Search

Over three-quarters of Taiwan's landmass consists of mountainous slopes with steep gradients, leading to frequent rockfall hazards that obstruct traffic and cause injuries and fatalities. This study used Google Alerts to compile internet news on rockfall incidents along Taiwan's highway system from April 2019 to February 2024. The locations of these rockfalls were geolocated using Google Earth and integrated with geographical, topographical, environmental, geological, and socioeconomic variables. Employing machine learning algorithms, particularly the Random Forest algorithm, we analyzed the potential for rockfall hazards along roadside slopes. The model achieved an overall accuracy of 0.8514 on the test dataset, with a sensitivity of 0.8378, correctly identifying 83.8% of rockfall locations. Shapley Additive Explanations (SHAP) analysis highlighted that factors such as slope angle and distance to geologically sensitive areas are pivotal in determining rockfall locations. The study underscores the utility of internet-based data collection in providing comprehensive coverage of Taiwan's highway system, and enabled the first broad analysis of rockfall hazard susceptibility for the entire highway network. The consistent importance of topographical and geographical features suggests that integrating detailed spatial data could further enhance predictive performance. The combined use of Random Forest and SHAP analyses offers a robust framework for understanding and improving predictive models, aiding in the development of effective strategies for risk management and mitigation in rockfall-prone areas, ultimately contributing to safer and more reliable transportation networks in mountainous regions. [ABSTRACT FROM AUTHOR]

This study used artificial intelligence techniques to identify clinical cancer biomarkers for recurrent gastric cancer survivors. From a hospital-based cancer registry database in Taiwan, the datasets of the incidence of recurrence and clinical risk features were included in 2476 gastric cancer survivors. We benchmarked Random Forest using MLP, C4.5, AdaBoost, and Bagging algorithms on metrics and leveraged the synthetic minority oversampling technique (SMOTE) for imbalanced dataset issues, cost-sensitive learning for risk assessment, and SHapley Additive exPlanations (SHAPs) for feature importance analysis in this study. Our proposed Random Forest outperformed the other models with an accuracy of 87.9%, a recall rate of 90.5%, an accuracy rate of 86%, and an F1 of 88.2% on the recurrent category by a 10-fold cross-validation in a balanced dataset. We identified clinical features of recurrent gastric cancer, which are the top five features, stage, number of regional lymph node involvement, Helicobacter pylori, BMI (body mass index), and gender; these features significantly affect the prediction model's output and are worth paying attention to in the following causal effect analysis. Using an artificial intelligence model, the risk factors for recurrent gastric cancer could be identified and cost-effectively ranked according to their feature importance. In addition, they should be crucial clinical features to provide physicians with the knowledge to screen high-risk patients in gastric cancer survivors as well. [ABSTRACT FROM AUTHOR]

Due to the impact of the COVID-19 pandemic, people have reduced eating out, resulting in a severe drop in the revenue of the catering industry. Health risks have become a major factor affecting the revenue of this industry. Predicting the revenue of the catering industry during the COVID-19 pandemic will not only allow practitioners to adjust their business strategies, but also provide a reference for governments to formulate relief measures. To this end, this study proposes a fuzzy big data analytics approach in which random forests, recursive feature elimination, fuzzy c-means, and deep neural networks are jointly applied. First, random forests and recursive feature elimination are used to select the most influential factors. The data is then divided into clusters by fuzzy c-means. Subsequently, a deep neural network is built for each cluster to make predictions. The prediction results of individual clusters are then aggregated to improve prediction accuracy. The proposed methodology has been applied to forecast the revenue of the catering industry in Taiwan. The results of the experiment showed that the impact of new deaths on the revenue of the catering industry was far greater than the number of newly diagnosed COVID-19 cases. [ABSTRACT FROM AUTHOR]

Background: Dengue fever is a well-studied vector-borne disease in tropical and subtropical areas of the world. Several methods for predicting the occurrence of dengue fever in Taiwan have been proposed. However, to the best of our knowledge, no study has investigated the relationship between air quality indices (AQIs) and dengue fever in Taiwan. Results: This study aimed to develop a dengue fever prediction model in which meteorological factors, a vector index, and AQIs were incorporated into different machine learning algorithms. A total of 805 meteorological records from 2013 to 2015 were collected from government open-source data after preprocessing. In addition to well-known dengue-related factors, we investigated the effects of novel variables, including particulate matter with an aerodynamic diameter < 10 µm (PM10), PM2.5, and an ultraviolet index, for predicting dengue fever occurrence. The collected dataset was randomly divided into an 80% training set and a 20% test set. The experimental results showed that the random forests achieved an area under the receiver operating characteristic curve of 0.9547 for the test set, which was the best compared with the other machine learning algorithms. In addition, the temperature was the most important factor in our variable importance analysis, and it showed a positive effect on dengue fever at < 30 °C but had less of an effect at > 30 °C. The AQIs were not as important as temperature, but one was selected in the process of filtering the variables and showed a certain influence on the final results. Conclusions: Our study is the first to demonstrate that AQI negatively affects dengue fever occurrence in Taiwan. The proposed prediction model can be used as an early warning system for public health to prevent dengue fever outbreaks. [ABSTRACT FROM AUTHOR]

Wind energy has become a trend in Brazil, particularly in the northeastern region of the country. Despite its advantages, wind power generation has been hindered by the high volatility of exogenous factors, such as weather, temperature, and air humidity, making long-term forecasting a highly challenging task. Another issue is the need for reliable solutions, especially for large-scale wind farms, as this involves integrating specific optimization tools and restricted-access datasets collected locally at the power plants. Therefore, in this paper, the problem of forecasting the energy generated at the Praia Formosa wind farm, an eco-friendly park located in the state of Ceará, Brazil, which produces around 7% of the state's electricity, was addressed. To proceed with our data-driven analysis, publicly available data were collected from multiple Brazilian official sources, combining them into a unified database to perform exploratory data analysis and predictive modeling. Specifically, three machine-learning-based approaches were applied: Extreme Gradient Boosting, Random Forest, and Long Short-Term Memory Network, as well as feature-engineering strategies to enhance the precision of the machine intelligence models, including creating artificial features and tuning the hyperparameters. Our findings revealed that all implemented models successfully captured the energy-generation trends, patterns, and seasonality from the complex wind data. However, it was found that the LSTM-based model consistently outperformed the others, achieving a promising global MAPE of 4.55%, highlighting its accuracy in long-term wind energy forecasting. Temperature, relative humidity, and wind speed were identified as the key factors influencing electricity production, with peak generation typically occurring from August to November. [ABSTRACT FROM AUTHOR]

Fall-risk assessment studies generally focus on identifying characteristics that affect postural balance in a specific group of subjects. However, falls affect a multitude of individuals. Among the groups with the most recurrent fallers are the community-dwelling elderly and stroke survivors. Thus, this study focuses on identifying a set of features that can explain fall risk for these two groups of subjects. Sixty-five community dwelling elderly (forty-nine female, sixteen male) and thirty-five stroke-survivors (twenty-two male, thirteen male) participated in our study. With the use of an inertial sensor, some features are extracted from the acceleration data of a Timed Up and Go (TUG) test performed by both groups of individuals. A short-form berg balance scale (SFBBS) score and the TUG test score were used for labeling the data. With the use of a 100-fold cross-validation approach, Relief-F and Extra Trees Classifier algorithms were used to extract sets of the top 5, 10, 15, 20, 25, and 30 features. Random Forest classifiers were trained for each set of features. The best models were selected, and the repeated features for each group of subjects were analyzed and discussed. The results show that only the stand duration was an important feature for the prediction of fall risk across all clinical tests and both groups of individuals. [ABSTRACT FROM AUTHOR]

Antimicrobial resistance (AMR) is a global health issue and surveillance of AMR can be useful for understanding AMR trends and planning intervention strategies. Salmonella, widely distributed in food-producing animals, has been considered the first priority for inclusion in the AMR surveillance program by the World Health Organization (WHO). Recent advances in rapid and affordable whole-genome sequencing (WGS) techniques lead to the emergence of WGS as a one-stop test to predict the antimicrobial susceptibility. Since the variation of sequencing and minimum inhibitory concentration (MIC) measurement methods could result in different results, this study aimed to develop WGS-based random forest models for predicting MIC values of 24 drugs using data generated from the same laboratories in Taiwan. The WGS data have been transformed as a feature vector of 10-mers for machine learning. Based on rigorous validation and independent tests, a good performance was obtained with an average mean absolute error (MAE) less than 1 for both validation and independent test. Feature selection was then applied to identify top-ranked 10-mers that can further improve the prediction performance. For surveillance purposes, the genome sequence-based machine learning methods could be utilized to monitor the difference between predicted and experimental MIC, where a large difference might be worthy of investigation on the emerging genomic determinants. [ABSTRACT FROM AUTHOR]

This study aims to develop a statistical method to generate high-resolution historical and future climate datasets for modeling plant distributions in mountainous area. Two climate datasets that were from Taiwan Climate Change Projection Information and Adaptation Knowledge Platform (TCCIP) and meteorological stations were used to construct two historical climate datasets with 50 x 50 m² spatial resolution, respectively. The two historical climate datasets presented similar temperature pattern but distinct precipitation patterns in northern Taiwan (NTWN). Random Forests (RF) had predicted similar distribution range of natural grassland along mountain ridge when RF were applied by the two climate datasets, whereas RF had predicted restricted distribution range when it was applied by true absence data. The two historical climate datasets were added to the relative changes of climate variables representing four future climate scenarios. RF method based on the future climate datasets predicted habitat loss of natural grassland at the mid and end of this century, regardless of climate datasets and four warming scenarios. Due to the altitudinal limits of NTWN, there is almost no chance for natural grassland to track their climatic requirements toward higher elevations under climate change. High-resolution historical and future climate datasets generated by the statistical method were useful for species distribution model to project species potential distribution range in mountainous area and were available to examine species range shifts under climate change. Model performances based on the high-resolution climate dataset may have better expressed the climatic requirements and exact climatic niches of species in mountainous areas. [ABSTRACT FROM AUTHOR]

Patients at risk of chronic kidney disease (CKD) must be identified early and precisely in order to prevent complications, save lives, and limit expenditures for patients and health systems. This study aimed to develop a simple, high-precision machine learning model to identify individuals at risk of developing CKD in the near future, using a novel metabolic index with or without creatinine. This retrospective cohort study used data from the MJ medical record database collected between 2001 and 2015 in Taiwan. We used Cox hazard regression to identify potential predictors, including the novel metabolic index, for use as variables in the models. To develop a machine learning-based CKD risk model with fewer variables, we performed several experimental analyses to combine interacting variables into subsets. Those subsets were used to train three models, random forest, logistic regression, and XGBoost, with or without adding creatinine. The study included 12,189 participants, 20% with and 80% without CKD. The most important conventional predictors of CKD are age and gender. The novel metabolic index, TyG-Index, TG/HDL-ratio and VAI, had stronger predictive power than the conventional risk factors. Without including creatinine data, the XGBoost provided the best predictive performance. After adding creatinine, the performance of all the models was excellent, outperforming both conventional indicators and existing clinical algorithms for CKD. Using novel metabolic index in machine learning-based CKD risk prediction can accurately identify individuals at risk of diagnosis with CKD in the next year, with or without including creatinine. [ABSTRACT FROM AUTHOR]

Obstructive sleep apnea (OSA) is a global health concern and is typically diagnosed using in-laboratory polysomnography (PSG). However, PSG is highly time-consuming and labor-intensive. We, therefore, developed machine learning models based on easily accessed anthropometric features to screen for the risk of moderate to severe and severe OSA. We enrolled 3503 patients from Taiwan and determined their PSG parameters and anthropometric features. Subsequently, we compared the mean values among patients with different OSA severity and considered correlations among all participants. We developed models based on the following machine learning approaches: logistic regression, k-nearest neighbors, naïve Bayes, random forest (RF), support vector machine, and XGBoost. Collected data were first independently split into two data sets (training and validation: 80%; testing: 20%). Thereafter, we adopted the model with the highest accuracy in the training and validation stage to predict the testing set. We explored the importance of each feature in the OSA risk screening by calculating the Shapley values of each input variable. The RF model achieved the highest accuracy for moderate to severe (84.74%) and severe (72.61%) OSA. The level of visceral fat was found to be a predominant feature in the risk screening models of OSA with the aforementioned levels of severity. Our machine learning models can be employed to screen for OSA risk in the populations in Taiwan and in those with similar craniofacial structures. [ABSTRACT FROM AUTHOR]

Defect pattern detection and classification are challenging for thin-film-transistor liquid-crystal display (TFT-LCD) manufacturing. Limitations of the existing solutions for automatic optical inspection can be traced in part to the lack of a framework within which different existing and new defect patterns can be analyzed, while integrating domain knowledge and effective technologies. This study aims to develop a framework for image-based defect classification that employs the convolution neural networks without using complex and time-consuming image-processing processes in advance. An empirical study was conducted in a leading TFT-LCD manufacturing in Taiwan for validation. The results have shown that the defect patterns can be effectively classified by the proposed convolutional neural networks that outperform the existing approaches such as Support Vector Machine and Random Forest. The developed solution is implemented to effectively support the engineers. [ABSTRACT FROM AUTHOR]

The steady and healthy development of strategic emerging industries (SEI) can enable the Republic of China to smoothly transform its economic progression mode and realize the upgradation and optimization of its industrial organization. In this way, China can take the initiative in the process of globalization. The strategic emerging industry economy has long argued for the influence of resilient firm dynamic aptitudes on innovative product growth. This study takes strategic emerging industrial economies as an important part and establishes a comprehensive evaluation index system for economic efficiency of SEI in the Republic of China from the perception of interorganizational interactions and entrepreneurial coordination features. We use the Random Forest (RF) algorithm to perform multifeature fusion and establish an economic efficiency measurement algorithm for SET through survey data from several Chinese companies. We find, in addition, that entrepreneurial coordination affects the firms' inclination and aptitude to exploit relationship profits and by this means significantly strengthening the effects of vertical interactions however weakening the effects of horizontal interactions. When compared with the BP and MLP approaches, the suggested approach has achieved 86.98% accuracy, while the other has 86.02% and 85.75%, respectively. [ABSTRACT FROM AUTHOR]

One of the most critical aspects of integrating renewable energy sources into the smart grid is photovoltaic (PV) power generation forecasting. This ensemble forecasting technique combines several forecasting models to increase the forecasting accuracy of the individual models. This study proposes a regression-based ensemble method for day-ahead PV power forecasting. The general framework consists of three steps: model training, creating the optimal set of weights, and testing the model. In step 1, a Random forest (RF) with different parameters is used for a single forecasting method. Five RF models (RF1, RF2, RF3, RF4, and RF5) and a support vector machine (SVM) for classification are established. The hyperparameters for the regression-based method involve learners (linear regression (LR) or support vector regression (SVR)), regularization (least absolute shrinkage and selection operator (LASSO) or Ridge), and a penalty coefficient for regularization (λ). Bayesian optimization is performed to find the optimal value of these three hyperparameters based on the minimum function. The optimal set of weights is obtained in step 2 and each set of weights contains five weight coefficients and a bias. In the final step, the weather forecasting data for the target day is used as input for the five RF models and the average daily weather forecasting data is also used as input for the SVM classification model. The SVM output selects the weather conditions, and the corresponding set of weight coefficients from step 2 is combined with the output from each RF model to obtain the final forecasting results. The stacking recurrent neural network (RNN) is used as a benchmark ensemble method for comparison. Historical PV power data for a PV site in Zhangbin Industrial Area, Taiwan, with a 2000 kWp capacity is used to test the methodology. The results for the single best RF model, the stacking RNN, and the proposed method are compared in terms of the mean relative error (MRE), the mean absolute error (MAE), and the coefficient of determination (R2) to verify the proposed method. The results for the MRE show that the proposed method outperforms the best RF method by 20% and the benchmark method by 2%. [ABSTRACT FROM AUTHOR]

Liquefaction is one of the disasters caused due to earthquake. In 1999, Chi-Chi, Taiwan, earthquake is an example of liquefaction prone disasters induced due to Mw 7.6 earthquake. This becomes major cause for prediction of the liquefaction in the soil with respect to geotechnical property. In this paper, we have use Artificial Neural Networks (ANN) model based on Resilient Back propagation (Rprop), Decision tree model (DT) and classifier are C 4.5 and Random Forest is done for comparing the performance and evaluation of liquefaction potential based on the obtained field CPT data (Juang et al. [1]) consisting 125 datasets over the simplified procedures that are being traditionally use for the classification of liquefaction of the soil by different researchers. It is observe that Resilient Back propagation Algorithm prediction is 100% whereas C 4.5 algorithm and Random forest Algorithm are 97.6% and 98.4% accurate for the evaluation of seismic soil liquefaction potential. [ABSTRACT FROM AUTHOR]

Background: A brief gonadotropin-releasing hormone analogues (GnRHa) stimulation test which solely focused on LH 30-minute post-stimulation was considered to identify girls with central precocious puberty (CPP). However, it was tested using traditional statistical methods. With advanced computer science, we aimed to develop a machine learning-based diagnostic model that processed baseline CPP-related variables and a brief GnRHa stimulation test for CPP diagnosis. Methods: We recruited girls suspected of precocious puberty and underwent a GnRHa stimulation test at Children Hospital 2, Vietnam, and Cathay General Hospital, Taiwan. Clinical data, bone age measurement, and 30-min post-stimulation blood test were used to build up the predictive model. The candidate model was developed by different machine learning algorithms that were mainly evaluated by sensitivity, specificity, the area under the receiver operator characteristic curve (AUC), and F1-score in internal and external validation data to classify girls as CPP and non-CPP at different time-points (0-min, 30-min, 60-min, and 120-min post-stimulation). Results: Among the 614 girls diagnosed with PP, 524 (85.3%) had CPP. The random forest algorithm yielded the highest value of F1-score (0.976), specificity (0.893), positive predicted value (0.987), and relatively high value of AUC (0.972) that contributed to high probability to identify CPP. The performance metrics of the 30-min post-stimulation diagnostic model including sensitivity and specificity surpassed those of the 0-minute model (0-min) and were equivalent to those of the model obtained 60-min and 120-min post-stimulation. Hence, our machine learning-based model helps shorten the stimulation test to 30 minutes after GnRHa injection, in general, it requires 120 minutes for a completed GnRHa stimulation test. Conclusions: We developed a diagnostic model based on clinical features and a single sample 30-minute post-stimulation to identify CPP in girls that can reduce distress for children caused by multiple blood samplings. [ABSTRACT FROM AUTHOR]

The discretionary damage of mental suffering in fatal car accident cases in Taiwan is subjective, uncertain, and unpredictable; thus, plaintiffs, defendants, and their lawyers find it difficult to judge whether spending much of their money and time on the lawsuit is worthwhile and which legal factors judges will consider important and dominant when they are assessing the mental suffering damages. To address these problems, we propose k-nearest neighbor, classification and regression trees, and random forests as learning algorithms for regression to build optimal predictive models. In addition, we reveal the importance ranking of legal factors by permutation feature importance. The experimental results show that the random forest model outperformed the other models and achieved good performance, and "the mental suffering damages that plaintiff claims" and "the age of the victim" play important roles in assessments of mental suffering damages in fatal car accident cases in Taiwan. Therefore, litigants and their lawyers can predict the discretionary damages of mental suffering in advance and wisely decide whether they should litigate or not, and then they can focus on the crucial legal factors and develop the best litigation strategy. [ABSTRACT FROM AUTHOR]

Providing early predictions of academic performance is necessary for identifying at-risk students and subsequently providing them with timely intervention for critical factors affecting their academic performance. Although e-book systems are often used to provide students with teaching/learning materials in university courses, seldom has research made the early prediction based on their online reading behaviours by implementing machine learning classifiers. This study explored to what extent university students' academic achievement can be predicted, based on their reading behaviours in an e-book supported course, using the classifiers. It further investigated which of the features extracted from the reading logs influence the predictions. The participants were 100 first-year undergraduates enrolled in a compulsory course at a university in Taiwan. The results suggest that logistic regression Gaussian naïve Bayes, supports vector classification, decision trees, and random forests, and neural networks achieved moderate prediction performance with accuracy, precision, and recall metrics. Furthermore, the Bayes classifier identified almost all at-risk students. Additionally, student online reading behaviours affecting the prediction models included: turning pages, going back to previous pages and jumping to other pages, adding/deleting markers, and editing/removing memos. These behaviours were significantly positively correlated to academic achievement and should be encouraged during courses supported by e-books. [ABSTRACT FROM AUTHOR]

Air, an essential natural resource, has been compromised in terms of quality by economic activities. Considerable research has been devoted to predicting instances of poor air quality, but most studies are limited by insufficient longitudinal data, making it difficult to account for seasonal and other factors. Several prediction models have been developed using an 11-year dataset collected by Taiwan's Environmental Protection Administration (EPA). Machine learning methods, including adaptive boosting (AdaBoost), artificial neural network (ANN), random forest, stacking ensemble, and support vector machine (SVM), produce promising results for air quality index (AQI) level predictions. A series of experiments, using datasets for three different regions to obtain the best prediction performance from the stacking ensemble, AdaBoost, and random forest, found the stacking ensemble delivers consistently superior performance for R2 and RMSE, while AdaBoost provides best results for MAE. [ABSTRACT FROM AUTHOR]

Effective and thorough credit-risk management is a key factor for lending institutions, as significant financial losses can arise from the borrowers' default. Consequently, machine learning methods can measure and analyze credit risk objectively when at the same time they face increasingly attention. This study analyzes default payment data from a credit cards' portfolio containing some 30,000 clients from Taiwan with twenty-three attributes and with no missing information. We compare prediction accuracy of seven classification methods used, i.e. KNN, Logistic Regression, Naïve Bayes, Decision Trees, Random Forest, SVC, and Linear SVC. The results indicate that only few out of most of the typical variables used can adequately analyze default characteristics in terms of lending decisions. The results provide effective feedback to credit evaluators, lending institutions and business analysts for in-depth analysis. Also, they mention to the importance of the precautionary borrowing techniques to be used to better understand credit-card borrowers' behavior, along with specific accounting, historical and demographical characteristics. [ABSTRACT FROM AUTHOR]

Landslide sources and runout features of typical natural terrain landslides can be observed from a geotechnical perspective. Landslide sources are the major area of occurrences, whereas runout signatures reveal the subsequent phenomena caused by unstable gravity. Remotely sensed landslide detection generally includes runout areas, unless these results have been excluded manually through detailed comparison with stereo aerial photos and other auxiliary data. Areas detected using remotely sensed landslide detection can be referred to as "landslide-affected" areas. The runout areas should be separated from landslide-affected areas when upgrading landslide detections into a landslide inventory to avoid unreliable results caused by impure samples. A supervised data mining procedure was developed to separate landslide sources and runout areas based on four topographic attributes derived from a 10–m digital elevation model with a random forest algorithm and cost-sensitive analysis. This approach was compared with commonly used methods, namely support vector machine (SVM) and logistic regression (LR). The Typhoon Morakot event in the Laonong River watershed, southern Taiwan, was modeled. The developed models constructed using the limited training data sets could separate landslide source and runout signatures verified using the polygon and area constraint-based datasets. Furthermore, the performance of developed models outperformed SVM and LR algorithms, achieving over 80% overall accuracy, area under the curve of the receiver operating characteristic, user's accuracy, and producer's accuracy in most cases. The agreement of quantitative evaluations between the area sizes of inventory polygons for training and the predicted targets was also observed when applying the supervised modeling strategy. [ABSTRACT FROM AUTHOR]

Landslide-generated seismic waves (landslide-quakes), exhibiting distinctive waveforms and frequency characteristics, can be recorded by nearby seismometers. Implementing an automatic classifier for landslide-quakes could help provide objective and accurate initiation times of landslides with efficiency. This study collected and analyzed the time information of 214 landslide seismic records due to 33 documented landslide events, from the Broadband Array in Taiwan for Seismology (BATS). In addition, equal numbers of earthquake and noise signals were also incorporated. The 642 seismic signals and time information were carefully examined using the random forest algorithm to create an automatic landslide-quake classifier. By validating the signal attributes of the landslide, earthquake, and noise events, specifically in the time and frequency domains, it was shown that the proposed classifier can reach an accuracy (the proportion of all correctly classified events to the total number of events) of 91.3%. To further evaluate the applicability of the automatic classifier, landslide-quakes generated during the devastating Typhoon Morakot (2009) and Typhoon Soudelor (2015) were also verified, showing that the sensitivity of the classifier is higher than 98%. [ABSTRACT FROM AUTHOR]

Reservoirs are important for households and impact the national economy. This paper proposed a time-series forecasting model based on estimating a missing value followed by variable selection to forecast the reservoir’s water level. This study collected data from the Taiwan Shimen Reservoir as well as daily atmospheric data from 2008 to 2015. The two datasets are concatenated into an integrated dataset based on ordering of the data as a research dataset. The proposed time-series forecasting model summarily has three foci. First, this study uses five imputation methods to directly delete the missing value. Second, we identified the key variable via factor analysis and then deleted the unimportant variables sequentially via the variable selection method. Finally, the proposed model uses a Random Forest to build the forecasting model of the reservoir’s water level. This was done to compare with the listing method under the forecasting error. These experimental results indicate that the Random Forest forecasting model when applied to variable selection with full variables has better forecasting performance than the listing model. In addition, this experiment shows that the proposed variable selection can help determine five forecast methods used here to improve the forecasting capability. [ABSTRACT FROM AUTHOR]

In this study, long-term variations in the concentrations of PM 2.5 , water-soluble inorganic salts (WIS), and gaseous precursors measured by a roadside air quality monitoring station were investigated from 2017 to February 2021 to examine the formation mechanism of secondary inorganic PM 2.5. A new machine learning model using WIS data as input variables was further developed to predict PM 2.5 and nitrate concentrations for source tracing and effective control strategy development. The results showed that a reduction in the NO x concentration under VOC-limited O 3 formation regime could offset the consumption of OH and O 3 , causing an increase in secondary NO 3 − and PM 2.5 formation during fall and winter seasons. A good agreement was obtained between the predicted and measured PM 2.5 values, with R2, root mean square error (RMSE), and mean absolute error (MAE) values of 0.81, 6.81 μg/m3, and 5.10 μg/m3, respectively. The nitrate ([NO 3 −]) prediction model could predict ∼59% of the atmospheric nitrate concentration. The sensitivity analysis of the input variables in the present model further revealed that NO 3 − and VOC were two important pollutants dominating the variation trend of PM 2.5. It is recommended that decision makers should focus more on the reduction of VOC and O 3 to reduce secondary PM 2.5 formation during winter in central Taiwan. Real-time measurements of the chemical composition of PM 2.5 , taken as the regulatory air quality monitoring items are needed in the future. [Display omitted] • An increase in secondary NO 3 − and PM 2.5 formation under a VOC-limited O 3 formation regime. • The present random forest (RF) model can predict 81% of PM 2.5. • 59% of nitrate can be captured by the present RF model. • NO 3 − and VOC play the most important roles in the variation of PM 2.5 [ABSTRACT FROM AUTHOR]

Fine particulate matter (PM 2.5) is associated with various adverse health outcomes and poses serious concerns for public health. However, ground monitoring stations for PM 2.5 measurements are mostly installed in population-dense or urban areas. Thus, satellite retrieved aerosol optical depth (AOD) data, which provide spatial and temporal surrogates of exposure, have become an important tool for PM 2.5 estimates in a study area. In this study, we used AOD estimates of surface PM 2.5 together with meteorological and land use variables to estimate monthly PM 2.5 concentrations at a spatial resolution of 3 km2 over Taiwan Island from 2015 to 2019. An ensemble two-stage estimation procedure was proposed, with a generalized additive model (GAM) for temporal-trend removal in the first stage and a random forest model used to assess residual spatiotemporal variations in the second stage. We obtained a model-fitting R2 of 0.98 with a root mean square error (RMSE) of 1.40 μ g / m 3. The leave-one-out cross-validation (LOOCV) R2 with seasonal stratification was 0.82, and the RMSE was 3.85 μ g / m 3 , whereas the R2 and RMSE obtained by using the pure random forest approach produced R2 and RMSE values of 0.74 and 4.60 μ g / m 3 , respectively. The results indicated that the ensemble modeling approach had a higher predictive ability than the pure machine learning method and could provide reliable PM 2.5 estimates over the entire island, which has complex terrain in terms of land use and topography. [Display omitted] • An improved two-stage ensemble modeling compared to random forest approach. • Leave one out cross-validation R2 0.82 and root mean squared error 3.85 μ g/ m 3. • Higher correlation between ground PM2.5 concentrations and AOD estimate of PM2.5 • Long-term PM2.5 concentrations across Taiwan substantially declined from 2015 to 2019. [ABSTRACT FROM AUTHOR]

Chronic obstructive pulmonary disease (COPD) is one of the leading causes of mortality and contributes to high morbidity worldwide. Patients with COPD have a higher risk for acute respiratory failure, ventilator dependence, and mortality after hospitalization compared with the general population. Accurate and early risk detection will provide more information for early management and better decision making. This study aimed to build prediction models using patients' characteristics, laboratory data, and comorbidities for early detection of acute respiratory failure, ventilator dependence, and mortality in patients with COPD after hospitalization. We retrospectively collected the electronic medical records of 5061 patients with COPD in three hospitals of the Chi Mei Medical Group, Taiwan. After data cleaning, we built three prediction models for acute respiratory failure, ventilator dependence, and mortality using seven machine learning algorithms. Based on the AUC value, the best model for mortality was built by the XGBoost algorithm (AUC = 0.817), the best model for acute respiratory failure was built by random forest algorithm (AUC = 0.804), while the best model for ventilator dependence was built by LightGBM algorithm (AUC = 0.809). A web service application was implemented with the best models and integrated into the existing hospital information system for physician's trials and evaluations. Our machine learning models exhibit excellent predictive quality and can therefore provide physicians with a useful decision-making reference for the adverse prognosis of COPD patients. [ABSTRACT FROM AUTHOR]

With the unique rainfall patterns of typhoons, plum rains, and short-term heavy rainfalls, the frequent landslide and debris flow disasters have caused severe loss to people in Taiwan. In the studies of landslide susceptibility, the information of factors used for analysis was usually annual-based content, and it was assumed that the same elements from different years were independent between each year. However, the occurrence of landslides was usually not simply due to the changes within a year. Instead, landslides were triggered because the factors that affected the potential of landslides reached critical conditions after a cumulative change with time. Therefore, this study had well evaluated the influence of temporal characteristics and the ratios of antecedent landslide areas in the past five years in the landslide potential evaluation model. The analysis was conducted through the random forest (RF) algorithm. Additional rainfall events of 2017 were used to test the proposed model's performance to understand its practicality. The analysis results show that in the study area, the RF model had considerably acceptable performance. The results have also demonstrated that the antecedent landslide ratios in the past five years were essential to describe the significance of cumulative change with time when conducting potential landslide evaluation. [ABSTRACT FROM AUTHOR]

First-year students' learning performance has received much attention in educational practice and theory. Previous works used some variables, which should be obtained during the course or in the progress of the semester through questionnaire surveys and interviews, to build prediction models. These models cannot provide enough timely support for the poor performance students, caused by economic factors. Therefore, other variables are needed that allow us to reach prediction results earlier. This study attempts to use family background variables that can be obtained prior to the start of the semester to build learning performance prediction models of freshmen using random forest (RF), C5.0, CART, and multilayer perceptron (MLP) algorithms. The real sample of 2407 freshmen who enrolled in 12 departments of a Taiwan vocational university will be employed. The experimental results showed that CART outperforms C5.0, RF, and MLP algorithms. The most important features were mother's occupations, department, father's occupations, main source of living expenses, and admission status. The extracted knowledge rules are expected to be indicators for students' early performance prediction so that strategic intervention can be planned before students begin the semester. [ABSTRACT FROM AUTHOR]

This paper is the first of its kind to use machine learning algorithms in conjunction with a Land-use Regression (LUR) model for predicting the spatiotemporal variation of CO concentrations in Taiwan. We used daily CO concentration from 2000 to 2016 to develop model and data from 2017 to 2018 as external data to verify the model reliability. Location of temples was used as a predictor to account for Asian culturally specific sources. With the ability to capture nonlinear relationship between observations and predictions, three LUR-based machine learning algorithms were used to estimate CO concentrations, including deep neural network (DNN), random forest (RF), and extreme gradient boosting (XGBoost). The results showed that LUR-based machine-learning model (LUR-XGBoost) has the best computation efficiency and improved adjusted R2 from 0.69 to 0.85. Our studies demonstrate the ability of the LUR-based machine learning algorithms to estimate long-term spatiotemporal CO concentration variations in fine resolution. [Display omitted] • Long-term daily CO concentrations were estimated with LUR-machine learning models. • Land-use patterns were included in machine learning models by using land-use regression. • The most contributed predictors were identified by stepwise variable selection. • Explanatory power of daily CO concentration was increased from 0.69 to 0.85. • XGboost outperformed RF and DNN machine learning algorithms. [ABSTRACT FROM AUTHOR]

Time-series prediction of a river stage during typhoons or storms is essential for flood control or flood disaster prevention. Data-driven models using machine learning (ML) techniques have become an attractive and effective approach to modeling and analyzing river stage dynamics. However, relatively new ML techniques, such as the light gradient boosting machine regression (LGBMR), have rarely been applied to predict the river stage in a tidal river. In this study, data-driven ML models were developed under a multistep-ahead prediction framework and evaluated for river stage modeling. Four ML techniques, namely support vector regression (SVR), random forest regression (RFR), multilayer perceptron regression (MLPR), and LGBMR, were employed to establish data-driven ML models with Bayesian optimization. The models were applied to simulate river stage hydrographs of the tidal reach of the Lan-Yang River Basin in Northeastern Taiwan. Historical measurements of rainfall, river stages, and tidal levels were collected from 2004 to 2017 and used for training and validation of the four models. Four scenarios were used to investigate the effect of the combinations of input variables on river stage predictions. The results indicated that (1) the tidal level at a previous stage significantly affected the prediction results; (2) the LGBMR model achieves more favorable prediction performance than the SVR, RFR, and MLPR models; and (3) the LGBMR model could efficiently and accurately predict the 1–6-h river stage in the tidal river. This study provides an extensive and insightful comparison of four data-driven ML models for river stage forecasting that can be helpful for model selection and flood mitigation. [ABSTRACT FROM AUTHOR]

Cost–benefit analysis is widely used to elucidate the association between foraging group size and resource size. Despite advances in the development of theoretical frameworks, however, the empirical systems used for testing are hindered by the vagaries of field surveys and incomplete data. This study developed the three approaches to data imputation based on machine learning (ML) algorithms with the aim of rescuing valuable field data. Using 163 host spider webs (132 complete data and 31 incomplete data), our results indicated that the data imputation based on random forest algorithm outperformed classification and regression trees, the k-nearest neighbor, and other conventional approaches (Wilcoxon signed-rank test and correlation difference have p-value from < 0.001–0.030). We then used rescued data based on a natural system involving kleptoparasitic spiders from Taiwan and Vietnam (Argyrodes miniaceus, Theridiidae) to test the occurrence and group size of kleptoparasites in natural populations. Our partial least-squares path modelling (PLS-PM) results demonstrated that the size of the host web (T = 6.890, p = 0.000) is a significant feature affecting group size. The resource size (T = 2.590, p = 0.010) and the microclimate (T = 3.230, p = 0.001) are significant features affecting the presence of kleptoparasites. The test of conformation of group size distribution to the ideal free distribution (IFD) model revealed that predictions pertaining to per-capita resource size were underestimated (bootstrap resampling mean slopes

Landslides have been identified as one of the costliest and deadliest natural disasters, causing tremendous damage to humans and societies. Information regarding the spatial extent of landslides is thus important to allow officials to devise successful strategies to mitigate landslide hazards. This study aims to develop a machine-learning approach for predicting landslide areas in the Tsengwen River Watershed (TRW), which is one of the most landslide-prone areas in Central Taiwan. Various spatial datasets were collected from 2009 to 2015 to derive 36 predictive variables used for landslide modeling with random forests (RF). The results of landslide prediction, compared with ground reference data, indicated an overall accuracy of 91.4% and Kappa coefficient of 0.83, respectively. The findings achieved from estimates of predictor importance also indicated to officials that the land-use/land-cover (LULC) type, distance to previous landslides, distance to roads, bank erosion, annual groundwater recharge, geological line density, aspect, and slope are the most influential factors that trigger landslides in the study region. [ABSTRACT FROM AUTHOR]

Although machine learning has been extensively used in various fields, it has only recently been applied to soil erosion pin modeling. To improve upon previous methods of quantifying soil erosion based on erosion pin measurements, this study explored the possible application of ensemble machine learning algorithms to the Shihmen Reservoir watershed in northern Taiwan. Three categories of ensemble methods were considered in this study: (a) Bagging, (b) boosting, and (c) stacking. The bagging method in this study refers to bagged multivariate adaptive regression splines (bagged MARS) and random forest (RF), and the boosting method includes Cubist and gradient boosting machine (GBM). Finally, the stacking method is an ensemble method that uses a meta-model to combine the predictions of base models. This study used RF and GBM as the meta-models, decision tree, linear regression, artificial neural network, and support vector machine as the base models. The dataset used in this study was sampled using stratified random sampling to achieve a 70/30 split for the training and test data, and the process was repeated three times. The performance of six ensemble methods in three categories was analyzed based on the average of three attempts. It was found that GBM performed the best among the ensemble models with the lowest root-mean-square error (RMSE = 1.72 mm/year), the highest Nash-Sutcliffe efficiency (NSE = 0.54), and the highest index of agreement (d = 0.81). This result was confirmed by the spatial comparison of the absolute differences (errors) between model predictions and observations using GBM and RF in the study area. In summary, the results show that as a group, the bagging method and the boosting method performed equally well, and the stacking method was third for the erosion pin dataset considered in this study. [ABSTRACT FROM AUTHOR]

Prediction of functional outcome in ischemic stroke patients is useful for clinical decisions. Previous studies mostly elaborate on the prediction of favorable outcomes. Miserable outcomes, which are usually defined as modified Rankin Scale (mRS) 5–6, should be considered as well before further invasive intervention. By using a machine learning algorithm, we aimed to develop a multiclass classification model for outcome prediction in acute ischemic stroke patients requiring reperfusion therapy. This was a retrospective study performed at a stroke medical center in Taiwan. Patients with acute ischemic stroke who visited between January 2016 and December 2019 and who were candidates for reperfusion therapy were included. Clinical outcomes were classified as favorable outcome, intermediate outcome, and miserable outcome. We developed four different multiclass machine learning models (Logistic Regression, Supportive Vector Machine, Random Forest, and Extreme Gradient Boosting) to predict clinical outcomes and compared their performance to the DRAGON score. A sample of 590 patients was included in this study. Of them, 180 (30.5%) had favorable outcomes and 152 (25.8%) had miserable outcomes. All selected machine learning models outperformed the DRAGON score on accuracy of outcome prediction (Logistic Regression: 0.70, Supportive Vector Machine: 0.67, Random Forest: 0.69, and Extreme Gradient Boosting: 0.67, vs. DRAGON: 0.51, p < 0.001). Among all selected models, Logistic Regression also had a better performance than the DRAGON score on positive predictive value, sensitivity, and specificity. Compared with the DRAGON score, the multiclass machine learning approach showed better performance on the prediction of the 3-month functional outcome of acute ischemic stroke patients requiring reperfusion therapy. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic, which first spread to the People of Republic of China and then to other countries in a short time, affected the whole world by infecting millions of people and have been increasing its impact day by day. Hundreds of researchers in many countries are in search of a solution to end up this pandemic. This study aims to contribute to the literature by performing detailed analyses via a new three-staged framework constructed based on data envelopment analysis and machine learning algorithms to assess the performances of 142 countries against the COVID-19 outbreak. Particularly, clustering analyses were made using k-means and hierarchic clustering methods. Subsequently, efficiency analysis of countries were performed by a novel model, the weighted stochastic imprecise data envelopment analysis. Finally, parameters were analyzed with decision tree and random forest algorithms. Results have been analyzed in detail, and the classification of countries are determined by providing the most influential parameters. The analysis showed that the optimum number of clusters for 142 countries is three. In addition, while 20 countries out of 142 countries were fully effective, 36% of them were found to be effective at a rate of 90%. Finally, it has been observed that the data such as GDP, smoking rates, and the rate of diabetes patients do not affect the effectiveness level of the countries. • Efficiency analysis of 142 countries against COVID-19 was made by WSIDEA. • DEA-based model and machine learning algorithms are integrated. • Clustering analysis was done and differences between clusters were examined. • Effects of factors on number of infections, recovered, death cases were examined. • Most important factors affecting the efficiency levels of countries were examined. [ABSTRACT FROM AUTHOR]

Airborne particulate matter 2.5 (PM2.5) can have a profound effect on the health of the population. Many researchers have been reporting highly accurate numerical predictions based on raw PM2.5 data imported directly into deep learning models; however, there is still considerable room for improvement in terms of implementation costs due to heavy computational overhead. From the perspective of environmental science, PM2.5 values in a given location can be attributed to local sources as well as external sources. Local sources tend to have a dramatic short-term impact on PM2.5 values, whereas external sources tend to have more subtle but longer-lasting effects. In the presence of PM2.5 from both sources at the same time, this combination of effects can undermine the predictive accuracy of the model. This paper presents a novel combinational Hammerstein recurrent neural network (CHRNN) to enhance predictive accuracy and overcome the heavy computational and monetary burden imposed by deep learning models. The CHRNN comprises a based-neural network tasked with learning gradual (long-term) fluctuations in conjunction with add-on neural networks to deal with dramatic (short-term) fluctuations. The CHRNN can be coupled with a random forest model to determine the degree to which short-term effects influence long-term outcomes. We also developed novel feature selection and normalization methods to enhance prediction accuracy. Using real-world measurement data of air quality and PM2.5 datasets from Taiwan, the precision of the proposed system in the numerical prediction of PM2.5 levels was comparable to that of state-of-the-art deep learning models, such as deep recurrent neural networks and long short-term memory, despite far lower implementation costs and computational overhead. [ABSTRACT FROM AUTHOR]

Since 2001, cardiovascular disease (CVD) has had the second-highest mortality rate, about 15,700 people per year, in Taiwan. It has thus imposed a substantial burden on medical resources. This study was triggered by the following three factors. First, the CVD problem reflects an urgent issue. A high priority has been placed on long-term therapy and prevention to reduce the wastage of medical resources, particularly in developed countries. Second, from the perspective of preventive medicine, popular data-mining methods have been well learned and studied, with excellent performance in medical fields. Thus, identification of the risk factors of CVD using these popular techniques is a prime concern. Third, the Framingham risk score is a core indicator that can be used to establish an effective prediction model to accurately diagnose CVD. Thus, this study proposes an integrated predictive model to organize five notable classifiers: the rough set (RS), decision tree (DT), random forest (RF), multilayer perceptron (MLP), and support vector machine (SVM), with a novel use of the Framingham risk score for attribute selection (i.e., F-attributes first identified in this study) to determine the key features for identifying CVD. Verification experiments were conducted with three evaluation criteria—accuracy, sensitivity, and specificity—based on 1190 instances of a CVD dataset available from a Taiwan teaching hospital and 2019 examples from a public Framingham dataset. Given the empirical results, the SVM showed the best performance in terms of accuracy (99.67%), sensitivity (99.93%), and specificity (99.71%) in all F-attributes in the CVD dataset compared to the other listed classifiers. The RS showed the highest performance in terms of accuracy (85.11%), sensitivity (86.06%), and specificity (85.19%) in most of the F-attributes in the Framingham dataset. The above study results support novel evidence that no classifier or model is suitable for all practical datasets of medical applications. Thus, identifying an appropriate classifier to address specific medical data is important. Significantly, this study is novel in its calculation and identification of the use of key Framingham risk attributes integrated with the DT technique to produce entropy-based decision rules of knowledge sets, which has not been undertaken in previous research. This study conclusively yielded meaningful entropy-based knowledgeable rules in tree structures and contributed to the differentiation of classifiers from the two datasets with three useful research findings and three helpful management implications for subsequent medical research. In particular, these rules provide reasonable solutions to simplify processes of preventive medicine by standardizing the formats and codes used in medical data to address CVD problems. The specificity of these rules is thus significant compared to those of past research. [ABSTRACT FROM AUTHOR]

Forecasting the demand for container throughput is a critical indicator to measure the development level of a port in global business management and industrial development. Time-series analysis approaches are crucial techniques for forecasting the demand for container throughput. However, accurate demand forecasting for container throughput remains a challenge in time-series analysis approaches. In this study, we proposed a mixed-precision neural architecture to forecasting the demand for container throughput. This study is the first work to use a mixed-precision neural network to forecast the container throughput—the mixed-precision architecture used the convolutional neural network for learning the strength of the features and used long short-term memory to identify the crucial internal representation of time series depending on the strength of the features. The experiments on the demand for container throughput of the five ports in Taiwan were conducted to compare our deep learning architecture with other forecasting approaches. The results indicated that our mixed-precision neural architecture exhibited higher forecasting performance than classic machine learning approaches, including adaptive boosting, random forest regression, and support vector regression. The proposed architecture can effectively predict the demand for port container throughput and effectively reduce the costs of planning and development of ports in the future. [ABSTRACT FROM AUTHOR]

Accurately quantifying the size–density relationships is important to predict stand development, estimate stand carrying capacity and prescribe silvicultural treatments. Parametric methods, such as segmented regression, were proposed to estimate the complicated trajectory of size–density relationships. However, applying nonparametric methods to assess stand development has not been explicitly examined. In this study, we compared parametric and nonparametric methods for estimating size–density relationships for Japanese cedar plantations in Taiwan. Specifically, we compared the efficacy of two segmented regression models with the penalized spline and random forest for regression methods. We also examined various stages in stand development for old-growth Japanese cedar stands. Data collected from 237 Japanese cedar permanent plots were used in model fitting and validation. Results indicated that the parametric and nonparametric methods used in this study can provide reliable estimates of the size–density relationship for Japanese cedar. Higher accuracy was achieved before the stands diverged from the self-thinning line. The penalized spline approach behaved consistently well regardless of datasets or stages in stand development, while the predictability of the random forest algorithm slightly decreased when the validation data was fitted. The results of this study provide insights on the use of methods to quantify the size–density relationships as well as enhance the understanding of long-term stand development. [ABSTRACT FROM AUTHOR]