An improved method to construct basic probability assignment based on the confusion matrix for classification problem

Enhancing fraud detection in auto insurance and credit card transactions: a novel approach integrating CNNs and machine learning algorithms

Fraudulent activities especially in auto insurance and credit card transactions impose significant financial losses on businesses and individuals. To overcome this issue, we propose a novel approach for fraud detection, combining convolutional neural networks (CNNs) with support vector machine (SVM), k nearest neighbor (KNN), naive Bayes (NB), and decision tree (DT) algorithms. The core of this methodology lies in utilizing the deep features extracted from the CNNs as inputs to various machine learning models, thus significantly contributing to the enhancement of fraud detection accuracy and efficiency. Our results demonstrate superior performance compared to previous studies, highlighting our model's potential for widespread adoption in combating fraudulent activities.

An enhanced speech emotion recognition using vision transformer

In human-computer interaction systems, speech emotion recognition (SER) plays a crucial role because it enables computers to understand and react to users' emotions. In the past, SER has significantly emphasised acoustic properties extracted from speech signals. The use of visual signals for enhancing SER performance, however, has been made possible by recent developments in deep learning and computer vision. This work utilizes a lightweight Vision Transformer (ViT) model to propose a novel method for improving speech emotion recognition. We leverage the ViT model's capabilities to capture spatial dependencies and high-level features in images which are adequate indicators of emotional states from mel spectrogram input fed into the model. To determine the efficiency of our proposed approach, we conduct a comprehensive experiment on two benchmark speech emotion datasets, the Toronto English Speech Set (TESS) and the Berlin Emotional Database (EMODB). The results of our extensive experiment demonstrate a considerable improvement in speech emotion recognition accuracy attesting to its generalizability as it achieved 98%, 91%, and 93% (TESS-EMODB) accuracy respectively on the datasets. The outcomes of the comparative experiment show that the non-overlapping patch-based feature extraction method substantially improves the discipline of speech emotion recognition. Our research indicates the potential for integrating vision transformer models into SER systems, opening up fresh opportunities for real-world applications requiring accurate emotion recognition from speech compared with other state-of-the-art techniques.

GNBoost-Based Ensemble Machine Learning for Predicting Tribological Properties of Liquid-Crystal Lubricants

The intricate development of liquid-crystal lubricants necessitates the timely and accurate prediction of their tribological performance in different environments and an assessment of the importance of relevant parameters. In this study, a classification model using Gaussian noise extreme gradient boosting (GNBoost) to predict tribological performance is proposed. Three additives, polysorbate-85, polysorbate-80, and graphene oxide, were selected to fabricate liquid-crystal lubricants. The coefficients of friction of these lubricants were tested in the rotational mode using a universal mechanical tester. A model was designed to predict the coefficient of friction through data augmentation of the initial data. The model parameters were optimized using particle swarm optimization techniques. This study provides an effective example for lubricant performance evaluation and formulation optimization.

Detection of the separated endodontic instrument on periapical radiographs using a deep learning-based convolutional neural network algorithm

The study evaluated the diagnostic performance of an artificial intelligence system to detect separated endodontic instruments on periapical radiograph radiographs. Three hundred seven periapical radiographs were collected and divided into 222 for training and 85 for testing to be fed to the Mask R-CNN model. Periapical radiographs were assigned to the training and test set and labelled on the DentiAssist labeling platform. Labelled polygonal objects had their bounding boxes automatically generated by the DentiAssist system. Fractured instruments were classified and segmented. As a result of the proposed method, the mean average precision (mAP) metric was 98.809%, the precision value was 95.238, while the recall reached 98.765 and the f1 score 96.969%. The threshold value of 80% was chosen for the bounding boxes working with the Intersection over Union (IoU) technique. The Mask R-CNN distinguished separated endodontic instruments on periapical radiographs.

Exploiting nearest neighbor data and fuzzy membership function to address missing values in classification

The accuracy of most classification methods is significantly affected by missing values. Therefore, this study aimed to propose a data imputation method to handle missing values through the application of nearest neighbor data and fuzzy membership function as well as to compare the results with standard methods. A total of five datasets related to classification problems obtained from the UCI Machine Learning Repository were used. The results showed that the proposed method had higher accuracy than standard imputation methods. Moreover, triangular method performed better than Gaussian fuzzy membership function. This showed that the combination of nearest neighbor data and fuzzy membership function was more effective in handling missing values and improving classification accuracy.

Discrimination among Fresh, Frozen-Stored and Frozen-Thawed Beef Cuts by Hyperspectral Imaging

The detection of the storage state of frozen meat, especially meat frozen-thawed several times, has always been important for food safety inspections. Hyperspectral imaging (HSI) is widely applied to detect the freshness and quality of meat or meat products. This study investigated the feasibility of the low-cost HSI system, combined with the chemometrics method, to classify beef cuts among fresh (F), frozen-stored (F-S), frozen-thawed three times (F-T-3) and frozen-thawed five times (F-T-5). A compact, low-cost HSI system was designed and calibrated for beef sample measurement. The classification model was developed for meat analysis with a method to distinguish fat and muscle, a CARS algorithm to extract the optimal wavelength subset and three classifiers to identify each beef cut among different freezing processes. The results demonstrated that classification models based on feature variables extracted from differentiated tissue spectra achieved better performances, with ACCs of 92.75% for PLS-DA, 97.83% for SVM and 95.03% for BP-ANN. A visualization map was proposed to provide detailed information about the changes in freshness of beef cuts after freeze-thawing. Furthermore, this study demonstrated the potential of implementing a reasonably priced HSI system in the food industry.

YOLOv5 based object detection in reel package X-ray images of semiconductor component

The industrial manufacturing landscape is currently shifting toward the incorporation of technologies based on artificial intelligence (AI). This transition includes an evolution toward smart factory infrastructure, with a specific focus on AI-driven strategies in production and quality control. Specifically, AI-empowered computer vision has emerged as a potent tool that offers a departure from extant rule-based systems and provides enhanced operational efficiency at manufacturing sites. As the manufacturing sector embraces this new paradigm, the impetus to integrate AI-integrated manufacturing is evident. Within this framework, one salient application is AI deep learning-facilitated small-object detection, which is poised to have extensive implications for diverse industrial applications. This study describes an optimized iteration of the YOLOv5 model, which is known for its efficacious single-stage object-detection abilities underpinned by PyTorch. Our proposed "improved model" incorporates an additional layer to the model's canonical three-layer architecture, augmenting accuracy and computational expediency. Empirical evaluations using semiconductor X-ray imagery reveal the model's superior performance metrics. Given the intricate specifications of surface-mount technologies, which are characterized by a plethora of micro-scale components, our model makes a seminal contribution to real-time, in-line production assessments. Quantitative analyses show that our improved model attained a mean average precision of 0.622, surpassing YOLOv5's 0.349, and a marked accuracy enhancement of 0.865, which is a significant improvement on YOLOv5's 0.552. These findings bolster the model's robustness and potential applicability, particularly in discerning objects at reel granularities during real-time inferencing.

Survival prediction of glioblastoma patients using modern deep learning and machine learning techniques

In this study, we utilized data from the Surveillance, Epidemiology, and End Results (SEER) database to predict the glioblastoma patients' survival outcomes. To assess dataset skewness and detect feature importance, we applied Pearson's second coefficient test of skewness and the Ordinary Least Squares method, respectively. Using two sampling strategies, holdout and five-fold cross-validation, we developed five machine learning (ML) models alongside a feed-forward deep neural network (DNN) for the multiclass classification and regression prediction of glioblastoma patient survival. After balancing the classification and regression datasets, we obtained 46,340 and 28,573 samples, respectively. Shapley additive explanations (SHAP) were then used to explain the decision-making process of the best model. In both classification and regression tasks, as well as across holdout and cross-validation sampling strategies, the DNN consistently outperformed the ML models. Notably, the accuracy were 90.25% and 90.22% for holdout and five-fold cross-validation, respectively, while the corresponding R2 values were 0.6565 and 0.6622. SHAP analysis revealed the importance of age at diagnosis as the most influential feature in the DNN's survival predictions. These findings suggest that the DNN holds promise as a practical auxiliary tool for clinicians, aiding them in optimal decision-making concerning the treatment and care trajectories for glioblastoma patients.

A Deep Learning-Based Decision Support Tool for Plant-Parasitic Nematode Management

Plant-parasitic nematodes (PPN), especially sedentary endoparasitic nematodes like root-knot nematodes (RKN), pose a significant threat to major crops and vegetables. They are responsible for causing substantial yield losses, leading to economic consequences, and impacting the global food supply. The identification of PPNs and the assessment of their population is a tedious and time-consuming task. This study developed a state-of-the-art deep learning model-based decision support tool to detect and estimate the nematode population. The decision support tool is integrated with the fast inferencing YOLOv5 model and used pretrained nematode weight to detect plant-parasitic nematodes (juveniles) and eggs. The performance of the YOLOv5-640 model at detecting RKN eggs was as follows: precision = 0.992; recall = 0.959; F1-score = 0.975; and mAP = 0.979. YOLOv5-640 was able to detect RKN eggs with an inference time of 3.9 milliseconds, which is faster compared to other detection methods. The deep learning framework was integrated into a user-friendly web application system to build a fast and reliable prototype nematode decision support tool (NemDST). The NemDST facilitates farmers/growers to input image data, assess the nematode population, track the population growths, and recommend immediate actions necessary to control nematode infestation. This tool has the potential for rapid assessment of the nematode population to minimise crop yield losses and enhance financial outcomes.

Bone marrow sparing oriented multi-model image registration in cervical cancer radiotherapy

Cervical cancer poses a serious threat to the health of women and radiotherapy is one of the primary treatment methods for this condition. However, this treatment is associated with a high risk of causing acute hematologic toxicity. Delineating the bone marrow (BM) for sparing based on computer tomography (CT) images before radiotherapy can effectively avoid this risk. Unfortunately, compared to magnetic resonance (MR) images, CT images lack the ability to express the activity of BM. Therefore, medical practitioners currently manually delineate the BM on CT images by corresponding to MR images. However, the manual delineation of BM is time-consuming and cannot guarantee accuracy due to the inconsistency of the CT-MR multimodal images. This study proposes a multimodal image-oriented automatic registration method for pelvic BM sparing. The proposed method includes three-dimensional (3D) bone point clouds reconstruction and an iterative closest point registration based on a local spherical system for marking BM on CT images. By introducing a joint coordinate system that combines the global Cartesian coordinate system with the local point clouds' spherical coordinate system, the increasement of point descriptive dimension avoids the local optimal registration and improves the registration accuracy. Experiments on the dataset of patients demonstrate that our proposed method can enhance the multimodal image registration accuracy and efficiency for medical practitioners in BM-sparing of cervical cancer radiotherapy. The method proposed in this contribution might also provide a solution to multimodal registration, especially in multimodal sequential images in other clinical applications, such as the diagnosis of cervical cancer and the preservation of normal organs during radiotherapy.

Machine learning for distinguishing saudi children with and without autism via eye-tracking data

BACKGROUND

Despite the prevalence of Autism Spectrum Disorder (ASD) globally, there's a knowledge gap pertaining to autism in Arabic nations. Recognizing the need for validated biomarkers for ASD, our study leverages eye-tracking technology to understand gaze patterns associated with ASD, focusing on joint attention (JA) and atypical gaze patterns during face perception. While previous studies typically evaluate a single eye-tracking metric, our research combines multiple metrics to capture the multidimensional nature of autism, focusing on dwell times on eyes, left facial side, and joint attention.

METHODS

We recorded data from 104 participants (41 neurotypical, mean age: 8.21 ± 4.12 years; 63 with ASD, mean age 8 ± 3.89 years). The data collection consisted of a series of visual stimuli of cartoon faces of humans and animals, presented to the participants in a controlled environment. During each stimulus, the eye movements of the participants were recorded and analyzed, extracting metrics such as time to first fixation and dwell time. We then used these data to train a number of machine learning classification algorithms, to determine if these biomarkers can be used to diagnose ASD.

RESULTS

We found no significant difference in eye-dwell time between autistic and control groups on human or animal eyes. However, autistic individuals focused less on the left side of both human and animal faces, indicating reduced left visual field (LVF) bias. They also showed slower response times and shorter dwell times on congruent objects during joint attention (JA) tasks, indicating diminished reflexive joint attention. No significant difference was found in time spent on incongruent objects during JA tasks. These results suggest potential eye-tracking biomarkers for autism. The best-performing algorithm was the random forest one, which achieved accuracy = 0.76 ± 0.08, precision = 0.78 ± 0.13, recall = 0.84 ± 0.07, and F1 = 0.80 ± 0.09.

CONCLUSIONS

Although the autism group displayed notable differences in reflexive joint attention and left visual field bias, the dwell time on eyes was not significantly different. Nevertheless, the machine algorithm model trained on these data proved effective at diagnosing ASD, showing the potential of these biomarkers. Our study shows promising results and opens up potential for further exploration in this under-researched geographical context.

Assessment of Carbon Sequestration Capacity of E. ulmoides in Ruyang County and Its Ecological Suitability Zoning Based on Satellite Images of GF-6

Eucommia ulmoides Oliver. (E. ulmoides) is a species of small tree native to China. It is a valuable medicinal herb that can be used to treat Alzheimer's disease, diabetes, hypertension, and other diseases. In addition, E. ulmoides is a source of rubber. It has both medicinal and ecological value. As ecological problems become increasingly prominent, accurate information on the cultivated area of E. ulmoides is important for understanding the carbon sequestration capacity and ecological suitability zoning of E. ulmoides. In previous tree mapping studies, no studies on the spectral characteristics of E. ulmoides and its remote sensing mapping have been seen. We use Ruyang County, Henan Province, China, as the study area. Firstly, using the 2021 Gao Fen-6 (GF-6) Wide Field of View (WFV) time series images covering the different growth stages of E. ulmoides based on the participation of red-edge bands, several band combination schemes were constructed. The optimal time window to identify E. ulmoides was selected by calculating the separability of E. ulmoides from other land cover types for different schemes. Secondly, a random forest algorithm based on several band combination schemes was investigated to map the E. ulmoides planting areas in Ruyang County. Thirdly, the annual NPP values of E. ulmoides were estimated using an improved Carnegie Ames Stanford Approach (CASA) to a light energy utilization model, which, in turn, was used to assess the carbon sequestration capacity. Finally, the ecologically suitable distribution zone of E. ulmoides under near current and future (2041-2060) climatic conditions was predicted using the MaxEnt model. The results showed that the participation of the red-edge band of the GF-6 data in the classification could effectively improve the recognition accuracy of E. ulmoides, making its overall accuracy reach 96.62%; the high NPP value of E. ulmoides was mainly concentrated in the south of Ruyang County, with a total annual carbon sequestration of 540.104835 t CM-2·a-1. The ecological suitability zone of E. ulmoides can be divided into four classes: unsuitable area, low suitable area, medium suitable area, and high suitable area. The method proposed in this paper applies to the real-time monitoring of E. ulmoides, highlighting its potential ecological value and providing theoretical reference and data support for the reasonable layout of E. ulmoides.

Coal-gangue recognition via multi-branch convolutional neural network based on MFCC in noisy environment

Traditional coal-gangue recognition methods usually do not consider the impact of equipment noise, which severely limits its adaptability and recognition accuracy. This paper mainly studies the more accurate recognition of coal-gangue in the noise site environment with the operation of shearer, conveyor, transfer machine and other device in the process of top coal caving. Mel Frequency Cepstrum Coefficients (MFCC) smoothing method was introduced to express the intrinsic feature of sound pressure more clearly in the coal-gangue recognition site. Then, a multi-branch convolution neural network (MBCNN) model with three branches was developed, and the smoothed MFCC feature was incorporated into this model to realize the recognition of falling coal and gangue in noisy environment. The sound pressure signal datasets under the operation of different device were constructed through a great deal of laboratory and site data acquisition. Comparative experiments were carried out on noiseless dataset, single noise dataset and simulated site dataset, and the results show that our method can provide higher correct recognition accuracy and better robustness. The proposed coal-gangue recognition approach based on MBCNN and MFCC smoothing can not only recognize the state of falling coal or gangue, but also recognize the operational state of site device.

A Metabolomics-Based Study on the Discriminative Classification Models and Toxicological Mechanism of Estazolam Fatal Intoxication

Fatal intoxication with sedative-hypnotic drugs is increasing yearly. However, the plasma drug concentration data for fatal intoxication involving these substances are not systematic and even overlap with the intoxication group. Therefore, developing a more precise and trustworthy approach to determining the cause of death is necessary. This study analyzed mice plasma and brainstem samples using the liquid chromatography-high resolution tandem mass spectrometry (LC-HR MS/MS)-based metabolomics method to create discriminative classification models for estazolam fatal intoxication (EFI). The most perturbed metabolic pathway between the EFI and EIND (estazolam intoxication non-death) was examined, Both EIND and EFI groups were administered 500 mg of estazolam per 100 g of body weight. Mice that did not die beyond 8 hours were treated with cervical dislocation and were classified into the EIND groups; the lysine degradation pathway was verified by qPCR (Quantitative Polymerase Chain Reaction), metabolite quantitative and TEM (transmission electron microscopy) analysis. Non-targeted metabolomics analysis with EFI were the experimental group and four hypoxia-related non-drug-related deaths (NDRDs) were the control group. Mass spectrometry data were analyzed with Compound Discoverer (CD) 3.1 software and multivariate statistical analyses were performed using the online software MetaboAnalyst 5.0. After a series of analyses, the results showed the discriminative classification model in plasma was composed of three endogenous metabolites: phenylacetylglycine, creatine and indole-3-lactic acid, and in the brainstem was composed of palmitic acid, creatine, and indole-3-lactic acid. The specificity validation results showed that both classification models distinguished between the other four sedatives-hypnotics, with an area under ROC curve (AUC) of 0.991, and the classification models had an extremely high specificity. When comparing different doses of estazolam, the AUC value of each group was larger than 0.80, and the sensitivity was also high. Moreover, the stability results showed that the AUC value was equal to or very close to 1 in plasma samples stored at 4 °C for 0, 1, 5, 10 and 15 days; the predictive power of the classification model was stable within 15 days. The results of lysine degradation pathway validation revealed that the EFI group had the highest lysine and saccharopine concentrations (mean (ng/mg) = 1.089 and 1.2526, respectively) when compared to the EIND and control group, while the relative expression of SDH (saccharopine dehydrogenase) showed significantly lower in the EFI group (mean = 1.206). Both of these results were statistically significant. Furthermore, TEM analysis showed that the EFI group had the more severely damaged mitochondria. This work gives fresh insights into the toxicological processes of estazolam and a new method for identifying EFI-related causes of mortality.

Neural Network Models for Driving Control of Indoor Autonomous Vehicles in Mobile Edge Computing

Mobile edge computing has been proposed as a solution for solving the latency problem of traditional cloud computing. In particular, mobile edge computing is needed in areas such as autonomous driving, which requires large amounts of data to be processed without latency for safety. Indoor autonomous driving is attracting attention as one of the mobile edge computing services. Furthermore, it relies on its sensors for location recognition because indoor autonomous driving cannot use a GPS device, as is the case with outdoor driving. However, while the autonomous vehicle is being driven, the real-time processing of external events and the correction of errors are required for safety. Furthermore, an efficient autonomous driving system is required because it is a mobile environment with resource constraints. This study proposes neural network models as a machine-learning method for autonomous driving in an indoor environment. The neural network model predicts the most appropriate driving command for the current location based on the range data measured with the LiDAR sensor. We designed six neural network models to be evaluated according to the number of input data points. In addition, we made an autonomous vehicle based on the Raspberry Pi for driving and learning and an indoor circular driving track for collecting data and performance evaluation. Finally, we evaluated six neural network models in terms of confusion matrix, response time, battery consumption, and driving command accuracy. In addition, when neural network learning was applied, the effect of the number of inputs was confirmed in the usage of resources. The result will influence the choice of an appropriate neural network model for an indoor autonomous vehicle.

The prediction of contact force networks in granular materials based on graph neural networks

The contact force network, usually organized inhomogeneously by the inter-particle forces on the bases of the contact network topologies, is essential to the rigidity and stability in amorphous solids. How to capture such a "backbone" is crucial to the understanding of various anomalous properties or behaviors in those materials, which remains a central challenge presently in physics, engineering, or material science. Here, we use a novel graph neural network to predict the contact force network in two-dimensional granular materials under uniaxial compression. With the edge classification model in the framework of the deep graph library, we show that the inter-particle contact forces can be accurately estimated purely from the knowledge of the static microstructures, which can be acquired from a discrete element method or directly visualized from experimental methods. By testing the granular packings with different structural disorders and pressure, we further demonstrate the robustness of the optimized graph neural network to changes in various model parameters. Our research tries to provide a new way of extracting the information about the inter-particle forces, which substantially improves the efficiency and reduces the costs compared to the traditional experiments.

Visual Detection of Water Content Range of Seabuckthorn Fruit Based on Transfer Deep Learning

To realize the classification of sea buckthorn fruits with different water content ranges, a convolution neural network (CNN) detection model of sea buckthorn fruit water content ranges was constructed. In total, 900 images of seabuckthorn fruits with different water contents were collected from 720 seabuckthorn fruits. Eight classic network models based on deep learning were used as feature extraction for transfer learning. A total of 180 images were randomly selected from the images of various water content ranges for testing. Finally, the identification accuracy of the network model for the water content range of seabuckthorn fruit was 98.69%, and the accuracy on the test set was 99.4%. The program in this study can quickly identify the moisture content range of seabuckthorn fruit by collecting images of the appearance and morphology changes during the drying process of seabuckthorn fruit. The model has a good detection effect for seabuckthorn fruits with different moisture content ranges with slight changes in characteristics. The migration deep learning can also be used to detect the moisture content range of other agricultural products, providing technical support for the rapid nondestructive testing of moisture contents of agricultural products.

Structural Analysis and Classification of Low-Molecular-Weight Hyaluronic Acid by Near-Infrared Spectroscopy: A Comparison between Traditional Machine Learning and Deep Learning

Confusing low-molecular-weight hyaluronic acid (LMWHA) from acid degradation and enzymatic hydrolysis (named LMWHA-A and LMWHA-E, respectively) will lead to health hazards and commercial risks. The purpose of this work is to analyze the structural differences between LMWHA-A and LMWHA-E, and then achieve a fast and accurate classification based on near-infrared (NIR) spectroscopy and machine learning. First, we combined nuclear magnetic resonance (NMR), Fourier transform infrared (FTIR) spectroscopy, two-dimensional correlated NIR spectroscopy (2DCOS), and aquaphotomics to analyze the structural differences between LMWHA-A and LMWHA-E. Second, we compared the dimensionality reduction methods including principal component analysis (PCA), kernel PCA (KPCA), and t-distributed stochastic neighbor embedding (t-SNE). Finally, the differences in classification effect of traditional machine learning methods including partial least squares-discriminant analysis (PLS-DA), support vector classification (SVC), and random forest (RF) as well as deep learning methods including one-dimensional convolutional neural network (1D-CNN) and long short-term memory (LSTM) were compared. The results showed that genetic algorithm (GA)-SVC and RF were the best performers in traditional machine learning, but their highest accuracy in the test dataset was 90%, while the accuracy of 1D-CNN and LSTM models in the training dataset and test dataset classification was 100%. The results of this study show that compared with traditional machine learning, the deep learning models were better for the classification of LMWHA-A and LMWHA-E. Our research provides a new methodological reference for the rapid and accurate classification of biological macromolecules.

A transfer learning approach for multiclass classification of Alzheimer's disease using MRI images

Alzheimer's is an acute degenerative disease affecting the elderly population all over the world. The detection of disease at an early stage in the absence of a large-scale annotated dataset is crucial to the clinical treatment for the prevention and early detection of Alzheimer's disease (AD). In this study, we propose a transfer learning base approach to classify various stages of AD. The proposed model can distinguish between normal control (NC), early mild cognitive impairment (EMCI), late mild cognitive impairment (LMCI), and AD. In this regard, we apply tissue segmentation to extract the gray matter from the MRI scans obtained from the Alzheimer's Disease National Initiative (ADNI) database. We utilize this gray matter to tune the pre-trained VGG architecture while freezing the features of the ImageNet database. It is achieved through the addition of a layer with step-wise freezing of the existing blocks in the network. It not only assists transfer learning but also contributes to learning new features efficiently. Extensive experiments are conducted and results demonstrate the superiority of the proposed approach.

Dynamic probability integration for electroencephalography-based rapid serial visual presentation performance enhancement: Application in nighttime vehicle detection

Background

Rapid serial visual presentation (RSVP) has become a popular target detection method by decoding electroencephalography (EEG) signals, owing to its sensitivity and effectiveness. Most current research on EEG-based RSVP tasks focused on feature extraction algorithms developed to deal with the non-stationarity and low signal-to-noise ratio (SNR) of EEG signals. However, these algorithms cannot handle the problem of no event-related potentials (ERP) component or miniature ERP components caused by the attention lapses of human vision in abnormal conditions. The fusion of human-computer vision can obtain complementary information, making it a promising way to become an efficient and general way to detect objects, especially in attention lapses.

Methods

Dynamic probability integration (DPI) was proposed in this study to fuse human vision and computer vision. A novel basic probability assignment (BPA) method was included, which can fully consider the classification capabilities of different heterogeneous information sources for targets and non-targets and constructs the detection performance model for the weight generation based on classification capabilities. Furthermore, a spatial-temporal hybrid common spatial pattern-principal component analysis (STHCP) algorithm was designed to decode EEG signals in the RSVP task. It is a simple and effective method of distinguishing target and non-target using spatial-temporal features.

Results

A nighttime vehicle detection based on the RSVP task was performed to evaluate the performance of DPI and STHCP, which is one of the conditions of attention lapses because of its decrease in visual information. The average AUC of DPI was 0.912 ± 0.041 and increased by 11.5, 5.2, 3.4, and 1.7% compared with human vision, computer vision, naive Bayesian fusion, and dynamic belief fusion (DBF), respectively. A higher average balanced accuracy of 0.845 ± 0.052 was also achieved using DPI, representing that DPI has the balanced detection capacity of target and non-target. Moreover, STHCP obtained the highest AUC of 0.818 ± 0.06 compared with the other two baseline methods and increased by 15.4 and 23.4%.

Conclusion

Experimental results indicated that the average AUC and balanced accuracy of the proposed fusion method were higher than individual detection methods used for fusion, as well as two excellent fusion methods. It is a promising way to improve detection performance in RSVP tasks, even in abnormal conditions.

A Study on Deep Learning-Based Fault Diagnosis and Classification for Marine Engine System Auxiliary Equipment

Maritime autonomous surface ships (MASS) are proposed as a future technology of the maritime industry. One of the key technologies for the development of MASS is condition-based maintenance (CBM) based on prognostics and health management (PHM). The CBM technology can be used for early detection of abnormalities based on the database and for a prediction of the fault occurring in the future. However, this technology has a problem that requires a high-quality database that reproduces the operation state of the actual ships and quantitatively and systematically indicates the characteristics for the various fault state of the device. To solve this problem, this paper presents a study on the development method of the fault database based on the reliability. Firstly, the reliability analysis of the target device was performed to select five types of the core fault modes. After that, a fault simulation scenario that defined the fault simulation test methodology was drawn. A land-based testbed was built for the fault simulation test. The fault simulation database was developed with a total of 109 sets through the fault simulation test. Additionally, a fault classification algorithm based on deep learning is proposed. The classification performance was evaluated with a confusion matrix. The developed database will be expected to serve as the basis for the development CBM technology of MASS in the future.

U-Net-Based Segmentation of Microscopic Images of Colorants and Simplification of Labeling in the Learning Process

Colored product textures correspond to particle size distributions. The microscopic images of colorants must be divided into regions to determine the particle size distribution. The conventional method used for this process involves manually dividing images into areas, which may be inefficient. In this paper, we have overcome this issue by developing two different modified architectures of U-Net convolution neural networks to automatically determine the particle sizes. To develop these modified architectures, a significant amount of ground truth data must be prepared to train the U-Net, which is difficult for big data as the labeling is performed manually. Therefore, we also aim to reduce this process by using incomplete labeling data. The first objective of this study is to determine the accuracy of our modified U-Net architectures for this type of image. The second objective is to reduce the difficulty of preparing the ground truth data by testing the accuracy of training on incomplete labeling data. The results indicate that efficient segmentation can be realized using our modified U-Net architectures, and the generation of ground truth data can be simplified. This paper presents a preliminary study to improve the efficiency of determining particle size distributions with incomplete labeling data.

Deep Learning Approach for Damage Classification Based on Acoustic Emission Data in Composite Materials

Damage detection and the classification of carbon fiber-reinforced composites using non-destructive testing (NDT) techniques are of great importance. This paper applies an acoustic emission (AE) technique to obtain AE data from three tensile damage tests determining fiber breakage, matrix cracking, and delamination. This article proposes a deep learning approach that combines a state-of-the-art deep learning technique for time series classification: the InceptionTime model with acoustic emission data for damage classification in composite materials. Raw AE time series and frequency-domain sequence data are used as the input for the InceptionTime network, and both obtain very high classification performances, achieving high accuracy scores of about 99%. The InceptionTime network produces better training, validation, and test accuracy with the raw AE time series data than it does with the frequency-domain sequence data. Simultaneously, the InceptionTime model network shows its potential in dealing with data imbalances.

Convolutional Neural Network for Measurement of Suspended Solids and Turbidity

The great potential of the convolutional neural networks (CNNs) provides novel and alternative ways to monitor important parameters with high accuracy. In this study, we developed a soft sensor model for dynamic processes based on a CNN for the measurement of suspended solids and turbidity from a single image of the liquid sample to be measured by using a commercial smartphone camera (Android or IOS system) and light-emitting diode (LED) illumination. For this, an image dataset of liquid samples illuminated with white, red, green, and blue LED light was taken and used to train a CNN and fit a multiple linear regression (MLR) by using different color lighting, we evaluated which color gives more accurate information about the concentration of suspended particles in the sample. We implemented a pre-trained AlexNet model, and an MLR to estimate total suspended solids (TSS), and turbidity values in liquid samples based on suspended particles. The proposed technique obtained high goodness of fit (R2 = 0.99). The best performance was achieved using white light, with an accuracy of 98.24% and 97.20% for TSS and turbidity, respectively, with an operational range of 0–800 mgL−1, and 0–306 NTU. This system was designed for aquaculture environments and tested with both commercial fish feed and paprika. This motivates further research with different aquatic environments such as river water, domestic and industrial wastewater, and potable water, among others.

An ideographic study into physiology, alcohol craving and lapses during one hundred days of daily life monitoring

•

In a 100 days intensive study considerable intra- and interindividual differences were established in physiology and psychological craving of people with alcohol use disorder.

•

For one third of the people heightened heart rate was associated with high craving.

•

For most participants that reported lapses, lapses cooccurred with craving in at least 92% of the time.

•

During treatment ambulatory physiological data can support the detection and discussion of possible high risk craving situations in innovative and reliable ways.

Introduction

Alcohol craving is a highly challenging obstacle to achieve long-term abstinence. Making alcohol use disorder patients timely aware of high-risk craving situations may protect them against relapse by prompting them to mobilize their coping resources. Current advances in wearable and smart-phone technology provide novel opportunities for the development of detecting these situations of heightened risk of craving, by enabling continuous tracking of fluctuations in psychological and physiological parameters. The present study therefore aims to determine the association between self-reported craving and relapses, and between heightened physiological activity. Specifically, we measured cardiovascular and electrodermal activity, and self-reported craving during one hundred days in the daily life of people trying to recover from alcoholism. The secondary aim is to study whether the association between physiology and craving can be strengthened by the inclusion of context related psychological parameters.

Methods

An intensive repeated and continuous measures in naturalistic settings case-study design was employed. Ten participants were monitored with wearable bio-sensors and answered multiple questions every three hours on a smartphone app about craving, lapsing and multiple evidence based contextual variables. The association between physiology, craving and lapses was explored using Matthews correlation coefficients both with a current and 3 h lagged design. The contextual variables were included in a decision tree together with the physiological parameters to explore the added effect on the correlation of these contextual variables.

Results

The association between lapses and craving was highly different across individuals, varying between a weak to a strong association. The association between cardiovascular activity and heightened self-reported craving was negligible to weak, however with a high specificity, meaning that most craving events were accompanied by increase heart rate. However, the association between electrodermal activity and craving was lower than with cardiovascular activity for most participants, both prior (lagged) and during craving. For two of the participants the association between physiology and craving improved by adding contextual variables, however, precision was too low.

Conclusions

People differ strongly in their bodily reactions and psychological experiences during the first months of their addiction treatment. No individual in our study had unique one-to-one mappings between on the one hand physiological or psychological precursors, and on the other hand craving and (re)lapses. Therefore, detecting high risk craving situations with both physiological activity measured with wearables and psychological precursors to alert people specifically for an imminent (re)lapse, does not seem viable on the basis of the current results. We do see an added benefit of using physiology during treatment, as physiology can help start the conversation about possible high risk craving situations during that week. This would also help the counselor to gain added insights into the fluctuating states of the clients, and help to ameliorate the recall bias of clients. The present study showed the possibility and paved the way for future intensive longitudinal designs integrating both physiological, psychological and contextual factors during the challenging and lengthy recovery from addiction.

A new approach for determining SARS-CoV-2 epitopes using machine learning-based in silico methods

The emergence of machine learning-based in silico tools has enabled rapid and high-quality predictions in the biomedical field. In the COVID-19 pandemic, machine learning methods have been used in many topics such as predicting the death of patients, modeling the spread of infection, determining future effects, diagnosis with medical image analysis, and forecasting the vaccination rate. However, there is a gap in the literature regarding identifying epitopes that can be used in fast, useful, and effective vaccine design using machine learning methods and bioinformatics tools. Machine learning methods can give medical biotechnologists an advantage in designing a faster and more successful vaccine. The motivation of this study is to propose a successful hybrid machine learning method for SARS-CoV-2 epitope prediction and to identify nonallergen, nontoxic, antigen peptides that can be used in vaccine design from the predicted epitopes with bioinformatics tools. The identified epitopes will be effective not only in the design of the COVID-19 vaccine but also against viruses from the SARS family that may be encountered in the future. For this purpose, epitope prediction performances of random forest, support vector machine, logistic regression, bagging with decision tree, k-nearest neighbor and decision tree methods were examined. In the SARS-CoV and B-cell datasets used for education in the study, epitope estimation was performed again after the datasets were balanced with the synthetic minority oversampling technique (SMOTE) method since the epitope class samples were in the minority compared to the nonepitope class. The experimental results obtained were compared and the most successful predictions were obtained with the random forest (RF) method. The epitope prediction performance in balanced datasets was found to be higher than that in the original datasets (94.0% AUC and 94.4% PRC for the SMOTE-SARS-CoV dataset; 95.6% AUC and 95.3% PRC for the SMOTE-B-cell dataset). In this study, 252 peptides out of 20312 peptides were determined to be epitopes with the SMOTE-RF-SVM hybrid method proposed for SARS-CoV-2 epitope prediction. Determined epitopes were analyzed with AllerTOP 2.0, VaxiJen 2.0 and ToxinPred tools, and allergic, nonantigen, and toxic epitopes were eliminated. As a result, 11 possible nonallergic, high antigen and nontoxic epitope candidates were proposed that could be used in protein-based COVID-19 vaccine design ("VGGNYNY", "VNFNFNGLTG", "RQIAPGQTGKI", "QIAPGQTGKIA", "SYECDIPIGAGI", "STFKCYGVSPTKL", "GVVFLHVTYVPAQ", "KNHTSPDVDLGDI", "NHTSPDVDLGDIS", "AGAAAYYVGYLQPR", "KKSTNLVKNKCVNF"). It is predicted that the few epitopes determined by machine learning-based in silico methods will help biotechnologists design fast and accurate vaccines by reducing the number of trials in the laboratory environment.

Hourly Power Consumption Forecasting Using RobustSTL and TCN

Power consumption forecasting is a crucial need for power management to achieve sustainable energy. The power demand is increasing over time, while the forecasting of power consumption possesses challenges with nonlinearity patterns and various noise in the datasets. To this end, this paper proposes the RobustSTL and temporal convolutional network (TCN) model to forecast hourly power consumption. Through the RobustSTL, instead of standard STL, this decomposition method can extract time series data despite containing dynamic patterns, various noise, and burstiness. The trend, seasonality, and remainder components obtained from the decomposition operation can enhance prediction accuracy by providing significant information from the dataset. These components are then used as input for the TCN model applying deep learning for forecasting. TCN employing dilated causal convolutions and residual blocks to extract long-term data patterns outperforms recurrent networks in time series forecasting studies. To assess the proposed model, this paper conducts a comparison experiment between the proposed model and counterpart models. The result shows that the proposed model can grasp the rules of historical time series data related to hourly power consumption. Our proposed model overcomes the counterpart schemes in MAPE, MAE, and RMSE metrics. Additionally, the proposed model obtains the best results in precision, recall, and F1-score values. The result also indicates that the predicted data can fit the pattern of the actual data.

Enhancing Front-Vehicle Detection in Large Vehicle Fleet Management

Transportation safety has been widely discussed for avoiding forward collisions. The broad concept of remote sensing can be applied to detect the front of vehicles without contact. The traditional Haar features use adjacent rectangular areas for many ordinary vehicle studies to detect the front vehicle images in practice. This paper focused on large vehicles using a front-installed digital video recorder (DVR) with a near-infrared (NIR) camera. The views of large and ordinary vehicles are different; thus, this study used a deep learning method to process progressive improvement in moving vehicle detection. This study proposed a You Only Look Once version 4 (YOLOv4) supplemented with the fence method, called YOLOv4(III), to enhance vehicle detection. This method had high detection accuracy and low false omission rates using the general DVR equipment, and it provided comparison results. There was no need to have a high specification front camera, and the proposed YOLOv4(III) was found to have competitive performance. YOLOv4(III) reduced false detection rates and had a more stable frame per second (FPS) performance than with Haar features. This improved detection method can give an alert for large vehicle drivers to avoid serious collisions, leading to a reduction in the waste of social resources.

Vectorized rooftop area data for 90 cities in China

Reliable information on building rooftops is crucial for utilizing limited urban space effectively. In recent decades, the demand for accurate and up-to-date data on the areas of rooftops on a large-scale is increasing. However, obtaining these data is challenging due to the limited capability of conventional computer vision methods and the high cost of 3D modeling involving aerial photogrammetry. In this study, a geospatial artificial intelligence framework is presented to obtain data for rooftops using high-resolution open-access remote sensing imagery. This framework is used to generate vectorized data for rooftops in 90 cities in China. The data was validated on test samples of 180 km² across different regions with spatial resolution, overall accuracy, and F1 score of 1 m, 97.95%, and 83.11%, respectively. In addition, the generated rooftop area conforms to the urban morphological characteristics and reflects urbanization level. These results demonstrate that the generated dataset can be used for data support and decision-making that can facilitate sustainable urban development effectively.

Automatic classification of nerve discharge rhythms based on sparse auto-encoder and time series feature

Background

Nerve discharge is the carrier of information transmission, which can reveal the basic rules of various nerve activities. Recognition of the nerve discharge rhythm is the key to correctly understand the dynamic behavior of the nervous system. The previous methods for the nerve discharge recognition almost depended on the traditional statistical features, and the nonlinear dynamical features of the discharge activity. The artificial extraction and the empirical judgment of the features were required for the recognition. Thus, these methods suffered from subjective factors and were not conducive to the identification of a large number of discharge rhythms.

Results

The ability of automatic feature extraction along with the development of the neural network has been greatly improved. In this paper, an effective discharge rhythm classification model based on sparse auto-encoder was proposed. The sparse auto-encoder was used to construct the feature learning network. The simulated discharge data from the Chay model and its variants were taken as the input of the network, and the fused features, including the network learning features, covariance and approximate entropy of nerve discharge, were classified by Softmax. The results showed that the accuracy of the classification on the testing data was 87.5%, which could provide more accurate classification results. Compared with other methods for the identification of nerve discharge types, this method could extract the characteristics of nerve discharge rhythm automatically without artificial design, and show a higher accuracy.

Conclusions

The sparse auto-encoder, even neural network has not been used to classify the basic nerve discharge from neither biological experiment data nor model simulation data. The automatic classification method of nerve discharge rhythm based on the sparse auto-encoder in this paper reduced the subjectivity and misjudgment of the artificial feature extraction, saved the time for the comparison with the traditional method, and improved the intelligence of the classification of discharge types. It could further help us to recognize and identify the nerve discharge activities in a new way.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04592-3.

Classification of fluorescent anisotropy decay based on the distance approach in the frequency domain

Frequency-domain (FD) fluorometry is a widely utilized tool to probe unique features of complex biological structures, which may serve medical diagnostic purposes. The conventional data analysis approaches used today to extract the fluorescence intensity or fluorescence anisotropy (FA) decay data suffer from several drawbacks and are inherently limited by the characteristics and complexity of the decay models. This paper presents the squared distance (D²) technique, which categorized samples based on the direct frequency response data (FRD) of the FA decay. As such, it improves the classification ability of the FD measurements of the FA decay as it avoids any distortion that results from the challenged translation into time domain data. This paper discusses the potential use of the D² approach to classify biological systems. Mathematical formulation of D² technique adjusted to the FRD of the FA decay is described. In addition, it validates the D² approach using 2 simulated data sets of 6 groups with similar widely and closely spaced FA decay data as well as in experimental data of 4 samples of a fluorophore-solvent (fluorescein-glycerol) system. In the simulations, the classification accuracy was above 95% for all 6 groups. In the experimental data, the classification accuracy was 100%. The D² approach can help classify samples whose FA decay data are difficult to extract making FA in the FD a realistic diagnostic tool. The D² approach offers an advanced method for sorting biological samples with differences beyond the practical temporal resolution limit in a reliable and efficient manner based on the FRD of their time-resolved fluorescence measurements thereby achieving better diagnostic quality in a shorter time.

Mapping Forest Restoration Probability and Driving Archetypes Using a Bayesian Belief Network and SOM: Towards Karst Ecological Restoration in Guizhou, China

To address ecological threats such as land degradation in the karst regions, several ecological restoration projects have been implemented for improved vegetation coverage. Forests are the most important types of vegetation. However, the evaluation of forest restoration is often uncertain, primarily owing to the complexity of the underlying factors and lack of information related to changes in forest coverage in the future. To address this issue, a systematic case study based on the Guizhou Province, China, was carried out. First, three archetypes of driving factors were recognized through the self-organizing maps (SOM) algorithm: the high-strength ecological archetype, marginal archetype, and high-strength archetype dominated by human influence. Then, the probability of forest restoration in the context of ecological restoration was predicted using Bayesian belief networks in an effort to decrease the uncertainty of evaluation. Results show that the overall probability of forest restoration in the study area ranged from 22.27 to 99.29%, which is quite high. The findings from regions with different landforms suggest that the forest restoration probabilities of karst regions in the grid and the regional scales were lower than in non-karst regions. However, this difference was insignificant mainly because the ecological restoration in the karst regions accelerated local forest restoration and decreased the ecological impact. The proposed method of driving-factor clustering based on restoration as well as the method of predicting restoration probability have a certain reference value for forest management and the layout of ecological restoration projects in the mid-latitude ecotone.

Identification of coumarin-based food additives using terahertz spectroscopy combined with manifold learning and improved support vector machine

The purpose of this paper is to use terahertz (THz) spectroscopy combined with manifold learning and improved support vector machine (SVM) model to identify the coumarin-based food additives. The 216 THz absorbance spectra (144 for calibration set and 72 for prediction set) of six coumarin-based food additives are measured by using THz time-domain spectroscopy (THz-TDS) in the range of 0.5-2.0 THz. The method (P-t-SNE) combined principal component analysis (PCA) with manifold learning t-distributed stochastic neighbor embedding (t-SNE) is used for feature extraction of the THz spectra. Then, an improved SVM using differential evolution (DE) to improve gray wolf optimization (GWO) to optimize parameters is proposed. Finally, the result shows that the prediction set accuracy of PCA-DEGWO-SVM, P-t-SNE-DEGWO-SVM, and P-t-SNE-GWO-SVM models are 97.22%, 98.61%, and 95.83%, respectively, indicating that the accuracy by P-t-SNE is increased by about 1.39% compared with that processed by PCA, and the accuracy by DEGWO is also increased by about 2.78% compared with that processed by GWO. In conclusion, the improved model (P-t-SNE-DEGWO-SVM) has the best identification effect, and it is proved to be an effective method to identify coumarin-based food additives. PRACTICAL APPLICATION: The method used in this paper can be applied in the field of food safety detection. When detecting coumarin-based food additives, the method proposed in this paper is more time-saving and efficient than traditional detection methods. Through some more tests and adjustments, it will be possible to achieve rapid and on-site identification of various food additives.

Study on Machine Learning Models for Building Resilience Evaluation in Mountainous Area: A Case Study of Banan District, Chongqing, China

‘Resilience’ is a new concept in the research and application of urban construction. From the perspective of building adaptability in a mountainous environment and maintaining safety performance over time, this paper innovatively proposes machine learning methods for evaluating the resilience of buildings in a mountainous area. Firstly, after considering the comprehensive effects of geographical and geological conditions, meteorological and hydrological factors, environmental factors and building factors, the database of building resilience evaluation models in a mountainous area is constructed. Then, machine learning methods such as random forest and support vector machine are used to complete model training and optimization. Finally, the test data are substituted into models, and the models’ effects are verified by the confusion matrix. The results show the following: (1) Twelve dominant impact factors are screened. (2) Through the screening of dominant factors, the models are comprehensively optimized. (3) The accuracy of the optimization models based on random forest and support vector machine are both 97.4%, and the F1 scores are greater than 94.4%. Resilience has important implications for risk prevention and the control of buildings in a mountainous environment.

Hierarchical Boosting Dual-Stage Feature Reduction Ensemble Model for Parkinson's Disease Speech Data

As a neurodegenerative disease, Parkinson's disease (PD) is hard to identify at the early stage, while using speech data to build a machine learning diagnosis model has proved effective in its early diagnosis. However, speech data show high degrees of redundancy, repetition, and unnecessary noise, which influence the accuracy of diagnosis results. Although feature reduction (FR) could alleviate this issue, the traditional FR is one-sided (traditional feature extraction could construct high-quality features without feature preference, while traditional feature selection could achieve feature preference but could not construct high-quality features). To address this issue, the Hierarchical Boosting Dual-Stage Feature Reduction Ensemble Model (HBD-SFREM) is proposed in this paper. The major contributions of HBD-SFREM are as follows: (1) The instance space of the deep hierarchy is built by an iterative deep extraction mechanism. (2) The manifold features extraction method embeds the nearest neighbor feature preference method to form the dual-stage feature reduction pair. (3) The dual-stage feature reduction pair is iteratively performed by the AdaBoost mechanism to obtain instances features with higher quality, thus achieving a substantial improvement in model recognition accuracy. (4) The deep hierarchy instance space is integrated into the original instance space to improve the generalization of the algorithm. Three PD speech datasets and a self-collected dataset are used to test HBD-SFREM in this paper. Compared with other FR algorithms and deep learning algorithms, the accuracy of HBD-SFREM in PD speech recognition is improved significantly and would not be affected by a small sample dataset. Thus, HBD-SFREM could give a reference for other related studies.

An Improved K-Means Algorithm Based on Evidence Distance

The main influencing factors of the clustering effect of the k-means algorithm are the selection of the initial clustering center and the distance measurement between the sample points. The traditional k-mean algorithm uses Euclidean distance to measure the distance between sample points, thus it suffers from low differentiation of attributes between sample points and is prone to local optimal solutions. For this feature, this paper proposes an improved k-means algorithm based on evidence distance. Firstly, the attribute values of sample points are modelled as the basic probability assignment (BPA) of sample points. Then, the traditional Euclidean distance is replaced by the evidence distance for measuring the distance between sample points, and finally k-means clustering is carried out using UCI data. Experimental comparisons are made with the traditional k-means algorithm, the k-means algorithm based on the aggregation distance parameter, and the Gaussian mixture model. The experimental results show that the improved k-means algorithm based on evidence distance proposed in this paper has a better clustering effect and the convergence of the algorithm is also better.

Monitoring particulate composition changes during the flocculation process using polarized light scattering

Monitoring the particulate composition changes during the flocculation process is still challenging for the research community. We use an experimental setup based on polarized light scattering to measure the polarization states of the scattered light of the individual particles. We build a classifier based on the support vector machine and feed it with the measured parameters. Results show that the classifier can effectively classify the particulate compositions, such as the sediment particles, flocculants, and flocs, which can be used to monitor the particulate composition changes during the flocculation process. Discussions on the intensity and polarization parameters find that the polarization parameters play a vital role in the classification of the particulate compositions in the flocculation suspensions. Additionally, the further analysis of the experimental data and the related simulations show that the degree of polarization can be an indicator of the flocculation process. We prove that the method based on polarized light scattering may be a potential in situ monitoring tool in the future for the study of the flocculation process.

Applying Bayesian Belief Networks to Assess Alpine Grassland Degradation Risks: A Case Study in Northwest Sichuan, China

Grasslands are crucial components of ecosystems. In recent years, owing to certain natural and socio-economic factors, alpine grassland ecosystems have experienced significant degradation. This study integrated the frequency ratio model (FR) and Bayesian belief networks (BBN) for grassland degradation risk assessment to mitigate several issues found in previous studies. Firstly, the identification of non-encroached degraded grasslands and shrub-encroached grasslands could help stakeholders more accurately understand the status of different types of alpine grassland degradation. In addition, the index discretization method based on the FR model can more accurately ascertain the relationship between grassland degradation and driving factors to improve the accuracy of results. On this basis, the application of BBN not only effectively expresses the complex causal relationships among various variables in the process of grassland degradation, but also solves the problem of identifying key factors and assessing grassland degradation risks under uncertain conditions caused by a lack of information. The obtained result showed that the accuracies based on the confusion matrix of the slope of NDVI change (NDVIs), shrub-encroached grasslands, and grassland degradation indicators in the BBN model were 85.27, 88.99, and 74.37%, respectively. The areas under the curve based on the ROC curve of NDVIs, shrub-encroached grasslands, and grassland degradation were 75.39% (P < 0.05), 66.57% (P < 0.05), and 66.11% (P < 0.05), respectively. Therefore, this model could be used to infer the probability of grassland degradation risk. The results obtained using the model showed that the area with a higher probability of degradation (P > 30%) was 2.22 million ha (15.94%), with 1.742 million ha (78.46%) based on NDVIs and 0.478 million ha (21.54%) based on shrub-encroached grasslands. Moreover, the higher probability of grassland degradation risk was mainly distributed in regions with lower vegetation coverage, lower temperatures, less potential evapotranspiration, and higher soil sand content. Our research can provide guidance for decision-makers when formulating scientific measures for alpine grassland restoration.

COVID-19 Diagnosis from Routine Blood Tests using Artificial Intelligence Techniques

Coronavirus disease (COVID-19) is a unique worldwide pandemic. With new mutations of the virus with higher transmission rates, it is imperative to diagnose positive cases as quickly and accurately as possible. Therefore, a fast, accurate, and automatic system for COVID-19 diagnosis can be very useful for clinicians. In this study, seven machine learning and four deep learning models were presented to diagnose positive cases of COVID-19 from three routine laboratory blood tests datasets. Three correlation coefficient methods, i.e., Pearson, Spearman, and Kendall, were used to demonstrate the relevance among samples. A four-fold cross-validation method was used to train, validate, and test the proposed models. In all three datasets, the proposed deep neural network (DNN) model achieved the highest values of accuracy, precision, recall or sensitivity, specificity, F1-Score, AUC, and MCC. On average, accuracy 92.11%, specificity 84.56%, and AUC 92.20% values have been obtained in the first dataset. In the second dataset, on average, accuracy 93.16%, specificity 93.02%, and AUC 93.20% values have been obtained. Finally, in the third dataset, on average, the values of accuracy 92.5%, specificity 85%, and AUC 92.20% have been obtained. In this study, we used a statistical t-test to validate the results. Finally, using artificial intelligence interpretation methods, important and impactful features in the developed model were presented. The proposed DNN model can be used as a supplementary tool for diagnosing COVID-19, which can quickly provide clinicians with highly accurate diagnoses of positive cases in a timely manner.

A data-driven binary-classification framework for oil fingerprinting analysis

A marine oil spill is one of the most challenging environmental issues, resulting in severe long-term impacts on ecosystems and human society. Oil dispersants are widely applied as a treating agent in oil spill response operations. The usage of dispersants significantly changes the behaviors of dispersed oil and consequently challenges the oil fingerprinting analysis. In this study, machine learning was first introduced to analyze oil fingerprinting by developing a data-driven binary classification framework. The modeling integrated dimensionality reduction algorithms (e.g., principal component analysis, PCA) to distinguish. Five groups of biomarkers, including terpanes, steranes, triaromatic steranes (TA-steranes), monoaromatic steranes (MA-steranes), and diamantanes, were selected. Different feature spaces were created from the diagnostic index of biomarkers, and six ML algorithms were applied for comparative analysis and optimizing the modeling process, including k-nearest neighbor (KNN), support vector classifier (SVC), random forest classifier (RFC), decision tree classifier (DTC), logistic regression classifier (LRC), and ensemble vote classifier (EVC). Hyperparameter optimization and cross-validation through GridSearchCV were applied to prevent overfitting and increase the model accuracy. Model performance was evaluated by model score and F-score through confusion matrices. The results indicated that the RFC algorithm from the diamantanes dataset performed the best. It delivered the highest F-score (0.871) versus the lowest F-score (0.792) from the EVC algorithm from the TA-steranes dataset by PCA with a variance of 95%. Therefore, diamantanes were recommended as the most suitable biomarker for distinguishing WCO and CDO to aid oil fingerprinting under the conditions in this study. The results proved the proposed method as a potential analysis tool for oil spill source identification through ML-aided oil fingerprinting. The study also showed the value of ML methods in oil spill response research and practice.