关键词： SARS-COV-2   Data Mining   Deep Learning   Extreme Gradient Boosting   Machine Learning  

摘要： Background and Objective: The new type of Coronavirus (2019-nCov) epidemic spread rapidly, causing more than 250 thousand deaths worldwide. The virus, which first appeared as a sign of pneumonia, was later called the SARS-COV-2 with Severe Acute Respiratory Syndrome by the World Health Organization. The SARS-COV-2 virus is triggered by binding to the Angiotensin-Converting Enzyme 2 (ACE 2) inhibitor, which is vital in cardiovascular diseases and the immune system, especially in conditions such as cerebrovascular, hypertension, and diabetes. This study aims to evaluate the prediction performance of death status based on the demographic/clinical factors (including COVID-19 severity) by data mining methods. Methods: The dataset consists of 1603 SARS-COV-2 patients and 13 variables obtained from an open source web address. The current dataset contains age, gender, chronic disease (hypertension, diabetes, renal, cardiovascular, etc.), some enzymes (ACE, angiotensin II receptor blockers), and COVID-19 severity, which are used to predict death status using deep learning and machine learning approaches (random forest, k-nearest neighbor, extreme gradient boosting [XGBoost]). A grid search algorithm tunes hyperparameters of the models, and predictions are assessed through performance metrics. Steps of knowledge discovery in databases are applied to obtain the relevant information. Results: The accuracy rate of deep learning (97.15%) was more successful than the accuracy rate based on classical machine learning (92.15% for RF and 93.4% for k-NN), but the ensemble classifier XGBoost method gave the highest accuracy (99.7%). While COVID-19 severity and age calculated from XGBoost were the two most important factors associated with death status, the most determining variables for death status estimated from deep learning were COVID-19 severity and hypertension. Conclusions: The proposed model (XGBoost) achieved the best prediction of death status based on the factors as compared

关键词： Sea measurements   Acoustics   Acoustic measurements   Machine learning   Oceans   Wind speed   CatBoost   classification   machine learning   precipitation   random forest   regression   underwater acoustic measurements   wind  

摘要： One method of measuring precipitation and wind over the ocean is through analysis of the underwater ambient acoustics. In this study, the ambient ocean noises recorded by a passive aquatic listener (PAL) in the Mediterranean are used to compare the effectiveness of the machine learning techniques for measuring the wind speed and precipitation rate with the empirical methods from previous works. The data were collected over the timeframe of June 2011 to May 2012 and included two storms that caused severe coastal flooding in Genoa, Italy. A spar buoy at the surface above the PAL provided high-quality in situ measurements to act as the reference data for model training and validation. The results using the machine learning models show correlation coefficients of 0.95 between the acoustic data and wind speed and a reduction in unexplained variance by over a third from previous methods. For precipitation, CatBoost and random forest models are used to measure the total precipitation for 12 events using leave-one-event-out cross-validation, demonstrating mean errors of 28 and 34 and median errors of 18 and 17, respectively. The ability to measure wind and precipitation by applying machine learning on data from underwater acoustic recorders shows potential to help improve in situ measurements over oceans globally.

关键词： Southeast Asia   landslide   wildfire   susceptibility   ensemble machine learning  

摘要： Southeast Asia (SEA) is a region affected by landslide and wildfire;however, few studies on susceptibility modeling for the two hazards together have been conducted for this region, and the intersection and the uncertainty of the two hazards are rarely assessed. Thus, the intersection of landslide and wildfire susceptibility and the spatial uncertainty of the susceptibility maps were studied in this paper. Reliable landslide and wildfire susceptibility maps are necessary for disaster management and land use planning. This work used three advanced ensemble machine learning algorithms: RF (Random Forest), GBDT (Gradient Boosting Decision Tree) and AdaBoost (Adaptive Boosting) to assess the landslide and wildfire susceptibility for SEA. A geo-database was established with 2759 landslide locations, 1633 wildfire locations and 18 predictor variables in total. The performances of the models were assessed using the overall classification accuracy (ACC), Precision, the area under the ROC (receiver operating curve) (AUC) and confusion matrix values. The results showed RF performs superior in both landslide (ACC = 0.81, Precision = 0.78 and AUC= 0.89) and wildfire (ACC= 0.83, Precision = 0.83 and AUC = 0.91) susceptibility modeling, followed by GBDT and AdaBoost. The overall superiority of RF over other models indicates that it is potentially an efficient model for landslide and wildfire susceptibility mapping. The landslide and wildfire susceptibility were obtained using the RF model. This paper also conducted an overlay analysis of the two hazards. The uncertainty of the susceptibility was further assessed using the coefficient of variation (CV). Additionally, the distance to roads is relatively important in both landslide and wildfire susceptibility, which is the most important in landslides and the second most important in wildfires. The result of this paper is useful for mastering the whole situation of hazard susceptibility and proves that RF is a robust model in the ha

关键词： Acute contact toxicity   Applicability domain   Honey bees   Machine learning   Structural alerts  

摘要： In recent years, the decline of honey bees and the collapse of bee colonies have caught the attention of ecologists, and the use of pesticides is one of the main reasons for the decline. Therefore, ecological risk assessment of pesticides is essential and necessary. In silico tools, such as QSAR models can play an important role in predicting physicochemical and biological properties of chemicals. In this study, a total of 54 classification models were developed by combination of 6 machine learning methods along with 9 kinds of molecular fingerprints based on the experimental honey bees acute contact toxicity data (LD50) of 676 structurally diverse pesticides. The best model proposed was SVM algorithm combined with CDK extended fingerprint. The analysis of the applicability domain of the model successfully excluded some extreme molecules. Additionally, 9 structural alerts about honey bees acute contact toxicity were identified by information gain and substructure frequency analysis.

关键词： Online learning   Support vector machine   Genetic algorithm   Metallographic image  

摘要： The shape, size and distribution of the microstructure could certainly reveal mechanical properties. Therefore, it is important to segment the microstructure accurately. However, in view of the randomness in the metallographic detection period, a fixed segmentation algorithm with good performance in specific and known dataset may lead to wrong result when external condition changes. To solve the problem, we proposes an online learning pipeline to adjust the model adaptively by dynamic samples. First, a preprocessing method is deployed to smooth the uneven gray level. Next, local features and abstract features are extracted by a genetic algorithm-optimized boundary detector and a deep learning model, respectively. Then, online support vector machine is employed to update the model parameters under complex and changeable conditions in real time. Finally, a post-processing method is employed to get final result. A variety of experiments are presented to verify the effectiveness and the convergence of online algorithm. The experiment results indicate that the proposed pipeline can effectively extract local and abstract features, and the real-time updating model according to dynamic samples achieves state-of-the-art segmentation performance.

关键词： machine learning   artificial data   anomaly detection  

摘要： In the field of Cyber-Physical Systems (CPS), there is a large number of machine learning methods, and their intrinsic hyper-parameters are hugely varied. Since no agreed-on datasets for CPS exist, developers of new algorithms are forced to define their own benchmarks. This leads to a large number of algorithms each claiming benefits over other approaches but lacking a fair comparison. To tackle this problem, this paper defines a novel model for a generation process of data, similar to that found in CPS. The model is based on well-understood system theory and allows many datasets with different characteristics in terms of complexity to be generated. The data will pave the way for a comparison of selected machine learning methods in the exemplary field of unsupervised learning. Based on the synthetic CPS data, the data generation process is evaluated by analyzing the performance of the methods of the Self-Organizing Map, One-Class Support Vector Machine and Long Short-Term Memory Neural Net in anomaly detection.

关键词： Remote sensing   Geographic information system   Random forest classification   Artificial neural network   Logistic regression   Cellular automata   LU   LC prediction  

摘要： The land-use/land-cover (LU/LC) information can be extracted through continuous monitoring and observation of the global environment in the field of RS and GIS (remote sensing and geographic information system). With many inventions on satellite technologies, RS plays a crucial role throughout the world, and the researchers had shown their interest in finding the past, present, and future LU/LC information using the RS satellite data. In this research work, the non-forest- and forest-covered changes of Javadi Hills located in India were simulated and predicted using the hybrid machine learning models. The Markov chain-artificial neural network with cellular automata (MC-ANN-CA) and Markov chain-logistic regression with cellular automata (MC-LR-CA) were used and compared using the actual LU/LC maps of 2009, 2012, and 2015 along with the spatial variables (slope, aspect, hill shade, and distance road map). The results of the comparative analysis between the predicted and actual map of 2015 had shown a higher percentage of correctness in the MC-ANN-CA model for the spatial variables like slope, aspect, and distance road map. The LU/LC for 2021 and 2027 was predicted using the MC-ANN-CA model. By 2021, the forest-covered area will decrease by nearly - 0.38%, and the non-forest-covered area will increase by 0.79%. By 2027, forest-covered areas will decrease by - 0.52%, and non-forest-covered areas will increase by 1.06%, respectively, indicating the impacts of human and urbanization on LU/LC in Javadi Hills.

关键词： Short-term prediction   Airway congestion   Long short-term memory network   Graph convolutional network   Extreme gradient boosting  

摘要： This study presents a hybrid spatiotemporal convolutional long short-term memory network with tri-directional temporal features (SCLN-TTF) for airway congestion index prediction. The spatial features, temporal features and spatiotemporal features are taken as input variables, among which the temporal and spatiotemporal input datasets are categorized into forward time series datasets, backward time series datasets and datasets at the target time. The spatial features are extracted using the fast-approximate convolutions on graphs, and the temporal features are extracted using the long short-term memory network. The subsequent congestion index is the output variable. The model parameters are sequentially updated based on the recent collected data and the new predicting results. It is found that, in general, the proposed SCLN-TTF method achieves the most stable and satisfactory performance for all the selected airways. The models exhibit the best performance for the 60 min prediction time horizon in terms of MAPE, while the proposed SCLN-TTF shows superiority compared to the other benchmark methods particularly for the 10 min prediction. Then the graph convolutional network and extreme gradient boosting method are combined for high congestion index identification. The finding further confirms the superiority of the proposed method by considering both the spatial and temporal dependencies and integrating forward and backward features.

摘要： Soil cohesion (C) is one of the critical soil properties and is closely related to basic soil properties such as particle size distribution, pore size, and shear strength. Hence, it is mainly determined by experimental methods. However, the experimental methods are often time-consuming and costly. Therefore, developing an alternative approach based on machine learning (ML) techniques to solve this problem is highly recommended. In this study, machine learning models, namely, support vector machine (SVM), Gaussian regression process (GPR), and random forest (RF), were built based on a data set of 145 soil samples collected from the Da Nang-Quang Ngai expressway project, Vietnam. The database also includes six input parameters, that is, clay content, moisture content, liquid limit, plastic limit, specific gravity, and void ratio. The performance of the model was assessed by three statistical criteria, namely, the correlation coefficient (R), mean absolute error (MAE), and root mean square error (RMSE). The results demonstrated that the proposed RF model could accurately predict soil cohesion with high accuracy (R = 0.891) and low error (RMSE = 3.323 and MAE = 2.511), and its predictive capability is better than SVM and GPR. Therefore, the RF model can be used as a cost-effective approach in predicting soil cohesion forces used in the design and inspection of constructions.