degradation process as a sequential decision-making process,the dynamic distortion generation part can develop a strategy to add as many different distortions as possible to an image,which enriches the distortion space to alleviate overfitting.A reward function calculated from quality degradation after adding distortion is utilized to continuously optimize the strategy.Furthermore,the quality perception part can extract rich quality features from the quality degradation process without using subjective scores,and accurately predict the state values that represent the image quality.Experimental results reveal that our method achieves competitive quality prediction performance compared to other state-of-the-art BIQA methods.
文摘Background Non-invasive computed tomography angiography(CTA)-based fractional flow reserve(CT-FFR)could become a gatekeeper to invasive coronary angiography.Deep learning(DL)-based CT-FFR has shown promise when compared to invasive FFR.To evaluate the performance of a DL-based CT-FFR technique,DeepVessel FFR(DVFFR).Methods This retrospective study was designed for iScheMia Assessment based on a Retrospective,single-center Trial of CTFFR(SMART).Patients suspected of stable coronary artery disease(CAD)and undergoing both CTA and invasive FFR examinations were consecutively selected from the Beijing Anzhen Hospital between January 1,2016 to December 30,2018.FFR obtained during invasive coronary angiography was used as the reference standard.DVFFR was calculated blindly using a DL-based CTFFR approach that utilized the complete tree structure of the coronary arteries.Results Three hundred and thirty nine patients(60.5±10.0 years and 209 men)and 414 vessels with direct invasive FFR were included in the analysis.At per-vessel level,sensitivity,specificity,accuracy,positive predictive value(PPV)and negative predictive value(NPV)of DVFFR were 94.7%,88.6%,90.8%,82.7%,and 96.7%,respectively.The area under the receiver operating characteristics curve(AUC)was 0.95 for DVFFR and 0.56 for CTA-based assessment with a significant difference(P<0.0001).At patient level,sensitivity,specificity,accuracy,PPV and NPV of DVFFR were 93.8%,88.0%,90.3%,83.0%,and 95.8%,respectively.The computation for DVFFR was fast with the average time of 22.5±1.9 s.Conclusions The results demonstrate that DVFFR was able to evaluate lesion hemodynamic significance accurately and effectively with improved diagnostic performance over CTA alone.Coronary artery disease(CAD)is a critical disease in which coronary artery luminal narrowing may result in myocardial ischemia.Early and effective assessment of myocardial ischemia is essential for optimal treatment planning so as to improve the quality of life and reduce medical costs.
基金This work is supported by the National Natural Science Foundation of China(Nos.72071150,71871174).
文摘Cardiovascular disease(CVD)has gradually become one of the main causes of harm to the life and health of residents.Exploring the influencing factors and risk assessment methods of CVD has become a general trend.In this paper,a machine learning-based decision-making mechanism for risk assessment of CVD is designed.In this mechanism,the logistics regression analysismethod and factor analysismodel are used to select age,obesity degree,blood pressure,blood fat,blood sugar,smoking status,drinking status,and exercise status as the main pathogenic factors of CVD,and an index systemof risk assessment for CVD is established.Then,a two-stage model combining K-means cluster analysis and random forest(RF)is proposed to evaluate and predict the risk of CVD,and the predicted results are compared with the methods of Bayesian discrimination,K-means cluster analysis and RF.The results show that thepredictioneffect of theproposedtwo-stagemodel is better than that of the comparedmethods.Moreover,several suggestions for the government,the medical industry and the public are provided based on the research results.
基金supported by the projects of the China Geological Survey(DD20221729,DD20190291)Zhuhai Urban Geological Survey(including informatization)(MZCD–2201–008).
文摘Machine learning is currently one of the research hotspots in the field of landslide prediction.To clarify and evaluate the differences in characteristics and prediction effects of different machine learning models,Conghua District,which is the most prone to landslide disasters in Guangzhou,was selected for landslide susceptibility evaluation.The evaluation factors were selected by using correlation analysis and variance expansion factor method.Applying four machine learning methods namely Logistic Regression(LR),Random Forest(RF),Support Vector Machines(SVM),and Extreme Gradient Boosting(XGB),landslide models were constructed.Comparative analysis and evaluation of the model were conducted through statistical indices and receiver operating characteristic(ROC)curves.The results showed that LR,RF,SVM,and XGB models have good predictive performance for landslide susceptibility,with the area under curve(AUC)values of 0.752,0.965,0.996,and 0.998,respectively.XGB model had the highest predictive ability,followed by RF model,SVM model,and LR model.The frequency ratio(FR)accuracy of LR,RF,SVM,and XGB models was 0.775,0.842,0.759,and 0.822,respectively.RF and XGB models were superior to LR and SVM models,indicating that the integrated algorithm has better predictive ability than a single classification algorithm in regional landslide classification problems.
基金funded by National Natural Science Foundation of China(No.52303356)Natural Science Foundation of Jiangsu Province(No.BK20210494)+2 种基金Natural Science Foundation of Jiangsu Province(No.BK20221115)National Key Research and Development Project(No.2000YFC2006601)Scientific Research Foundation for Excellent Talents of Xuzhou Medical University(No.D2019032).
文摘The rapid advancement of gas sensitive properties in metal oxides is crucial for detecting hazardous gases in industrial and coal mining environments.However,the conventional experimental trial and error approach poses significant challenges and resource consumption for the high throughput screening of gas sensitive materials.Consequently,this paper introduced a novel screening approach that integrates first principles with machine learning(ML)to rapidly predict the gas sensitivity of materials.Initially,a comprehensive database of multi-physical parameters was established by modeling various adsorption sites on the surface of WO3,which serves as a representative material.Since density functional theory(DFT)is one of the first principles,DFT calculations were conducted to derive essential multi-physical parameters,including bandgap,density of states(DOS),Fermi level,adsorption energy,and structural modifications resulting from adsorption.The collected data was subsequently utilized to develop a cor-relation model linking the multi-physical parameters to gas sensitive performance using intelligent algo-rithms.The model’s performance was assessed through receiver operating characteristic(ROC)curves,confusion matrices,and other evaluation metrics,ultimately achieving a prediction accuracy of 90%for identifying key features influencing gas adsorption performance.This proposed strategy for predicting the gas sensitive characteristics of materials holds significant potential for application in identifying addi-tional gas sensitive properties across various materials.
基金funded by Key-Area Research and Development Program Project of Guangdong Province (2021B0101230003)China Southern Power Grid Science and Technology Project (ZBKJXM20220004).
文摘As the global demand for renewable energy grows,solar energy is gaining attention as a clean,sustainable energy source.Accurate assessment of solar energy resources is crucial for the siting and design of photovoltaic power plants.This study proposes an integrated deep learning-based photovoltaic resource assessment method.Ensemble learning and deep learning methods are fused for photovoltaic resource assessment for the first time.The proposed method combines the random forest,gated recurrent unit,and long short-term memory to effectively improve the accuracy and reliability of photovoltaic resource assessment.The proposed method has strong adaptability and high accuracy even in the photovoltaic resource assessment of complex terrain and landscape.The experimental results show that the proposed method outperforms the comparison algorithm in all evaluation indexes,indicating that the proposed method has higher accuracy and reliability in photovoltaic resource assessment with improved generalization performance traditional single algorithm.
文摘Zambia like any other country in most African regions is still grappling with the dynamics of harnessing technology for the betterment of Higher Education. The onset of the Covid 19 pandemic brought a test for the preparedness of the Zambian Higher Education Institutions (HEIs) in harnessing technology for pedagogical activities. As countries worldwide switched to electronic learning during the pandemic, the same could not be said for Zambian HEIs. Zambian HEIs struggled to conduct pedagogical activities on learning management platforms. This study investigated the factors affecting the implementation and assessment of learning Management systems in Zambia’s HEIs. With its focus on assessing: 1) system features, 2) compliance with regulatory standards, 3) quality of service and 4) technology acceptance as the four key assessment areas of an LMS, this article proposed a model for assessing learning management systems in Zambian HEIs. To test the proposed model, a software tool was also developed.
文摘Background:Frailty in older adults is linked to increased risks and lower quality of life.Pre-frailty,a condition preceding frailty,is intervenable,but its determinants and assessment are challenging.This study aims to develop and validate an explainable machine learning model for pre-frailty risk assessment among community-dwelling older adults.Methods:The study included 3141 adults aged 60 or above from the China Health and Retirement Longitudinal Study.Pre-frailty was characterized by one or two criteria from the physical frailty phenotype scale.We extracted 80 distinct features across seven dimensions to evaluate pre-frailty risk.A model was constructed using recursive feature elimination and a stacking-CatBoost distillation module on 80%of the sample and validated on a separate 20%holdout data set.Results:The study used data from 2508 community-dwelling older adults(mean age,67.24 years[range,60–96];1215[48.44%]females)to develop a pre-frailty risk assessment model.We selected 57 predictive features and built a distilled CatBoost model,which achieved the highest discrimination(AUROC:0.7560[95%CI:0.7169,0.7928])on the 20%holdout data set.The living city,BMI,and peak expiratory flow(PEF)were the three most significant contributors to pre-frailty risk.Physical and environmental factors were the top 2 impactful feature dimensions.Conclusions:An accurate and interpretable pre-frailty risk assessment framework using state-of-the-art machine learning techniques and explanation methods has been developed.Our framework incorporates a wide range of features and determinants,allowing for a comprehensive and nuanced understanding of pre-frailty risk.
基金Supported by the Shanghai Municipal Health Commission Project,No.20214Y0284.
文摘BACKGROUND Preoperative risk assessments are vital for identifying patients at high risk of postoperative mortality.However,traditional scoring systems can be time consuming.We hypothesized that the use of machine learning models would enable rapid and accurate risk assessments to be performed.AIM To assess the potential of machine learning algorithms to develop predictive models of mortality risk after abdominal surgery.METHODS This retrospective study included 230 individuals who underwent abdominal surgery at the Seventh People’s Hospital of Shanghai University of Traditional Chinese Medicine between January 2023 and December 2023.Demographic and surgery-related data were collected and used to develop nomogram,decision-tree,random-forest,gradient-boosting,support vector machine,and naïve Bayesian models to predict 30-day mortality risk after abdominal surgery.Models were assessed using receiver operating characteristic curves and compared using the DeLong test.RESULTS Of the 230 included patients,52 died and 178 survived.Models were developed using the training cohort(n=161)and assessed using the validation cohort(n=68).The areas under the receiver operating characteristic curves for the nomogram,decision-tree,random-forest,gradient-boosting tree,support vector machine,and naïve Bayesian models were 0.908[95%confidence interval(CI):0.824-0.992],0.874(95%CI:0.785-0.963),0.928(95%CI:0.869-0.987),0.907(95%CI:0.837-0.976),0.983(95%CI:0.959-1.000),and 0.807(95%CI:0.702-0.911),respectively.CONCLUSION Nomogram,random-forest,gradient-boosting tree,and support vector machine models all demonstrate strong performances for the prediction of postoperative mortality and can be selected based on the clinical circumstances.
基金supported by the Centre for Advanced Modelling and Geospatial Information Systems(CAMGIS),University of Technology Sydney(UTS),Australia and was supported by the Research Training Program(RTP)of the Australian Government.
文摘Hurricanes are one of the most destructive natural disasters that can cause catastrophic losses to both communities and infrastructure.Assessment of hurricane risk furnishes a spatial depiction of the interplay among hazard,vulnerability,exposure,and mitigation capacity,crucial for understanding and managing the risks hurricanes pose to communities.These assessments aid in gauging the efficacy of existing hurricane mitigation strategies and gauging their resilience across diverse climate change scenarios.A systematic review was conducted,encompassing 94 articles,to scrutinize the structure,data inputs,assumptions,methodologies,perils modelled,and key predictors of hurricane risk.This review identified key research gaps essential for enhancing future risk assessments.The complex interaction between hurricane perils may be disastrous and underestimated in the majority of risk assessments which focus on a single peril,commonly storm surge and flood,resulting in inadequacies in disaster resilience planning.Most risk assessments were based on hurricane frequency rather than hurricane damage,which is more insightful for policymakers.Furthermore,considering secondary indirect impacts stemming from hurricanes,including real estate market and business interruption,could enrich economic impact assessments.Hurricane mitigation measures were the most under-utilised category of predictors leveraged in only 5%of studies.The top six predictive factors for hurricane risk were land use,slope,precipitation,elevation,population density,and soil texture/drainage.Another notable research gap identified was the potential of machine learning techniques in risk assessments,offering advantages over traditional MCDM and numerical models due to their ability to capture complex nonlinear relationships and adaptability to different study regions.Existing machine learning based risk assessments leverage random forest models(42%of studies)followed by neural network models(19%of studies),with further research required to investigate diverse machine learning algorithms such as ensemble models.A further research gap is model validation,in particular assessing transferability to a new study region.Additionally,harnessing simulated data and refining projections related to demographic and built environment dynamics can bolster the sophistication of climate change scenario assessments.By addressing these research gaps,hurricane risk assessments can furnish invaluable insights for national policymakers,facilitating the development of robust hurricane mitigation strategies and the construction of hurricane-resilient communities.To the authors’knowledge,this represents the first literature review specifically dedicated to quantitative hurricane risk assessments,encompassing a comparison of Multi-criteria Decision Making(MCDM),numerical models,and machine learning models.Ultimately,advancements in hurricane risk assessments and modelling stand poised to mitigate potential losses to communities and infrastructure both in the immediate and long-term future.
基金supported by the National Natural Science Foundation of China under Grant No.62502040the ZTE Industry-University-Institute Cooperation Funds under Grant No.IA20230700001.
文摘We propose Mixed-Precision Multibranch Network(M+MNet)to compensate for the neglect of background information in image aesthetics assessment(IAA)while providing strategies for overcoming the dilemma between training costs and performance.First,two exponentially weighted pooling methods are used to selectively boost the extraction of background and salient information during downsampling.Second,we propose Corner Grid,an unsupervised data augmentation method that leverages the diffusive characteristics of convolution to force the network to seek more relevant background information.Third,we perform mixed-precision training by switching the precision format,thus significantly reducing the time and memory consumption of data representation and transmission.Most of our methods specifically designed for IAA tasks have demonstrated generalizability to other IAA works.For performance verification,we develop a large-scale benchmark(the most comprehensive thus far)by comparing 17 methods with M+MNet on two representative datasets:the Aesthetic Visual Analysis(AVA)dataset and FLICKR-Aesthetic Evaluation Subset(FLICKR-AES).M+MNet achieves state-of-the-art performance on all tasks.
基金supported by the project of National Natural Science Foundation of China(Grant No.42371203 and U21A2032)the project Financial Fund of Sichuan Institute of Geological Survey(SCIGSCZDXM-2024008)+1 种基金Sichuan Provincial Science and Technology Department Program Funding(No.2025YFHZ0010)Science and Technology Program of Aba City(NO.R24YYJSYJ0001)。
文摘The 2019 Typhoon Lekima triggered extensive landslides in Zhejiang Province.To explore the impact of typhoon paths on the distribution of landslide susceptibility,this study proposes a spatiotemporal zoning assessment framework based on typhoon paths and inner rainbands.According to the typhoon landing path and its rainfall impact range,the study area is divided into the typhoon event period(TEP)and the annual non-typhoon period(ANP).The model uses 14 environmental factors,with the only difference between TEP and ANP being the rainfall index:TEP uses 48-hour rainfall during the typhoon,while ANP uses multi-year average annual rainfall.Modeling and comparative analysis were conducted using six machine learning models including random forest(RF)and support vector machine(SVM).The results show that the distribution pattern of high-risk landslide areas during TEP is significantly correlated with typhoon intensity:when the intensity is level 12,high-risk areas are radially distributed;at levels 10-11,they tend to concentrate asymmetrically along the coast;and when the intensity drops to below level 9,the overall susceptibility decreases significantly.During ANP,the distribution of landslides is relatively uniform with no obvious spatial concentration.Analysis on the factor contribution rate indicates that the rainfall weight in TEP is as high as 32.1%,making it the dominant factor;in ANP,the rainfall weight drops to 13.6%while the influence of factors such as slope and topographic wetness index increases,revealing differences in landslide formation mechanisms between the two periods.This study demonstrates that the spatiotemporal zoning method based on typhoon paths can effectively characterize the spatial susceptibility patterns of landslides and improve disaster identification capabilities under extreme weather conditions.The finally generated annual susceptibility zoning map divides the study area into four types of risk regions,providing a reference for dynamic monitoring and differentiated risk management of landslides in typhoon-prone areas.
