Objective and accurate evaluation of rock mass quality classification is the prerequisite for reliable sta-bility assessment.To develop a tool that can deliver quick and accurate evaluation of rock mass quality,a deep...Objective and accurate evaluation of rock mass quality classification is the prerequisite for reliable sta-bility assessment.To develop a tool that can deliver quick and accurate evaluation of rock mass quality,a deep learning approach is developed,which uses stacked autoencoders(SAEs)with several autoencoders and a softmax net layer.Ten rock parameters of rock mass rating(RMR)system are calibrated in this model.The model is trained using 75%of the total database for training sample data.The SAEs trained model achieves a nearly 100%prediction accuracy.For comparison,other different models are also trained with the same dataset,using artificial neural network(ANN)and radial basis function(RBF).The results show that the SAEs classify all test samples correctly while the rating accuracies of ANN and RBF are 97.5%and 98.7%,repectively,which are calculated from the confusion matrix.Moreover,this model is further employed to predict the slope risk level of an abandoned quarry.The proposed approach using SAEs,or deep learning in general,is more objective and more accurate and requires less human inter-vention.The findings presented here shall shed light for engineers/researchers interested in analyzing rock mass classification criteria or performing field investigation.展开更多
A common difficulty in building prediction models with real-world environmental datasets is the skewed distribution of classes.There are significantly more samples for day-to-day classes,while rare events such as poll...A common difficulty in building prediction models with real-world environmental datasets is the skewed distribution of classes.There are significantly more samples for day-to-day classes,while rare events such as polluted classes are uncommon.Consequently,the limited availability of minority outcomes lowers the classifier’s overall reliability.This study assesses the capability of machine learning(ML)algorithms in tackling imbalanced water quality data based on the metrics of precision,recall,and F1 score.It intends to balance the misled accuracy towards the majority of data.Hence,10 ML algorithms of its performance are compared.The classifiers included are AdaBoost,SupportVector Machine,Linear Discriminant Analysis,k-Nearest Neighbors,Naive Bayes,Decision Trees,Random Forest,Extra Trees,Bagging,and the Multilayer Perceptron.This study also uses the Easy Ensemble Classifier,Balanced Bagging,andRUSBoost algorithm to evaluatemulti-class imbalanced learning methods.The comparison results revealed that a highaccuracy machine learning model is not always good in recall and sensitivity.This paper’s stacked ensemble deep learning(SE-DL)generalization model effectively classifies the water quality index(WQI)based on 23 input variables.The proposed algorithm achieved a remarkable average of 95.69%,94.96%,92.92%,and 93.88%for accuracy,precision,recall,and F1 score,respectively.In addition,the proposed model is compared against two state-of-the-art classifiers,the XGBoost(eXtreme Gradient Boosting)and Light Gradient Boosting Machine,where performance metrics of balanced accuracy and g-mean are included.The experimental setup concluded XGBoost with a higher balanced accuracy and G-mean.However,the SE-DL model has a better and more balanced performance in the F1 score.The SE-DL model aligns with the goal of this study to ensure the balance between accuracy and completeness for each water quality class.The proposed algorithm is also capable of higher efficiency at a lower computational time against using the standard SyntheticMinority Oversampling Technique(SMOTE)approach to imbalanced datasets.展开更多
Purpose:To assess the clinical efficacy of integrating deep learning reconstruction(DLR)with contrast-enhancement-boost(CE-boost)in 80 kVp head and neck CT angiography(CTA)using substantially lowered radiation and con...Purpose:To assess the clinical efficacy of integrating deep learning reconstruction(DLR)with contrast-enhancement-boost(CE-boost)in 80 kVp head and neck CT angiography(CTA)using substantially lowered radiation and contrast medium(CM)doses,compared to the standard 100 kVp protocol using hybrid iterative reconstruction(HIR).Methods:Sixty-six patients were prospectively enrolled and randomly assigned to one of two groups:the low-dose group(n=33),receiving 80 kVp and 28 mL contrast medium(CM)with a noise index(NI)of 15;and the regular-dose group(n=33),receiving 100 kVp and 40 mL CM with an NI of 10.For the lowdose group,images underwent reconstruction using both hybrid iterative reconstruction(HIR)and deep learning reconstruction(DLR)at mild-,standard-,and strong-strength levels,both before and after combination with contrast enhancement-boost(CE-boost).This generated eight distinct datasets:L-HIR,L-DLR_(mild),L-DLR_(standard),L-DLR_(strong),L-HIR-CE,L-DLR_(mild)-CE,L-DLR_(standard)-CE,and L-DLR_(strong)-CE.Images for the regular-dose group were reconstructed solely with HIR(R-HIR).Quantitative analysis involved calculating and comparing CT attenuation,image noise,signal-to-noise ratio(SNR),and contrast-to-noise ratio(CNR)within six key vessels:the aortic arch(AA),internal carotid artery(ICA),external carotid artery(ECA),vertebral arteries(VA),basilar artery(BA),and middle cerebral artery(MCA).Two radiologists independently assessed subjective image quality using a 5-point scale,with statistical significance defined as P<0.05.Results:Compared to the regular-dose group,the low-dose protocol achieved a substantial reduction in contrast media volume(28 mL versus 40 mL,a 30%decrease)and radiation exposure((0.41±0.08)mSv versus(1.18±0.12)mSv,a 65%reduction).Both L-DLR_(standard) and L-DLR_(strong) delivered comparable or superior SNR and CNR across all vascular segments relative to R-HIR.However,subjective image quality scores for L-DLR at all strength levels fell below those for R-HIR(all P<0.05 for both readers).Combining CE-boost with the low-dose protocol significantly enhanced the objective image performance of L-DLR_(strong)-CE(all P<0.05)and produced subjective image scores comparable to R-HIR(reader 1:P=0.15;reader 2:P=0.06).Conclusion:When compared to the standard 100 kVp head and neck CTA,the combination of the DLR and CE-boost techniques at 80 kVp can achieve a 30%reduction in contrast dose and a 65%reduction in radiation dose,while maintaining both objective and subjective image quality.展开更多
Software defect prediction plays a very important role in software quality assurance,which aims to inspect as many potentially defect-prone software modules as possible.However,the performance of the prediction model ...Software defect prediction plays a very important role in software quality assurance,which aims to inspect as many potentially defect-prone software modules as possible.However,the performance of the prediction model is susceptible to high dimensionality of the dataset that contains irrelevant and redundant features.In addition,software metrics for software defect prediction are almost entirely traditional features compared to the deep semantic feature representation from deep learning techniques.To address these two issues,we propose the following two solutions in this paper:(1)We leverage a novel non-linear manifold learning method-SOINN Landmark Isomap(SL-Isomap)to extract the representative features by selecting automatically the reasonable number and position of landmarks,which can reveal the complex intrinsic structure hidden behind the defect data.(2)We propose a novel defect prediction model named DLDD based on hybrid deep learning techniques,which leverages denoising autoencoder to learn true input features that are not contaminated by noise,and utilizes deep neural network to learn the abstract deep semantic features.We combine the squared error loss function of denoising autoencoder with the cross entropy loss function of deep neural network to achieve the best prediction performance by adjusting a hyperparameter.We compare the SL-Isomap with seven state-of-the-art feature extraction methods and compare the DLDD model with six baseline models across 20 open source software projects.The experimental results verify that the superiority of SL-Isomap and DLDD on four evaluation indicators.展开更多
In recent years,the rapid development of computer software has led to numerous security problems,particularly software vulnerabilities.These flaws can cause significant harm to users’privacy and property.Current secu...In recent years,the rapid development of computer software has led to numerous security problems,particularly software vulnerabilities.These flaws can cause significant harm to users’privacy and property.Current security defect detection technology relies on manual or professional reasoning,leading to missed detection and high false detection rates.Artificial intelligence technology has led to the development of neural network models based on machine learning or deep learning to intelligently mine holes,reducing missed alarms and false alarms.So,this project aims to study Java source code defect detection methods for defects like null pointer reference exception,XSS(Transform),and Structured Query Language(SQL)injection.Also,the project uses open-source Javalang to translate the Java source code,conducts a deep search on the AST to obtain the empty syntax feature library,and converts the Java source code into a dependency graph.The feature vector is then used as the learning target for the neural network.Four types of Convolutional Neural Networks(CNN),Long Short-Term Memory(LSTM),Bi-directional Long Short-Term Memory(BiLSTM),and Attention Mechanism+Bidirectional LSTM,are used to investigate various code defects,including blank pointer reference exception,XSS,and SQL injection defects.Experimental results show that the attention mechanism in two-dimensional BLSTM is the most effective for object recognition,verifying the correctness of the method.展开更多
Soil is the major source of infinite lives on Earth and the quality of soil plays significant role on Agriculture practices all around.Hence,the evaluation of soil quality is very important for determining the amount ...Soil is the major source of infinite lives on Earth and the quality of soil plays significant role on Agriculture practices all around.Hence,the evaluation of soil quality is very important for determining the amount of nutrients that the soil require for proper yield.In present decade,the application of deep learning models in many fields of research has created greater impact.The increasing soil data availability of soil data there is a greater demand for the remotely avail open source model,leads to the incorporation of deep learning method to predict the soil quality.With that concern,this paper proposes a novel model called Improved Soil Quality Prediction Model using Deep Learning(ISQP-DL).The work considers the chemical,physical and biological factors of soil in particular area to estimate the soil quality.Firstly,pH rating of soil samples has been collected from the soil testing laboratory from which the acidic range has been categorized through soil test and the same data has been taken as input to the Deep Neural Network Regression(DNNR)model.Secondly,soil nutrient data has been given as second input to the DNNR model.By utilizing this data set,the DNNR method is used to evaluate the fertility rate by which the soil quality has been estimated.For training and testing,the model uses Deep Neural Network Regression(DNNR),by utilizing the dataset.The results show that the proposed model is effective for SQP(Soil Quality Prediction Model)with efficient good fitting and generality is enhanced with input features with higher rate of classification accuracy.The results show that the proposed model achieves 96.7%of accuracy rate compared with existing models.展开更多
Recommending high-quality news to users is vital in improving user stickiness and news platforms’reputation.However,existing news quality evaluation methods,such as clickbait detection and popularity prediction,are c...Recommending high-quality news to users is vital in improving user stickiness and news platforms’reputation.However,existing news quality evaluation methods,such as clickbait detection and popularity prediction,are challenging to reflect news quality comprehensively and concisely.This paper defines news quality as the ability of news articles to elicit clicks and comments from users,which represents whether the news article can attract widespread attention and discussion.Based on the above definition,this paper first presents a straightforward method to measure news quality based on the comments and clicks of news and defines four news quality indicators.Then,the dataset can be labeled automatically by the method.Next,this paper proposes a deep learning model that integrates explicit and implicit news information for news quality evaluation(EINQ).The explicit information includes the headline,source,and publishing time of the news,which attracts users to click.The implicit information refers to the news article’s content which attracts users to comment.The implicit and explicit information affect users’click and comment behavior differently.For modeling explicit information,the typical convolution neural network(CNN)is used to get news headline semantic representation.For modeling implicit information,a hierarchical attention network(HAN)is exploited to extract news content semantic representation while using the latent Dirichlet allocation(LDA)model to get the subject distribution of news as a semantic supplement.Considering the different roles of explicit and implicit information for quality evaluation,the EINQ exploits an attention layer to fuse them dynamically.The proposed model yields the Accuracy of 82.31%and the F-Score of 80.51%on the real-world dataset from Toutiao,which shows the effectiveness of explicit and implicit information dynamic fusion and demonstrates performance improvements over a variety of baseline models in news quality evaluation.This work provides empirical evidence for explicit and implicit factors in news quality evaluation and a new idea for news quality evaluation.展开更多
Over the last couple of decades,community question-answering sites(CQAs)have been a topic of much academic interest.Scholars have often leveraged traditional machine learning(ML)and deep learning(DL)to explore the eve...Over the last couple of decades,community question-answering sites(CQAs)have been a topic of much academic interest.Scholars have often leveraged traditional machine learning(ML)and deep learning(DL)to explore the ever-growing volume of content that CQAs engender.To clarify the current state of the CQA literature that has used ML and DL,this paper reports a systematic literature review.The goal is to summarise and synthesise the major themes of CQA research related to(i)questions,(ii)answers and(iii)users.The final review included 133 articles.Dominant research themes include question quality,answer quality,and expert identification.In terms of dataset,some of the most widely studied platforms include Yahoo!Answers,Stack Exchange and Stack Overflow.The scope of most articles was confined to just one platform with few cross-platform investigations.Articles with ML outnumber those with DL.Nonetheless,the use of DL in CQA research is on an upward trajectory.A number of research directions are proposed.展开更多
Concrete exterior quality is one of the important metrics in evaluating construction project quality.Among the defects affecting concrete exterior quality,bughole is one of the most common imperfections,thus detecting...Concrete exterior quality is one of the important metrics in evaluating construction project quality.Among the defects affecting concrete exterior quality,bughole is one of the most common imperfections,thus detecting concrete bughole accurately is significant for improving concrete exterior quality and consequently the quality of the whole project.This paper presents a deep learning-based method for detecting concrete surface bugholes in a more objective and automatic way.The bugholes are identified in concrete surface images by Mask R-CNN.An evaluation metric is developed to indicate the scale of concrete bughole.The proposed approach can detect bugholes in an instance level automatically and output the mask of each bughole,based on which the bughole area ratio is automatically calculated and the quality grade of the concrete surfaces is assessed.For demonstration,a total of 273 raw concrete surface images taken by mobile phone cameras are collected as a dataset.The test results show that the average precision(AP)of bughole masks is 90.8%.展开更多
The paradigm shift towards the Internet of Things(IoT)phe-nomenon and the rise of edge-computing models provide massive poten-tial for several upcoming IoT applications like smart grid,smart energy,smart home,smart he...The paradigm shift towards the Internet of Things(IoT)phe-nomenon and the rise of edge-computing models provide massive poten-tial for several upcoming IoT applications like smart grid,smart energy,smart home,smart health and smart transportation services.However,it also provides a sequence of novel cyber-security issues.Although IoT networks provide several advantages,the heterogeneous nature of the network and the wide connectivity of the devices make the network easy for cyber-attackers.Cyberattacks result in financial loss and data breaches for organizations and individuals.So,it becomes crucial to secure the IoT environment from such cyberattacks.With this motivation,the current study introduces an effectual Enhanced Crow Search Algorithm with Deep Learning-Driven Cyberattack Detection(ECSADL-CAD)model for the Software-Defined Networking(SDN)-enabled IoT environment.The presented ECSADL-CAD approach aims to identify and classify the cyberattacks in the SDN-enabled IoT envi-ronment.To attain this,the ECSADL-CAD model initially pre-processes the data.In the presented ECSADL-CAD model,the Reinforced Deep Belief Network(RDBN)model is employed for attack detection.At last,the ECSA-based hyperparameter tuning process gets executed to boost the overall classification outcomes.A series of simulations were conducted to validate the improved outcomes of the proposed ECSADL-CAD model.The experimental outcomes confirmed the superiority of the proposed ECSADL-CAD model over other existing methodologies.展开更多
Combined sewer overflows represent significant risks to human health as untreated water is discharged to the environment.Municipalities,such as the Metropolitan Sewer District of Greater Cincinnati(MSDGC),recently beg...Combined sewer overflows represent significant risks to human health as untreated water is discharged to the environment.Municipalities,such as the Metropolitan Sewer District of Greater Cincinnati(MSDGC),recently began collecting large amounts of water-related data and considering the adoption of deep learning(DL)solutions like recurrent neural network(RNN)for predicting overflow events.Clearly,assessing the DL's fitness for the purpose requires a systematic understanding of the problem context.In this study,we propose a requirements engineering framework that uses the problem frames to identify and structure the stakeholder concerns,analyses the physical situations in which the highquality data assumptions may not hold,and derives the software testing criteria in the form of metamorphic relations that incorporate both input transformations and output comparisons.Applying our framework to MSDGC's overflow prediction problem enables a principled way to evaluate different RNN solutions in meeting the requirements.展开更多
Distributed denial-of-service(DDoS)attacks are designed to interrupt network services such as email servers and webpages in traditional computer networks.Furthermore,the enormous number of connected devices makes it d...Distributed denial-of-service(DDoS)attacks are designed to interrupt network services such as email servers and webpages in traditional computer networks.Furthermore,the enormous number of connected devices makes it difficult to operate such a network effectively.Software defined networks(SDN)are networks that are managed through a centralized control system,according to researchers.This controller is the brain of any SDN,composing the forwarding table of all data plane network switches.Despite the advantages of SDN controllers,DDoS attacks are easier to perpetrate than on traditional networks.Because the controller is a single point of failure,if it fails,the entire network will fail.This paper offers a Hybrid Deep Learning Intrusion Detection and Prevention(HDLIDP)framework,which blends signature-based and deep learning neural networks to detect and prevent intrusions.This framework improves detection accuracy while addressing all of the aforementioned problems.To validate the framework,experiments are done on both traditional and SDN datasets;the findings demonstrate a significant improvement in classification accuracy.展开更多
Recently, many open source software (OSS) developed by various OSS projects. Also, the reliability assessment methods of OSS have been proposed by several researchers. Many methods for software reliability assessment ...Recently, many open source software (OSS) developed by various OSS projects. Also, the reliability assessment methods of OSS have been proposed by several researchers. Many methods for software reliability assessment have been proposed by software reliability growth models. Moreover, our research group has been proposed the method of reliability assessment for the OSS. Many OSS use bug tracking system (BTS) to manage software faults after it released. It keeps a detailed record of the environment in terms of the faults. There are several methods of reliability assessment based on deep learning for OSS fault data in the past. On the other hand, the data registered in BTS differences depending on OSS projects. Also, some projects have the specific collection data. The BTS has the specific collection data for each project. We focus on the recorded data. Moreover, we investigate the difference between the general data and the specific one for the estimation of OSS reliability. As a result, we show that the reliability estimation results by using specific data are better than the method using general data. Then, we show the characteristics between the specified data and general one in this paper. We also develop the GUI-based software to perform these reliability analyses so that even those who are not familiar with deep learning implementations can perform reliability analyses of OSS.展开更多
Modern manufacturing processes have become more reliant on automation because of the accelerated transition from Industry 3.0 to Industry 4.0.Manual inspection of products on assembly lines remains inefficient,prone t...Modern manufacturing processes have become more reliant on automation because of the accelerated transition from Industry 3.0 to Industry 4.0.Manual inspection of products on assembly lines remains inefficient,prone to errors and lacks consistency,emphasizing the need for a reliable and automated inspection system.Leveraging both object detection and image segmentation approaches,this research proposes a vision-based solution for the detection of various kinds of tools in the toolkit using deep learning(DL)models.Two Intel RealSense D455f depth cameras were arranged in a top down configuration to capture both RGB and depth images of the toolkits.After applying multiple constraints and enhancing them through preprocessing and augmentation,a dataset consisting of 3300 annotated RGB-D photos was generated.Several DL models were selected through a comprehensive assessment of mean Average Precision(mAP),precision-recall equilibrium,inference latency(target≥30 FPS),and computational burden,resulting in a preference for YOLO and Region-based Convolutional Neural Networks(R-CNN)variants over ViT-based models due to the latter’s increased latency and resource requirements.YOLOV5,YOLOV8,YOLOV11,Faster R-CNN,and Mask R-CNN were trained on the annotated dataset and evaluated using key performance metrics(Recall,Accuracy,F1-score,and Precision).YOLOV11 demonstrated balanced excellence with 93.0%precision,89.9%recall,and a 90.6%F1-score in object detection,as well as 96.9%precision,95.3%recall,and a 96.5%F1-score in instance segmentation with an average inference time of 25 ms per frame(≈40 FPS),demonstrating real-time performance.Leveraging these results,a YOLOV11-based windows application was successfully deployed in a real-time assembly line environment,where it accurately processed live video streams to detect and segment tools within toolkits,demonstrating its practical effectiveness in industrial automation.The application is capable of precisely measuring socket dimensions by utilising edge detection techniques on YOLOv11 segmentation masks,in addition to detection and segmentation.This makes it possible to do specification-level quality control right on the assembly line,which improves the ability to examine things in real time.The implementation is a big step forward for intelligent manufacturing in the Industry 4.0 paradigm.It provides a scalable,efficient,and accurate way to do automated inspection and dimensional verification activities.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
Data-driven approaches that make timely predictions about pollutant concentrations in the effluent of constructed wetlands are essential for improving the treatment performance of constructed wetlands.However,the effe...Data-driven approaches that make timely predictions about pollutant concentrations in the effluent of constructed wetlands are essential for improving the treatment performance of constructed wetlands.However,the effect of the meteorological condition and flow changes in a real scenario are generally neglected in water quality prediction.To address this problem,in this study,we propose an approach based on multi-source data fusion that considers the following indicators:water quality indicators,water quantity indicators,and meteorological indicators.In this study,we establish four representative methods to simultaneously predict the concentrations of three representative pollutants in the effluent of a practical large-scale constructed wetland:(1)multiple linear regression;(2)backpropagation neural network(BPNN);(3)genetic algorithm combined with the BPNN to solve the local minima problem;and(4)long short-term memory(LSTM)neural network to consider the influence of past results on the present.The results suggest that the LSTM-predicting model performed considerably better than the other deep neural network-based model or linear method,with a satisfactory R^(2).Additionally,given the huge fluctuation of different pollutant concentrations in the effluent,we used a moving average method to smooth the original data,which successfully improved the accuracy of traditional neural networks and hybrid neural networks.The results of this study indicate that the hybrid modeling concept that combines intelligent and scientific data preprocessing methods with deep learning algorithms is a feasible approach for forecasting water quality in the effluent of actual engineering.展开更多
We investigate a problem of object-oriented (OO) software quality estimation from a multi-instance (MI) perspective. In detail,each set of classes that have an inheritance relation,named 'class hierarchy',is r...We investigate a problem of object-oriented (OO) software quality estimation from a multi-instance (MI) perspective. In detail,each set of classes that have an inheritance relation,named 'class hierarchy',is regarded as a bag,while each class in the set is regarded as an instance. The learning task in this study is to estimate the label of unseen bags,i.e.,the fault-proneness of untested class hierarchies. A fault-prone class hierarchy contains at least one fault-prone (negative) class,while a non-fault-prone (positive) one has no negative class. Based on the modification records (MRs) of the previous project releases and OO software metrics,the fault-proneness of an untested class hierarchy can be predicted. Several selected MI learning algorithms were evalu-ated on five datasets collected from an industrial software project. Among the MI learning algorithms investigated in the ex-periments,the kernel method using a dedicated MI-kernel was better than the others in accurately and correctly predicting the fault-proneness of the class hierarchies. In addition,when compared to a supervised support vector machine (SVM) algorithm,the MI-kernel method still had a competitive performance with much less cost.展开更多
The built environment sector is responsible for almost one-third of the world’s final energy consumption. Hence, seeking plausible solutions to minimise building energy demands and mitigate adverse environmental impa...The built environment sector is responsible for almost one-third of the world’s final energy consumption. Hence, seeking plausible solutions to minimise building energy demands and mitigate adverse environmental impacts is necessary. Artificial intelligence (AI) techniques such as machine and deep learning have been increasingly and successfully applied to develop solutions for the built environment. This review provided a critical summary of the existing literature on the machine and deep learning methods for the built environment over the past decade, with special reference to holistic approaches. Different AI-based techniques employed to resolve interconnected problems related to heating, ventilation and air conditioning (HVAC) systems and enhance building performances were reviewed, including energy forecasting and management, indoor air quality and occupancy comfort/satisfaction prediction, occupancy detection and recognition, and fault detection and diagnosis. The present study explored existing AI-based techniques focusing on the framework, methodology, and performance. The literature highlighted that selecting the most suitable machine learning and deep learning model for solving a problem could be challenging. The recent explosive growth experienced by the research area has led to hundreds of machine learning algorithms being applied to building performance-related studies. The literature showed that existing research studies considered a wide range of scope/scales (from an HVAC component to urban areas) and time scales (minute to year). This makes it difficult to find an optimal algorithm for a specific task or case. The studies also employed a wide range of evaluation metrics, adding to the challenge. Further developments and more specific guidelines are required for the built environment field to encourage best practices in evaluating and selecting models. The literature also showed that while machine and deep learning had been successfully applied in building energy efficiency research, most of the studies are still at the experimental or testing stage, and there are limited studies which implemented machine and deep learning strategies in actual buildings and conducted the post-occupancy evaluation.展开更多
Air quality estimation assesses the pollution level in the air,supports public health warnings,and is a valuable tool in environmental management.Although air sensors have proven helpful in this task,sensors are often...Air quality estimation assesses the pollution level in the air,supports public health warnings,and is a valuable tool in environmental management.Although air sensors have proven helpful in this task,sensors are often expensive and difficult to install,while cameras are becoming more popular and accessible,from which images can be collected as data for deep learning models to solve the above task.This leads to another problem:several labeled images are needed to achieve high accuracy when deep-learningmodels predict air quality.In this research,we have threemain contributions:(1)Collect and publish an air quality estimation dataset,namely PTIT_AQED,including environmental image data and air quality;(2)Propose a deep learning model to predict air quality with few data,called PTIT_FAQE(PTIT Few-shot air quality estimation).We build PTIT_FAQE based on EfficientNet-a CNN architecture that ensures high performance in deep learning applications and Few-shot Learning with Prototypical Networks.This helps the model use only a fewtraining data but still achieve high accuracy in air quality estimation.And(3)conduct experiments to prove the superiority of PTIT_FAQE compared to other studies on both PTIT_AQED and APIN datasets.The results show that our model achieves an accuracy of 0.9278 and an F1-Score of 0.9139 on the PTIT_AQED dataset and an accuracy of 0.9467 and an F1-Score of 0.9371 on the APIN dataset,which demonstrate a significant performance improvement compared to previous studies.We also conduct detailed experiments to evaluate the impact of each component on model performance.展开更多
Understanding the characteristics of O_(3)precursor contributions over multiple years is crucial for designing effective O_(3)control strategies over the Pearl River Delta(PRD)region of China.In this study,a deep lear...Understanding the characteristics of O_(3)precursor contributions over multiple years is crucial for designing effective O_(3)control strategies over the Pearl River Delta(PRD)region of China.In this study,a deep learning-based response surface model(DeepRSM)was developed and applied over the PRD(DeepRSM-PRD)to identify and quantify the main features of O_(3)regimes and regional contributions in the core PRD over multiple years(2019–2021).The Out-of-Sample(OOS)validation results indicated that DeepRSM-PRD effectively predicted the nonlinear response of O_(3)to emission controls,maintaining validity across non-training periods.Our study revealed that O_(3)generation was sensitive to volatile organic compounds(VOC)in the core PRD in 2019,with nitrogen oxides(NO_(x))-limited regimes emerging in most major cities in 2020 and 2021.Further investigation into source contributions showed that in our model domain,O_(3)formation in central cities of the PRD was primarily driven by local contributions and was susceptible to influence from nearby cities.With small emission reductions,VOC contributions predominantly drive O_(3)production in Guangzhou and Shenzhen.However,NO_(x)emissions were identified as the primary contributors in all central city receptors when anthropogenic emissions were removed,sharing 59.5%–69.3%in 2019,64.4%–72.3%in 2020,and 62.75%–73.2%in 2021.Our results highlight the need for a high focus on NO_(x)emissions control in the core PRD.In addition,for Guangzhou and Shenzhen,VOC reduction also plays a crucial role in the initial stages of modest emission reductions.展开更多
基金supported by the National Natural Science Foundation of China(Grant Nos.51979253,51879245)the Fundamental Research Funds for the Central Universities,China University of Geosciences(Wuhan)(Grant No.CUGCJ1821).
文摘Objective and accurate evaluation of rock mass quality classification is the prerequisite for reliable sta-bility assessment.To develop a tool that can deliver quick and accurate evaluation of rock mass quality,a deep learning approach is developed,which uses stacked autoencoders(SAEs)with several autoencoders and a softmax net layer.Ten rock parameters of rock mass rating(RMR)system are calibrated in this model.The model is trained using 75%of the total database for training sample data.The SAEs trained model achieves a nearly 100%prediction accuracy.For comparison,other different models are also trained with the same dataset,using artificial neural network(ANN)and radial basis function(RBF).The results show that the SAEs classify all test samples correctly while the rating accuracies of ANN and RBF are 97.5%and 98.7%,repectively,which are calculated from the confusion matrix.Moreover,this model is further employed to predict the slope risk level of an abandoned quarry.The proposed approach using SAEs,or deep learning in general,is more objective and more accurate and requires less human inter-vention.The findings presented here shall shed light for engineers/researchers interested in analyzing rock mass classification criteria or performing field investigation.
基金primarily supported by the Ministry of Higher Education through MRUN Young Researchers Grant Scheme(MY-RGS),MR001-2019,entitled“Climate Change Mitigation:Artificial Intelligence-Based Integrated Environmental System for Mangrove Forest Conservation,”received by K.H.,S.A.R.,H.F.H.,M.I.M.,and M.M.Asecondarily funded by the UM-RU Grant,ST065-2021,entitled Climate Smart Mitigation and Adaptation:Integrated Climate Resilience Strategy for Tropical Marine Ecosystem.
文摘A common difficulty in building prediction models with real-world environmental datasets is the skewed distribution of classes.There are significantly more samples for day-to-day classes,while rare events such as polluted classes are uncommon.Consequently,the limited availability of minority outcomes lowers the classifier’s overall reliability.This study assesses the capability of machine learning(ML)algorithms in tackling imbalanced water quality data based on the metrics of precision,recall,and F1 score.It intends to balance the misled accuracy towards the majority of data.Hence,10 ML algorithms of its performance are compared.The classifiers included are AdaBoost,SupportVector Machine,Linear Discriminant Analysis,k-Nearest Neighbors,Naive Bayes,Decision Trees,Random Forest,Extra Trees,Bagging,and the Multilayer Perceptron.This study also uses the Easy Ensemble Classifier,Balanced Bagging,andRUSBoost algorithm to evaluatemulti-class imbalanced learning methods.The comparison results revealed that a highaccuracy machine learning model is not always good in recall and sensitivity.This paper’s stacked ensemble deep learning(SE-DL)generalization model effectively classifies the water quality index(WQI)based on 23 input variables.The proposed algorithm achieved a remarkable average of 95.69%,94.96%,92.92%,and 93.88%for accuracy,precision,recall,and F1 score,respectively.In addition,the proposed model is compared against two state-of-the-art classifiers,the XGBoost(eXtreme Gradient Boosting)and Light Gradient Boosting Machine,where performance metrics of balanced accuracy and g-mean are included.The experimental setup concluded XGBoost with a higher balanced accuracy and G-mean.However,the SE-DL model has a better and more balanced performance in the F1 score.The SE-DL model aligns with the goal of this study to ensure the balance between accuracy and completeness for each water quality class.The proposed algorithm is also capable of higher efficiency at a lower computational time against using the standard SyntheticMinority Oversampling Technique(SMOTE)approach to imbalanced datasets.
基金National Natural Science Foundation of China(82001814)National High Level Hospital Clinical Research Funding(grant number 2022-PUMCH-B-067)+1 种基金National High Level Hospital Clinical Research Funding(grant number 2022-PUMCH-B-068)2021 SKY Imaging Research Fund of the Chinese Internatinal Medical Foundation(Z-2014-07-2101).
文摘Purpose:To assess the clinical efficacy of integrating deep learning reconstruction(DLR)with contrast-enhancement-boost(CE-boost)in 80 kVp head and neck CT angiography(CTA)using substantially lowered radiation and contrast medium(CM)doses,compared to the standard 100 kVp protocol using hybrid iterative reconstruction(HIR).Methods:Sixty-six patients were prospectively enrolled and randomly assigned to one of two groups:the low-dose group(n=33),receiving 80 kVp and 28 mL contrast medium(CM)with a noise index(NI)of 15;and the regular-dose group(n=33),receiving 100 kVp and 40 mL CM with an NI of 10.For the lowdose group,images underwent reconstruction using both hybrid iterative reconstruction(HIR)and deep learning reconstruction(DLR)at mild-,standard-,and strong-strength levels,both before and after combination with contrast enhancement-boost(CE-boost).This generated eight distinct datasets:L-HIR,L-DLR_(mild),L-DLR_(standard),L-DLR_(strong),L-HIR-CE,L-DLR_(mild)-CE,L-DLR_(standard)-CE,and L-DLR_(strong)-CE.Images for the regular-dose group were reconstructed solely with HIR(R-HIR).Quantitative analysis involved calculating and comparing CT attenuation,image noise,signal-to-noise ratio(SNR),and contrast-to-noise ratio(CNR)within six key vessels:the aortic arch(AA),internal carotid artery(ICA),external carotid artery(ECA),vertebral arteries(VA),basilar artery(BA),and middle cerebral artery(MCA).Two radiologists independently assessed subjective image quality using a 5-point scale,with statistical significance defined as P<0.05.Results:Compared to the regular-dose group,the low-dose protocol achieved a substantial reduction in contrast media volume(28 mL versus 40 mL,a 30%decrease)and radiation exposure((0.41±0.08)mSv versus(1.18±0.12)mSv,a 65%reduction).Both L-DLR_(standard) and L-DLR_(strong) delivered comparable or superior SNR and CNR across all vascular segments relative to R-HIR.However,subjective image quality scores for L-DLR at all strength levels fell below those for R-HIR(all P<0.05 for both readers).Combining CE-boost with the low-dose protocol significantly enhanced the objective image performance of L-DLR_(strong)-CE(all P<0.05)and produced subjective image scores comparable to R-HIR(reader 1:P=0.15;reader 2:P=0.06).Conclusion:When compared to the standard 100 kVp head and neck CTA,the combination of the DLR and CE-boost techniques at 80 kVp can achieve a 30%reduction in contrast dose and a 65%reduction in radiation dose,while maintaining both objective and subjective image quality.
基金This work is supported in part by the National Science Foundation of China(Grant Nos.61672392,61373038)in part by the National Key Research and Development Program of China(Grant No.2016YFC1202204).
文摘Software defect prediction plays a very important role in software quality assurance,which aims to inspect as many potentially defect-prone software modules as possible.However,the performance of the prediction model is susceptible to high dimensionality of the dataset that contains irrelevant and redundant features.In addition,software metrics for software defect prediction are almost entirely traditional features compared to the deep semantic feature representation from deep learning techniques.To address these two issues,we propose the following two solutions in this paper:(1)We leverage a novel non-linear manifold learning method-SOINN Landmark Isomap(SL-Isomap)to extract the representative features by selecting automatically the reasonable number and position of landmarks,which can reveal the complex intrinsic structure hidden behind the defect data.(2)We propose a novel defect prediction model named DLDD based on hybrid deep learning techniques,which leverages denoising autoencoder to learn true input features that are not contaminated by noise,and utilizes deep neural network to learn the abstract deep semantic features.We combine the squared error loss function of denoising autoencoder with the cross entropy loss function of deep neural network to achieve the best prediction performance by adjusting a hyperparameter.We compare the SL-Isomap with seven state-of-the-art feature extraction methods and compare the DLDD model with six baseline models across 20 open source software projects.The experimental results verify that the superiority of SL-Isomap and DLDD on four evaluation indicators.
基金This work is supported by the Provincial Key Science and Technology Special Project of Henan(No.221100240100)。
文摘In recent years,the rapid development of computer software has led to numerous security problems,particularly software vulnerabilities.These flaws can cause significant harm to users’privacy and property.Current security defect detection technology relies on manual or professional reasoning,leading to missed detection and high false detection rates.Artificial intelligence technology has led to the development of neural network models based on machine learning or deep learning to intelligently mine holes,reducing missed alarms and false alarms.So,this project aims to study Java source code defect detection methods for defects like null pointer reference exception,XSS(Transform),and Structured Query Language(SQL)injection.Also,the project uses open-source Javalang to translate the Java source code,conducts a deep search on the AST to obtain the empty syntax feature library,and converts the Java source code into a dependency graph.The feature vector is then used as the learning target for the neural network.Four types of Convolutional Neural Networks(CNN),Long Short-Term Memory(LSTM),Bi-directional Long Short-Term Memory(BiLSTM),and Attention Mechanism+Bidirectional LSTM,are used to investigate various code defects,including blank pointer reference exception,XSS,and SQL injection defects.Experimental results show that the attention mechanism in two-dimensional BLSTM is the most effective for object recognition,verifying the correctness of the method.
文摘Soil is the major source of infinite lives on Earth and the quality of soil plays significant role on Agriculture practices all around.Hence,the evaluation of soil quality is very important for determining the amount of nutrients that the soil require for proper yield.In present decade,the application of deep learning models in many fields of research has created greater impact.The increasing soil data availability of soil data there is a greater demand for the remotely avail open source model,leads to the incorporation of deep learning method to predict the soil quality.With that concern,this paper proposes a novel model called Improved Soil Quality Prediction Model using Deep Learning(ISQP-DL).The work considers the chemical,physical and biological factors of soil in particular area to estimate the soil quality.Firstly,pH rating of soil samples has been collected from the soil testing laboratory from which the acidic range has been categorized through soil test and the same data has been taken as input to the Deep Neural Network Regression(DNNR)model.Secondly,soil nutrient data has been given as second input to the DNNR model.By utilizing this data set,the DNNR method is used to evaluate the fertility rate by which the soil quality has been estimated.For training and testing,the model uses Deep Neural Network Regression(DNNR),by utilizing the dataset.The results show that the proposed model is effective for SQP(Soil Quality Prediction Model)with efficient good fitting and generality is enhanced with input features with higher rate of classification accuracy.The results show that the proposed model achieves 96.7%of accuracy rate compared with existing models.
基金supported by the Fundamental Research Funds for the Central Universities(CUC230B008).
文摘Recommending high-quality news to users is vital in improving user stickiness and news platforms’reputation.However,existing news quality evaluation methods,such as clickbait detection and popularity prediction,are challenging to reflect news quality comprehensively and concisely.This paper defines news quality as the ability of news articles to elicit clicks and comments from users,which represents whether the news article can attract widespread attention and discussion.Based on the above definition,this paper first presents a straightforward method to measure news quality based on the comments and clicks of news and defines four news quality indicators.Then,the dataset can be labeled automatically by the method.Next,this paper proposes a deep learning model that integrates explicit and implicit news information for news quality evaluation(EINQ).The explicit information includes the headline,source,and publishing time of the news,which attracts users to click.The implicit information refers to the news article’s content which attracts users to comment.The implicit and explicit information affect users’click and comment behavior differently.For modeling explicit information,the typical convolution neural network(CNN)is used to get news headline semantic representation.For modeling implicit information,a hierarchical attention network(HAN)is exploited to extract news content semantic representation while using the latent Dirichlet allocation(LDA)model to get the subject distribution of news as a semantic supplement.Considering the different roles of explicit and implicit information for quality evaluation,the EINQ exploits an attention layer to fuse them dynamically.The proposed model yields the Accuracy of 82.31%and the F-Score of 80.51%on the real-world dataset from Toutiao,which shows the effectiveness of explicit and implicit information dynamic fusion and demonstrates performance improvements over a variety of baseline models in news quality evaluation.This work provides empirical evidence for explicit and implicit factors in news quality evaluation and a new idea for news quality evaluation.
文摘Over the last couple of decades,community question-answering sites(CQAs)have been a topic of much academic interest.Scholars have often leveraged traditional machine learning(ML)and deep learning(DL)to explore the ever-growing volume of content that CQAs engender.To clarify the current state of the CQA literature that has used ML and DL,this paper reports a systematic literature review.The goal is to summarise and synthesise the major themes of CQA research related to(i)questions,(ii)answers and(iii)users.The final review included 133 articles.Dominant research themes include question quality,answer quality,and expert identification.In terms of dataset,some of the most widely studied platforms include Yahoo!Answers,Stack Exchange and Stack Overflow.The scope of most articles was confined to just one platform with few cross-platform investigations.Articles with ML outnumber those with DL.Nonetheless,the use of DL in CQA research is on an upward trajectory.A number of research directions are proposed.
基金This work is supported by Chongqing Municipal Natural Science Foundation(Grant Nos.cstc2021jcyj-bsh0189 and cstc2019jcyj-bshX0070)Chongqing Jiulongpo District Science and Technology Planning Project(Grant No.2020-01-001-Y).
文摘Concrete exterior quality is one of the important metrics in evaluating construction project quality.Among the defects affecting concrete exterior quality,bughole is one of the most common imperfections,thus detecting concrete bughole accurately is significant for improving concrete exterior quality and consequently the quality of the whole project.This paper presents a deep learning-based method for detecting concrete surface bugholes in a more objective and automatic way.The bugholes are identified in concrete surface images by Mask R-CNN.An evaluation metric is developed to indicate the scale of concrete bughole.The proposed approach can detect bugholes in an instance level automatically and output the mask of each bughole,based on which the bughole area ratio is automatically calculated and the quality grade of the concrete surfaces is assessed.For demonstration,a total of 273 raw concrete surface images taken by mobile phone cameras are collected as a dataset.The test results show that the average precision(AP)of bughole masks is 90.8%.
基金Princess Nourah bint Abdulrahman University Researchers Supporting Project Number(PNURSP2022R77)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.The authors would like to thank the Deanship of Scientific Research at Umm Al-Qura University for supporting this work by Grant Code:(22UQU4331004DSR15).
文摘The paradigm shift towards the Internet of Things(IoT)phe-nomenon and the rise of edge-computing models provide massive poten-tial for several upcoming IoT applications like smart grid,smart energy,smart home,smart health and smart transportation services.However,it also provides a sequence of novel cyber-security issues.Although IoT networks provide several advantages,the heterogeneous nature of the network and the wide connectivity of the devices make the network easy for cyber-attackers.Cyberattacks result in financial loss and data breaches for organizations and individuals.So,it becomes crucial to secure the IoT environment from such cyberattacks.With this motivation,the current study introduces an effectual Enhanced Crow Search Algorithm with Deep Learning-Driven Cyberattack Detection(ECSADL-CAD)model for the Software-Defined Networking(SDN)-enabled IoT environment.The presented ECSADL-CAD approach aims to identify and classify the cyberattacks in the SDN-enabled IoT envi-ronment.To attain this,the ECSADL-CAD model initially pre-processes the data.In the presented ECSADL-CAD model,the Reinforced Deep Belief Network(RDBN)model is employed for attack detection.At last,the ECSA-based hyperparameter tuning process gets executed to boost the overall classification outcomes.A series of simulations were conducted to validate the improved outcomes of the proposed ECSADL-CAD model.The experimental outcomes confirmed the superiority of the proposed ECSADL-CAD model over other existing methodologies.
基金the National Natural Science Foundation of China,Grant/Award Number:62177003。
文摘Combined sewer overflows represent significant risks to human health as untreated water is discharged to the environment.Municipalities,such as the Metropolitan Sewer District of Greater Cincinnati(MSDGC),recently began collecting large amounts of water-related data and considering the adoption of deep learning(DL)solutions like recurrent neural network(RNN)for predicting overflow events.Clearly,assessing the DL's fitness for the purpose requires a systematic understanding of the problem context.In this study,we propose a requirements engineering framework that uses the problem frames to identify and structure the stakeholder concerns,analyses the physical situations in which the highquality data assumptions may not hold,and derives the software testing criteria in the form of metamorphic relations that incorporate both input transformations and output comparisons.Applying our framework to MSDGC's overflow prediction problem enables a principled way to evaluate different RNN solutions in meeting the requirements.
文摘Distributed denial-of-service(DDoS)attacks are designed to interrupt network services such as email servers and webpages in traditional computer networks.Furthermore,the enormous number of connected devices makes it difficult to operate such a network effectively.Software defined networks(SDN)are networks that are managed through a centralized control system,according to researchers.This controller is the brain of any SDN,composing the forwarding table of all data plane network switches.Despite the advantages of SDN controllers,DDoS attacks are easier to perpetrate than on traditional networks.Because the controller is a single point of failure,if it fails,the entire network will fail.This paper offers a Hybrid Deep Learning Intrusion Detection and Prevention(HDLIDP)framework,which blends signature-based and deep learning neural networks to detect and prevent intrusions.This framework improves detection accuracy while addressing all of the aforementioned problems.To validate the framework,experiments are done on both traditional and SDN datasets;the findings demonstrate a significant improvement in classification accuracy.
文摘Recently, many open source software (OSS) developed by various OSS projects. Also, the reliability assessment methods of OSS have been proposed by several researchers. Many methods for software reliability assessment have been proposed by software reliability growth models. Moreover, our research group has been proposed the method of reliability assessment for the OSS. Many OSS use bug tracking system (BTS) to manage software faults after it released. It keeps a detailed record of the environment in terms of the faults. There are several methods of reliability assessment based on deep learning for OSS fault data in the past. On the other hand, the data registered in BTS differences depending on OSS projects. Also, some projects have the specific collection data. The BTS has the specific collection data for each project. We focus on the recorded data. Moreover, we investigate the difference between the general data and the specific one for the estimation of OSS reliability. As a result, we show that the reliability estimation results by using specific data are better than the method using general data. Then, we show the characteristics between the specified data and general one in this paper. We also develop the GUI-based software to perform these reliability analyses so that even those who are not familiar with deep learning implementations can perform reliability analyses of OSS.
文摘Modern manufacturing processes have become more reliant on automation because of the accelerated transition from Industry 3.0 to Industry 4.0.Manual inspection of products on assembly lines remains inefficient,prone to errors and lacks consistency,emphasizing the need for a reliable and automated inspection system.Leveraging both object detection and image segmentation approaches,this research proposes a vision-based solution for the detection of various kinds of tools in the toolkit using deep learning(DL)models.Two Intel RealSense D455f depth cameras were arranged in a top down configuration to capture both RGB and depth images of the toolkits.After applying multiple constraints and enhancing them through preprocessing and augmentation,a dataset consisting of 3300 annotated RGB-D photos was generated.Several DL models were selected through a comprehensive assessment of mean Average Precision(mAP),precision-recall equilibrium,inference latency(target≥30 FPS),and computational burden,resulting in a preference for YOLO and Region-based Convolutional Neural Networks(R-CNN)variants over ViT-based models due to the latter’s increased latency and resource requirements.YOLOV5,YOLOV8,YOLOV11,Faster R-CNN,and Mask R-CNN were trained on the annotated dataset and evaluated using key performance metrics(Recall,Accuracy,F1-score,and Precision).YOLOV11 demonstrated balanced excellence with 93.0%precision,89.9%recall,and a 90.6%F1-score in object detection,as well as 96.9%precision,95.3%recall,and a 96.5%F1-score in instance segmentation with an average inference time of 25 ms per frame(≈40 FPS),demonstrating real-time performance.Leveraging these results,a YOLOV11-based windows application was successfully deployed in a real-time assembly line environment,where it accurately processed live video streams to detect and segment tools within toolkits,demonstrating its practical effectiveness in industrial automation.The application is capable of precisely measuring socket dimensions by utilising edge detection techniques on YOLOv11 segmentation masks,in addition to detection and segmentation.This makes it possible to do specification-level quality control right on the assembly line,which improves the ability to examine things in real time.The implementation is a big step forward for intelligent manufacturing in the Industry 4.0 paradigm.It provides a scalable,efficient,and accurate way to do automated inspection and dimensional verification activities.
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.
基金funded by National Natural Science Foundation of China(No.51908161&52100044)Guangdong Basic and Applied Basic Research Foundation(No.2019A1515010807)+1 种基金State Key Laboratory of Urban Water Resource and Environment(Harbin Institute of Technology)(2021TS30)Shenzhen Science and Technology Program(No.KQTD20190929172630447,KCXFZ20211020163404007 and GXWD20201230155427003-20200824100026001).
文摘Data-driven approaches that make timely predictions about pollutant concentrations in the effluent of constructed wetlands are essential for improving the treatment performance of constructed wetlands.However,the effect of the meteorological condition and flow changes in a real scenario are generally neglected in water quality prediction.To address this problem,in this study,we propose an approach based on multi-source data fusion that considers the following indicators:water quality indicators,water quantity indicators,and meteorological indicators.In this study,we establish four representative methods to simultaneously predict the concentrations of three representative pollutants in the effluent of a practical large-scale constructed wetland:(1)multiple linear regression;(2)backpropagation neural network(BPNN);(3)genetic algorithm combined with the BPNN to solve the local minima problem;and(4)long short-term memory(LSTM)neural network to consider the influence of past results on the present.The results suggest that the LSTM-predicting model performed considerably better than the other deep neural network-based model or linear method,with a satisfactory R^(2).Additionally,given the huge fluctuation of different pollutant concentrations in the effluent,we used a moving average method to smooth the original data,which successfully improved the accuracy of traditional neural networks and hybrid neural networks.The results of this study indicate that the hybrid modeling concept that combines intelligent and scientific data preprocessing methods with deep learning algorithms is a feasible approach for forecasting water quality in the effluent of actual engineering.
文摘We investigate a problem of object-oriented (OO) software quality estimation from a multi-instance (MI) perspective. In detail,each set of classes that have an inheritance relation,named 'class hierarchy',is regarded as a bag,while each class in the set is regarded as an instance. The learning task in this study is to estimate the label of unseen bags,i.e.,the fault-proneness of untested class hierarchies. A fault-prone class hierarchy contains at least one fault-prone (negative) class,while a non-fault-prone (positive) one has no negative class. Based on the modification records (MRs) of the previous project releases and OO software metrics,the fault-proneness of an untested class hierarchy can be predicted. Several selected MI learning algorithms were evalu-ated on five datasets collected from an industrial software project. Among the MI learning algorithms investigated in the ex-periments,the kernel method using a dedicated MI-kernel was better than the others in accurately and correctly predicting the fault-proneness of the class hierarchies. In addition,when compared to a supervised support vector machine (SVM) algorithm,the MI-kernel method still had a competitive performance with much less cost.
基金supported by the Department of Architecture and Built Environment,University of Nottingham,and the PhD studentship from EPSRC,Project References:2100822(EP/R513283/1).
文摘The built environment sector is responsible for almost one-third of the world’s final energy consumption. Hence, seeking plausible solutions to minimise building energy demands and mitigate adverse environmental impacts is necessary. Artificial intelligence (AI) techniques such as machine and deep learning have been increasingly and successfully applied to develop solutions for the built environment. This review provided a critical summary of the existing literature on the machine and deep learning methods for the built environment over the past decade, with special reference to holistic approaches. Different AI-based techniques employed to resolve interconnected problems related to heating, ventilation and air conditioning (HVAC) systems and enhance building performances were reviewed, including energy forecasting and management, indoor air quality and occupancy comfort/satisfaction prediction, occupancy detection and recognition, and fault detection and diagnosis. The present study explored existing AI-based techniques focusing on the framework, methodology, and performance. The literature highlighted that selecting the most suitable machine learning and deep learning model for solving a problem could be challenging. The recent explosive growth experienced by the research area has led to hundreds of machine learning algorithms being applied to building performance-related studies. The literature showed that existing research studies considered a wide range of scope/scales (from an HVAC component to urban areas) and time scales (minute to year). This makes it difficult to find an optimal algorithm for a specific task or case. The studies also employed a wide range of evaluation metrics, adding to the challenge. Further developments and more specific guidelines are required for the built environment field to encourage best practices in evaluating and selecting models. The literature also showed that while machine and deep learning had been successfully applied in building energy efficiency research, most of the studies are still at the experimental or testing stage, and there are limited studies which implemented machine and deep learning strategies in actual buildings and conducted the post-occupancy evaluation.
文摘Air quality estimation assesses the pollution level in the air,supports public health warnings,and is a valuable tool in environmental management.Although air sensors have proven helpful in this task,sensors are often expensive and difficult to install,while cameras are becoming more popular and accessible,from which images can be collected as data for deep learning models to solve the above task.This leads to another problem:several labeled images are needed to achieve high accuracy when deep-learningmodels predict air quality.In this research,we have threemain contributions:(1)Collect and publish an air quality estimation dataset,namely PTIT_AQED,including environmental image data and air quality;(2)Propose a deep learning model to predict air quality with few data,called PTIT_FAQE(PTIT Few-shot air quality estimation).We build PTIT_FAQE based on EfficientNet-a CNN architecture that ensures high performance in deep learning applications and Few-shot Learning with Prototypical Networks.This helps the model use only a fewtraining data but still achieve high accuracy in air quality estimation.And(3)conduct experiments to prove the superiority of PTIT_FAQE compared to other studies on both PTIT_AQED and APIN datasets.The results show that our model achieves an accuracy of 0.9278 and an F1-Score of 0.9139 on the PTIT_AQED dataset and an accuracy of 0.9467 and an F1-Score of 0.9371 on the APIN dataset,which demonstrate a significant performance improvement compared to previous studies.We also conduct detailed experiments to evaluate the impact of each component on model performance.
基金supported by the National Key R&D Program of China(Nos.2023YFC3708505,2023YFC3708503,and 2023YFE0121300)the High-End Foreign Expert Recruitment Program,China(No.G2023163014L)。
文摘Understanding the characteristics of O_(3)precursor contributions over multiple years is crucial for designing effective O_(3)control strategies over the Pearl River Delta(PRD)region of China.In this study,a deep learning-based response surface model(DeepRSM)was developed and applied over the PRD(DeepRSM-PRD)to identify and quantify the main features of O_(3)regimes and regional contributions in the core PRD over multiple years(2019–2021).The Out-of-Sample(OOS)validation results indicated that DeepRSM-PRD effectively predicted the nonlinear response of O_(3)to emission controls,maintaining validity across non-training periods.Our study revealed that O_(3)generation was sensitive to volatile organic compounds(VOC)in the core PRD in 2019,with nitrogen oxides(NO_(x))-limited regimes emerging in most major cities in 2020 and 2021.Further investigation into source contributions showed that in our model domain,O_(3)formation in central cities of the PRD was primarily driven by local contributions and was susceptible to influence from nearby cities.With small emission reductions,VOC contributions predominantly drive O_(3)production in Guangzhou and Shenzhen.However,NO_(x)emissions were identified as the primary contributors in all central city receptors when anthropogenic emissions were removed,sharing 59.5%–69.3%in 2019,64.4%–72.3%in 2020,and 62.75%–73.2%in 2021.Our results highlight the need for a high focus on NO_(x)emissions control in the core PRD.In addition,for Guangzhou and Shenzhen,VOC reduction also plays a crucial role in the initial stages of modest emission reductions.
