Search

Objective: To develop and validate machine learning models for predicting coronary artery disease (CAD) within a Taiwanese cohort, with an emphasis on identifying significant predictors and comparing the performance of various models. Methods: This study involved a comprehensive analysis of clinical, demographic, and laboratory data from 8,495 subjects in Taiwan Biobank (TWB) after propensity score matching to address potential confounding factors. Key variables included age, gender, lipid profiles (T-CHO, HDL_C, LDL_C, TG), smoking and alcohol consumption habits, and renal and liver function markers. The performance of multiple machine learning models was evaluated. Results: The cohort comprised 1,699 individuals with CAD identified through self-reported questionnaires. Significant differences were observed between CAD and non-CAD individuals regarding demographics and clinical features. Notably, the Gradient Boosting model emerged as the most accurate, achieving an AUC of 0.846 (95% confidence interval [CI] 0.819–0.873), sensitivity of 0.776 (95% CI, 0.732–0.820), and specificity of 0.759 (95% CI, 0.736–0.782), respectively. The accuracy was 0.762 (95% CI, 0.742–0.782). Age was identified as the most influential predictor of CAD risk within the studied dataset. Conclusion: The Gradient Boosting machine learning model demonstrated superior performance in predicting CAD within the Taiwanese cohort, with age being a critical predictor. These findings underscore the potential of machine learning models in enhancing the prediction accuracy of CAD, thereby supporting early detection and targeted intervention strategies. Trial registration: Not applicable. [ABSTRACT FROM AUTHOR]

For the advancement of smart grids, solar power generation predictions have become an important research topic. In the case of using traditional modeling methods, excessive computational costs may be incurred and it is difficult for these methods to learn the multi-variable dependencies of the data. Therefore, in this paper, a deep learning model was used to combine convolutional neural networks and long short-term memory recurrent network predictions. This method enables hourly power generation one day into the future. Convolutional neural networks are used to extract the features of multiple time series, while long short-term memory neural networks predict multivariate outcomes simultaneously. In order to obtain more accurate prediction results, we performed feature selection on meteorological features and combined the selected weather features to train the prediction model. We further distinguished sunny- and rainy-day models according to the predicted daily rainfall conditions. In the experiment, it was shown that the method of combining meteorological features further reduced the error. Finally, taking into account the differences in climate conditions between the northern and southern regions of Taiwan, the experimental results of case studies involving multiple regions were evaluated to verify the proposed method. The results showed that training combined with selected meteorological features can be widely used in regions with different climates in Taiwan. [ABSTRACT FROM AUTHOR]

Huge waves caused by typhoons often induce severe disasters along coastal areas, making the effective prediction of typhoon-induced waves a crucial research issue for researchers. In recent years, the development of the Internet of Underwater Things (IoUT) has rapidly increased the prediction of oceanic environmental disasters. Past studies have utilized meteorological data and feedforward neural networks (e.g., BPNN) with static network structures to establish short lead time (e.g., 1 h) typhoon wave prediction models for the coast of Taiwan. However, sufficient lead time for prediction remains essential for preparedness, early warning, and response to minimize the loss of lives and properties during typhoons. The aim of this research is to construct a novel long lead time typhoon-induced wave prediction model using Long Short-Term Memory (LSTM), which incorporates a dynamic network structure. LSTM can capture long-term information through its recurrent structure and selectively retain necessary signals using memory gates. Compared to earlier studies, this method extends the prediction lead time and significantly improves the learning and generalization capability, thereby enhancing prediction accuracy markedly. [ABSTRACT FROM AUTHOR]

This study explores the potential of machine learning algorithms for earthquake prediction, utilizing fluid chemical anomaly data from hot springs. Six hot springs, located within an active fault zone along the southeastern coast of China, were carefully chosen as hydrochemical monitoring sites for an extended period of two and a half years. Using this data, a prediction model integrating six algorithms was developed to forecast M ≥ 5 earthquakes in Taiwan. The model's performance was validated against recorded earthquake events, and the factors influencing its predictive capability were analyzed. Our comprehensive analysis conclusively demonstrates the superiority of machine learning algorithms over traditional statistical methods for earthquake prediction. Additionally, including sampling time in the data sets significantly improves the model's predictive performance. However, it is important to note that the model's predictive performance varies across different hot spring and indicators type, highlighting the importance of identifying optimal indicators for specific scenarios. The model parameters, including the anomaly detection rate (P) and earthquake response time threshold (M), significantly impact the model's predictive capabilities. Therefore, adjustments are needed to optimize the model's performance for practical use. Despite limitations such as the inability to differentiate pre‐earthquake anomalies from post‐earthquake anomalies and pinpoint the precise location of earthquakes, this study successfully showcases the potential of machine learning algorithms in earthquake prediction, paving the way for further research and improved prediction methods. Plain Language Summary: This study explores the potential of utilizing machine learning algorithms for earthquake prediction based on hydrothermal fluid chemical anomaly data. An earthquake prediction model integrating six machine learning algorithms was developed using continuous hot spring hydrochemical monitoring data. The model's prediction performance was validated using actual earthquake event data, and various factors influencing its predictive capability were analyzed. The results demonstrate that the machine learning algorithm's prediction performance exceeds that of traditional methods, and the selection of hot springs, indicator types, and model parameters significantly impacts the model's prediction performance. This study successfully showcased the potential of machine learning algorithms in earthquake prediction, while also indicating new avenues for further research and improvement of prediction methods. Key Points: An earthquake prediction model based on machine learning using hydrochemical data from hot springs is proposed with sufficient accuracyThe earthquake prediction accuracy of the local outlier factor algorithm outperformed the traditional statistical methodThe predictive performance of the model varies across different hot springs, indicators type, and model parameters [ABSTRACT FROM AUTHOR]

To achieve successful prevention of fire incidents originating from human activities, it is imperative to possess a thorough understanding. This paper introduces a machine learning approach, specifically utilizing deep neural networks (DNN), to develop predictive models for fire occurrence in Keelung City, Taiwan. It investigates ten factors across demographic, architectural, and economic domains through spatial analysis and thematic maps generated from geographic information system data. These factors are then integrated as inputs for the DNN model. Through 50 iterations, performance indices including the coefficient of determination (R2), root mean square error (RMSE), variance accounted for (VAF), prediction interval (PI), mean absolute error (MAE), weighted index (WI), weighted mean absolute percentage error (WMAPE), Nash–Sutcliffe efficiency (NS), and the ratio of performance to deviation (RPD) are computed, with average values of 0.89, 7.30 × 10−2, 89.21, 1.63, 4.90 × 10−2, 0.97, 2.92 × 10−1, 0.88, and 4.84, respectively. The model's predictions, compared with historical data, demonstrate its efficacy. Additionally, this study explores the impact of various urban renewal strategies using the DNN model, highlighting the significant influence of economic factors on fire incidence. This underscores the importance of economic factors in mitigating fire incidents and emphasizes their consideration in urban renewal planning. [ABSTRACT FROM AUTHOR]

Objective: Chronic kidney disease (CKD) is one of the most common diseases worldwide. The increasing prevalence and incidence of CKD have contributed to the critical problem of high medical costs. Due to stressful environments, aircrew members may have a high risk of renal dysfunction. A better strategy to prevent CKD progression in Air Force personnel would be to diagnosis CKD at an early stage. Since few studies have been conducted in Taiwan to examine the long-term trends in early CKD in Air Force aircrew members, this study is highly important. We investigated the prevalence of CKD and established a predictive model of disease variation among aircrew members. Materials and Methods: In this retrospective study, we included all subjects who had received physical examinations at a military hospital from 2004 to 2010 and who could be tracked for four years. The Abbreviated Modification of Diet in Renal Disease Formula (aMDRD) was used to estimate the glomerular filtration rate (GFR) and was combined with the National Kidney Foundation/ Kidney Disease Outcomes Quality Initiative (NKF-K/DOQI) to identify CKD patients. Results: A total of 212 aircrew members were assessed. The results showed that the prevalence of CKD was 3.8%, 9.4%, 9.0%, and 9.4% in each of the four years. According to the logistic regression analysis, abnormal urobilinogen levels, ketones, and white blood cell (WBC) counts in urine and a positive urine occult blood test increased the risk of CKD. A positive urine occult blood test can be used to predict the future risk of CKD. Moreover, the generalized estimating equation (GEE) model showed that a greater risk of CKD with increased examination time, age and seniority had a negative effect. In conclusion, abnormal urobilinogen levels, ketones, and urine WBC counts in urine as well as a positive urine occult blood test might serve as independent predictors for CKD. Conclusion: In the future, we can focus not only on annual physical examinations but also on simple and accurate examinations, such as urine occult blood testing, to determine the risk of CKD and prevent its progression in our aircrew members. [ABSTRACT FROM AUTHOR]

Simple Summary: In patients with esophageal cancer bleeding, the presence of tumor ulcer and arterial–esophageal fistula (AEF) is a common occurrence. Notably, AEF is associated with an exceptionally poor prognosis, yet there is no prediction score to estimate its occurrence rate. Therefore, we introduce a novel model, the HEARTS-Score, for predicting AEF in esophageal cancer bleeding patients. This predictive model effectively distinguishes patients at risk, as evidenced by a c-statistic of 0.82 (95% CI 0.72–0.92). By employing this model, clinicians can more objectively differentiate between high-risk and low-risk patients, facilitating more efficient clinical decision-making, diagnostic planning, and subsequent treatment strategies. Purpose: To develop and internally validate a novel prediction score to predict the occurrence of arterial–esophageal fistula (AEF) in esophageal cancer bleeding. Methods: This retrospective cohort study enrolled patients with esophageal cancer bleeding in the emergency department. The primary outcome was the diagnosis of AEF. The patients were randomly divided into a derivation group and a validation group. In the derivation stage, a predictive model was developed using logistic regression analysis. Subsequently, internal validation of the model was conducted in the validation cohort during the validation stage to assess its discrimination ability. Results: A total of 257 patients were enrolled in this study. All participants were randomized to a derivation cohort (n = 155) and a validation cohort (n = 102). AEF occurred in 22 patients (14.2%) in the derivation group and 14 patients (13.7%) in the validation group. A predictive model (HEARTS-Score) comprising five variables (hematemesis, active bleeding, serum creatinine level >1.2 mg/dL, prothrombin time >13 s, and previous stent implantation) was established. The HEARTS-Score demonstrated a high discriminative ability in both the derivation and validation cohorts, with c-statistics of 0.90 (95% CI 0.82–0.98) and 0.82 (95% CI 0.72–0.92), respectively. Conclusions: By employing this novel prediction score, clinicians can make more objective risk assessments, optimizing diagnostic strategies and tailoring treatment approaches. [ABSTRACT FROM AUTHOR]

Simple Summary: In this retrospective study conducted at Taichung Veterans General Hospital, 54,692 participants were examined, including 385 with esophageal, oropharyngeal, or hypopharyngeal squamous cell carcinoma (SCC). Researchers investigated the correlation between cancer incidence and prognosis and a polygenic risk score (PRS) derived from 8353 single-nucleotide polymorphisms. Those with a high PRS (Q4) exhibited a 1.83-fold higher risk of SCCs compared to the low-PRS group (Q1), particularly notable in esophageal and hypopharyngeal cancers. Notably, PGS001087 displayed discernible associations with SCC risk, especially in specific subtypes and advanced stages, although not significantly linked to clinical staging. The study underscores the significance of genetic factors in upper aerodigestive tract cancers, notably esophageal SCC, offering insights for future research and risk assessment strategies. These findings highlight the importance of PRS in understanding cancer susceptibility, guiding targeted interventions, and informing personalized treatment approaches. Esophageal cancer shares strong associations with oropharyngeal and hypopharyngeal cancers, primarily due to shared risk factors like excessive tobacco and alcohol use. This retrospective study at Taichung Veterans General Hospital involved 54,692 participants, including 385 with squamous cell carcinoma (SCC) of the esophagus, oropharynx, or hypopharynx. Using a polygenic risk score (PRS) derived from 8353 single-nucleotide polymorphisms, researchers aimed to assess its correlation with cancer incidence and prognosis. The study found a 1.83-fold higher risk of esophageal, oropharyngeal, and hypopharyngeal SCCs in participants with a high PRS (Q4) compared to the low-PRS group (Q1). Esophageal cancer risk demonstrated a significant positive association with the PRS, as did hypopharyngeal cancer. Clinical parameters and staging showed limited associations with PRS quartiles, and the PRS did not significantly impact recurrence or mortality rates. The research highlighted that a higher PRS is linked to increased susceptibility to esophageal and hypopharyngeal cancer. Notably, a specific polygenic risk score, PGS001087, exhibited a discernible association with SCC risk, particularly in specific subtypes and advanced disease stages. However, it was not significantly linked to clinical cancer staging, emphasizing the multifactorial nature of cancer development. This hospital study reveals that a higher PRS correlates with increased susceptibility to esophageal and hypopharyngeal cancers. Notably, PGS001087 shows a discernible association with SCC risk in specific subtypes and advanced stages, although not significantly linked to clinical cancer staging. These findings enhance our understanding of genetic factors in upper aerodigestive tract cancers, particularly esophageal SCC, guiding future research and risk assessment strategies. [ABSTRACT FROM AUTHOR]

We aimed to explore the clinical outcomes and prognostic factors for PTCL-NOS patients in the real world. Clinical data were retrospectively collected from adult patients with PTCL-NOS treated at a single center in Taiwan. 104 PTCL-NOS patients with a median age of 53.0 years were enrolled. Patients with the International Prognostic Index (IPI) or prognostic index for peripheral T-cell lymphoma (PIT) scores of zero had a longer overall survival (OS) and progression free survival (PFS), while patients with IPI or PIT scores ≥1 did poorly. For patients who are eligible for transplantation, the use of pralatrexate as salvage chemotherapy has shown better OS (2-year OS 83.3% vs. 24.4%, P = 0.011) compared to patients who did not. By multivariate analysis, age >60 years, male, B symptoms, ECOG >1, lung involvement, and thrombocytopenia were independent adverse factors for OS. Incorporating factors in multivariate analysis, we established a novel predictive index for PTCL-NOS which efficiently stratifies patients into low (0–1 factor), intermediate-1 (2 factors), intermediate-2 (3 factors), and high risk (4–6 factors) groups with 2-year OS rates of 81.5%, 32.9%, 8.8%, and 0%, respectively (P < 0.001). PTCL-NOS patients have a dismal prognosis in Taiwan. Novel agents may improve the outcomes of PTCL-NOS patients. The usefulness of the novel prognostic index for PTCL-NOS needs further validation. [ABSTRACT FROM AUTHOR]

Chang, H.-K.; Chen, W.-W.; Cheng, C.-C.; Liou, J.-C., and Lin, S.-F., 2024. Assessment of tide model prediction and discrepancy in shallow waters of Taiwan to improve data fusion methods. Journal of Coastal Research, 40(1), 129–137. Charlotte (North Carolina), ISSN 0749-0208. Shallow bathymetry in the Taiwan Strait and corresponding complex hydrodynamics are key factors that make it difficult for hydrodynamic tide models to accurately calculate tide levels. Based on the Princeton Ocean Model with data assimilation from 2010 to 2014, the Ministry of Interior (MOI) developed a local tide model (MOI.18v1) for the waters around Taiwan that was imposed by harmonic analysis to have fast tide calculation at any time and at any location. This study first evaluated the prediction accuracy of the MOI.18v1 model using tide data at 22 gauge stations in Taiwan for 2016 and compared the results with those of NAO.99b. The MOI.18v1 root-mean-square error (RMSE) of residuals by subtraction with observations can reach about 13.49 to 18.26 cm, results lower than those of NAO.99b. A data fusion method is also proposed to further reduce the prediction discrepancy with RMSE of 3.67 to 6.35 cm, which means that the percentage of improvement increases the level of 25.85 to 47.08%. Data fusion can be used as a spatial correction for the predicted tide level using hydrodynamic tide models. [ABSTRACT FROM AUTHOR]

Construct-based models have become a mainstay of management and information systems research. However, these models are likely overfit to the data samples upon which they are estimated, making them risky to use in explanatory, prescriptive, or predictive ways outside a given sample. Empirical researchers currently lack tools to analyze why and how their models may not generalize out of sample. We propose a composite overfit analysis (COA) framework that applies predictive tools to describe the sources and ramifications of overfit in terms of the focal concepts important to empirical researchers: cases, constructs, and causal paths. The COA framework begins by using a leave-one-out crossvalidation procedure to identify cases with unusually high predictive error given their in-sample fit—a difference we describe as predictive deviance. The framework then employs a novel deviance tree method to group deviant cases that have similar predictive deviance and for similar theoretical reasons. We then employ a leave-deviant-group-out method, which sequentially analyzes how each deviant group affects model parameters, thereby identifying potentially unstable paths in the model. We can then infer descriptive reasons for why and how overfit affects a given model and data sample using the grouping criteria of the deviance tree, construct scores of deviant groups, and resulting unstable paths. These insights allow researchers to identify unexpected behavior that could define boundary conditions of their theory or point to new theoretical phenomena. We demonstrate the practical utility of our analytical framework on a technology adoption model in a new context. This paper was accepted by Dongjun Wu, information systems. Funding: This work was partially supported by the Ministry of Science and Technology, Taiwan [Grants 109-2811-H-007-503 and 108-2410-H-007-091-MY3] and by the 2021 Arts and Social Sciences Benefaction Fund of Trinity College Dublin, Ireland. Supplemental Material: The data files and online appendix are available at https://doi.org/10.1287/mnsc.2023.4705. [ABSTRACT FROM AUTHOR]

Background: Salary impacts nurse retention rates and thus is a factor affecting the nursing shortage both in Taiwan and around the world. Nurses in Taiwan earn a low salary compared with other health professionals and may be undervalued compared with their international counterparts. Purpose: This study was designed to analyze the factors associated with nurse salary (NS) in Organisation for Economic Co-operation and Development (OECD) countries and to compare NS in Taiwan with those in OECD member states. Methods: Data were extracted from the OECD statistics database and official statistics for Taiwan. For the 28 OECD member countries considered in this study and Taiwan, 21 indicators characterizing healthcare systems, including demographics, socioeconomic status, health behaviors and risks, healthcare resources, health financing, healthcare utilization, health outcomes, and economic inequality, were examined for the period of 2009-2018. A random-effects model (REM) and a fixed-effects model (FEM) were used to investigate the associations between these indicators and annual NS levels. The expected annual NS for Taiwan was estimated and compared with the actual NS for Taiwan using the REM. Results: In the REM, higher NS in OECD countries was shown to be positively associated with gross domestic product per capita (0.49, 95% confidence interval [CI] [0.41, 0.56]), proportion of population aged 65 years and over (2.72, 95% CI [2.17, 3.26]), crude birth rate (1.02, 95% CI [0.56, 1.49]), number of computerized tomography scanners per million population (0.26, 95% CI [0.17, 0.35]), alcohol consumption per person (0.94, 95% CI [0.26, 1.61]), and prevalence of obesity (0.64, 95% CI [0.40, 0.89]) and to be in inversely associated with infant mortality rate (-3.13, 95% CI [-3.94, -2.32]), bed density (-0.99, 95% CI [-1.72, -0.25]), number of hospital discharges (-0.08, 95% CI [-0.11, -0.05]), household out-of-pocket expenditure as a percentage of health expenditure (-0.34, 95% CI [-0.56, -0.11]), and the Gini coefficient (-0.25, 95% CI [-0.50, -0.01]). The FEM results were similar to those of the REM. The predicted annual NS for Taiwan based on the REM rose from 29,390 U.S. dollars (corrected for purchasing power parity; 95% CI [22,532, 36,247]) in 2009 to 49,891 U.S. dollars (95% CI [42,344, 57,438]) in 2018. The actual annual NS in Taiwan in 2018 was approximately 12% lower than the model-predicted value. Conclusions/Implications for Practice: Taiwan has a lower NS compared with its OECD counterparts. These findings may help policymakers, healthcare managers, and nurse organizations develop effective strategies to improve the remuneration system for nurses in Taiwan. [ABSTRACT FROM AUTHOR]

The strategies of the extrapolation adjusted by model prediction (ExAMP) blending scheme, which trusts the field pattern predicted by extrapolation and allows the field intensity to be adjusted by numerical weather prediction (NWP), for rainfall nowcasting are analyzed in this study. The McGill algorithm for precipitation nowcasting by Lagrangian extrapolation (MAPLE) and the Weather Research and Forecasting (WRF) model serve as the extrapolation and NWP models, respectively. Seven 150-min rainfall nowcasting experiments with different strategies are carried out for 37 sampled periods from seven heavy rainfall events in Taiwan in 2019. The results of the overall statistics indicate that, for the extrapolation component, extrapolating the current rainfall rate estimated from the lowest dual-polarimetric radar observations is a superior strategy. The ExAMP scheme that blends the MAPLE and WRF forecasts can surpass both components in 150-min rainfall nowcasting, and an empirical limitation on the innovation of intensity during the blending procedure is found unnecessary in this study. Moreover, the spatial performance for two contrasting events reveals the ability of ExAMP in grasping the rainfall strengthening and weakening in different areas. The skill statistics separately at rainfall strengthening gauges and weakening gauges further prove the effectiveness of ExAMP even though it is effective in intensity correction instead of pattern correction. [ABSTRACT FROM AUTHOR]

Background: Aging leads to changes in the body system, such as sarcopenia. This can result in several health issues, particularly physical and mobility dysfunction. Asian people typically have little awareness of sarcopenia. Thus, this study incorporated nursing instruction into the mobile application design to allow users to easily learn about sarcopenia. Objective: This study evaluated a model for predicting high-risk populations for sarcopenia in home settings. We further developed a sarcopenia nursing guidance mobile application and assessed the effectiveness of this application in influencing sarcopenia-related knowledge and self-care awareness among participants. Methods: Using a one-group pretest–posttest design, data were collected from 120 participants at a teaching hospital in northern Taiwan. This study used an artificial intelligence algorithm to evaluate a model for predicting high-risk populations for sarcopenia. We developed and assessed the sarcopenia nursing guidance mobile application using a questionnaire based on the Mobile Application Rating Scale. Results: The application developed in this study enhanced participants' sarcopenia-related knowledge and awareness regarding self-care. After the three-month intervention, the knowledge and awareness was effectively increase, total score was from 4.15 ± 2.35 to 6.65 ± 0.85 and were significant for all questionnaire items (p values < 0.05). On average, 96.1% of the participants were satisfied with the mobile app. The artificial intelligence algorithm positively evaluated the home-use model for predicting high-risk sarcopenia groups. Conclusions: The mobile application of the sarcopenia nursing guidance for public use in home settings may help alleviate sarcopenia symptoms and reduce complications by enhancing individuals' self-care awareness and ability. Trial registration: NCT05363033, registered on 02/05/2022. [ABSTRACT FROM AUTHOR]

Patient safety is a paramount concern in the medical field, and advancements in deep learning and Artificial Intelligence (AI) have opened up new possibilities for improving healthcare practices. While AI has shown promise in assisting doctors with early symptom detection from medical images, there is a critical need to prioritize patient safety by enhancing existing processes. To enhance patient safety, this study focuses on improving the medical operation process during X-ray examinations. In this study, we utilize EfficientNet for classifying the 49 categories of pre-X-ray images. To enhance the accuracy even further, we introduce two novel Neural Network architectures. The classification results are then compared with the doctor's order to ensure consistency and minimize discrepancies. To evaluate the effectiveness of the proposed models, a comprehensive dataset comprising 49 different categories and over 12,000 training and testing sheets was collected from Taichung Veterans General Hospital. The research demonstrates a significant improvement in accuracy, surpassing a 4% enhancement compared to previous studies. [ABSTRACT FROM AUTHOR]

Mammography is considered the gold standard for breast cancer screening. Multiple risk factors that affect breast cancer development have been identified; however, there is an ongoing debate regarding the significance of these factors. Machine learning (ML) models and Shapley Additive Explanation (SHAP) methodology can rank risk factors and provide explanatory model results. This study used ML algorithms with SHAP to analyze the risk factors between two different age groups and evaluate the impact of each factor in predicting positive mammography. The ML model was built using data from the risk factor questionnaires of women participating in a breast cancer screening program from 2017 to 2021. Three ML models, least absolute shrinkage and selection operator (lasso) logistic regression, extreme gradient boosting (XGBoost), and random forest (RF), were applied. RF generated the best performance. The SHAP values were then applied to the RF model for further analysis. The model identified age at menarche, education level, parity, breast self-examination, and BMI as the top five significant risk factors affecting mammography outcomes. The differences between age groups ranked by reproductive lifespan and BMI were higher in the younger and older age groups, respectively. The use of SHAP frameworks allows us to understand the relationships between risk factors and generate individualized risk factor rankings. This study provides avenues for further research and individualized medicine. [ABSTRACT FROM AUTHOR]

Simple Summary: Hepatocellular carcinoma (HCC) is a severe complication of chronic hepatitis B virus (HBV) infection. HCC can still develop in CHB patients undergoing antiviral therapy. Although several scoring systems for the prediction of HCC risk in CHB patients are available, very few of them are established for CHB patients receiving antiviral therapy. Our study developed a predictive scoring system for CHB patients on finite antiviral treatment, categorizing them into three risk groups based on a multivariate Cox proportional hazards model. Significant differences in HCC incidence were observed among these groups. PURPOSE: Hepatocellular carcinoma (HCC) is one of the most severe complications in chronic hepatitis B virus (HBV) infection. HCC can still develop in patients with chronic HBV (CHB) infection undergoing antiviral therapy. Several effective scoring systems for the prediction of HCC risk in CHB patients have been established. However, very few of them are designed for CHB patients receiving nucleos(t)ide analogues (NAs) therapy. Furthermore, none are available for HCC risk prediction in CHB patients receiving finite periods of antiviral therapy. METHODS: This study enrolled 790 consecutive treatment-naïve patients with CHB infection who had visited our liver clinics from 2008 to 2012 for pretreatment assessment before receiving antiviral therapies. The treatments were provided at finite periods according to the National Health Insurance Policy in Taiwan. The last follow-up date was 31 December 2021. We analyzed the virological and clinical factors in these 790 CHB patients receiving finite periods of NA treatments and identified the most significant risk factors for HCC to establish a novel predictive scoring system. By using stepwise selection in a multivariate Cox proportional hazards model, we divided the patients into three risk groups. RESULTS: Our predictive scoring system included five independent variables: genotype C (adjusted HR [aHR] = 2.23), NA-withdraw-related hepatitis relapse (aHR = 6.96), male (aHR = 4.19), liver cirrhosis (aHR = 11.14), and T1768A core promoter mutation (aHR = 3.21). This model revealed significant differences in HCC incidence among the three risk groups. The 5-year cumulative HCC risk significantly differed among the three risk groups (low risk: 1.33%, moderate risk: 4.99%, and high risk: 17.46%), with log-rank test p < 0.001. CONCLUSION: Our predictive scoring system is a promising tool for the prediction of HCC in CHB patients receiving finite NA treatments. Genotype C, NA-withdraw-related hepatitis relapse, male gender, liver cirrhosis, and the T1768A HBV core promoter mutation were significant independent risk factors. [ABSTRACT FROM AUTHOR]

Background: Survival is an important factor to consider when clinicians make treatment decisions for patients with skeletal metastasis. Several preoperative scoring systems (PSSs) have been developed to aid in survival prediction. Although we previously validated the Skeletal Oncology Research Group Machine‐learning Algorithm (SORG‐MLA) in Taiwanese patients of Han Chinese descent, the performance of other existing PSSs remains largely unknown outside their respective development cohorts. We aim to determine which PSS performs best in this unique population and provide a direct comparison between these models. Methods: We retrospectively included 356 patients undergoing surgical treatment for extremity metastasis at a tertiary center in Taiwan to validate and compare eight PSSs. Discrimination (c‐index), decision curve (DCA), calibration (ratio of observed:expected survivors), and overall performance (Brier score) analyses were conducted to evaluate these models' performance in our cohort. Results: The discriminatory ability of all PSSs declined in our Taiwanese cohort compared with their Western validations. SORG‐MLA is the only PSS that still demonstrated excellent discrimination (c‐indexes>0.8) in our patients. SORG‐MLA also brought the most net benefit across a wide range of risk probabilities on DCA with its 3‐month and 12‐month survival predictions. Conclusions: Clinicians should consider potential ethnogeographic variations of a PSS's performance when applying it onto their specific patient populations. Further international validation studies are needed to ensure that existing PSSs are generalizable and can be integrated into the shared treatment decision‐making process. As cancer treatment keeps advancing, researchers developing a new prediction model or refining an existing one could potentially improve their algorithm's performance by using data gathered from more recent patients that are reflective of the current state of cancer care. [ABSTRACT FROM AUTHOR]

Convenience store chains are many people's top choice for dining and leisure and have logistics procedures that involve each store receiving multiple deliveries because of the varying delivery periods and suitable temperatures for different goods. The estimated arrival time for each delivery has a huge impact on the route arrangement and convenience store preparation for dispatchers to schedule future deliveries. This study collected global positioning system travel data from a fleet of one of the top convenience store chains in Taiwan between April 2021 and March 2022 and proposed machine learning to establish a model to predict travel times. For unavailable data, we proposed the nonlinear regression equation to fill in the missing GPS data. Moreover, the study used the data between April 2022 and September 2022 with mean absolute percentage error to validate the prediction effects exceeding 97%. Therefore, the proposed model based on historical data and the machine learning algorithm in this study can help logistics fleets estimate accurate travel times for their scheduling of future delivery tasks and arranging routes. [ABSTRACT FROM AUTHOR]

Border management serves as a crucial control checkpoint for governments to regulate the quality and safety of imported food. In 2020, the first-generation ensemble learning prediction model (EL V.1) was introduced to Taiwan's border food management. This model primarily assesses the risk of imported food by combining five algorithms to determine whether quality sampling should be performed on imported food at the border. In this study, a second-generation ensemble learning prediction model (EL V.2) was developed based on seven algorithms to enhance the "detection rate of unqualified cases" and improve the robustness of the model. In this study, Elastic Net was used to select the characteristic risk factors. Two algorithms were used to construct the new model: The Bagging-Gradient Boosting Machine and Bagging-Elastic Net. In addition, Fβ was used to flexibly control the sampling rate, improving the predictive performance and robustness of the model. The chi-square test was employed to compare the efficacy of "pre-launch (2019) random sampling inspection" and "post-launch (2020–2022) model prediction sampling inspection". For cases recommended for inspection by the ensemble learning model and subsequently inspected, the unqualified rates were 5.10%, 6.36%, and 4.39% in 2020, 2021, and 2022, respectively, which were significantly higher (p < 0.001) compared with the random sampling rate of 2.09% in 2019. The prediction indices established by the confusion matrix were used to further evaluate the prediction effects of EL V.1 and EL V.2, and the EL V.2 model exhibited superior predictive performance compared with EL V.1, and both models outperformed random sampling. [ABSTRACT FROM AUTHOR]

Due to the increasing cost of health insurance, for decades, many countries have endeavored to constrain the cost of insurance by utilizing a DRG payment system. In most cases, under the DRG payment system, hospitals cannot exactly know which DRG code inpatients are until they are discharged. This paper focuses on the prediction of what DRG code appendectomy patients will be classified with when they are admitted to hospital. We utilize two models (or classifiers) constructed using the C4.5 algorithm and back-propagation neural network (BPN). We conducted experiments with the data collected from two hospitals. The results show that the accuracies of these two classification models can be up to 97.84% and 98.70%, respectively. According to the predicted DRG code, hospitals can effectively arrange medical resources with certainty, then, in turn, improve the quality of the medical care patients receive. [ABSTRACT FROM AUTHOR]

This study develops a process that helps admission committees of higher education institutions select interested and qualified students. This enables institutions to maintain their financial viability by reaching the quota given by the Education Administration of Taiwan. We aimed to predict the decision-making behavior of students in terms of enrollment. A logistic regression analysis was conducted on publicly and inexpensively accessible data; the selection criteria of the model are based on metrics from a confusion matrix comprising predicted and observed data. The results indicate a matching rate of close to 80% between the training data of a target university from 2018 to 2020 and the testing data from 2021. This system outputs a probability that the student will enroll and thus helps admission committees more effectively select students. [ABSTRACT FROM AUTHOR]

Background: Serum albumin level is a crucial nutritional indicator for patients on dialysis. Approximately one-third of patients on hemodialysis (HD) have protein malnutrition. Therefore, the serum albumin level of patients on HD is strongly correlated with mortality. Methods: In study, the data sets were obtained from the longitudinal electronic health records of the largest HD center in Taiwan from July 2011 to December 2015, included 1,567 new patients on HD who met the inclusion criteria. Multivariate logistic regression was performed to evaluate the association of clinical factors with low serum albumin, and the grasshopper optimization algorithm (GOA) was used for feature selection. The quantile g-computation method was used to calculate the weight ratio of each factor. Machine learning and deep learning (DL) methods were used to predict the low serum albumin. The area under the curve (AUC) and accuracy were calculated to determine the model performance. Results: Age, gender, hypertension, hemoglobin, iron, ferritin, sodium, potassium, calcium, creatinine, alkaline phosphatase, and triglyceride levels were significantly associated with low serum albumin. The AUC and accuracy of the GOA quantile g-computation weight model combined with the Bi-LSTM method were 98% and 95%, respectively. Conclusion: The GOA method was able to rapidly identify the optimal combination of factors associated with serum albumin in patients on HD, and the quantile g-computation with DL methods could determine the most effective GOA quantile g-computation weight prediction model. The serum albumin status of patients on HD can be predicted by the proposed model and accordingly provide patients with better a prognostic care and treatment. [ABSTRACT FROM AUTHOR]

Background: Targeted temperature management (TTM) is recommended for comatose out-of-hospital cardiac arrest (OHCA) survivors. Several prediction models have been proposed; however, most of these tools require data conversion and complex calculations. Early and easy predictive model of neurological prognosis in OHCA survivors with TTM warrant investigation.Materials and Methods: This multicenter retrospective cohort study enrolled 408 non-traumatic adult OHCA survivors with TTM from the TaIwan network of targeted temperature ManagEment for CARDiac arrest (TIMECARD) registry during January 2014 to June 2019. The primary outcome was unfavorable neurological outcome at discharge. The clinical variables associated with unfavorable neurological outcomes were identified and a risk prediction score-TIMECARD score was developed. The model was validated with data from National Taiwan University Hospital.Results: There were 319 (78.2%) patients presented unfavorable neurological outcomes at hospital discharge. Eight independent variables, including malignancy, no bystander cardiopulmonary resuscitation (CPR), non-shockable rhythm, call-to-start CPR duration >5 min, CPR duration >20 min, sodium bicarbonate use during resuscitation, Glasgow Coma Scale motor score of 1 at return of spontaneous circulation, and no emergent coronary angiography, revealed a significant correlation with unfavorable neurological prognosis in TTM-treated OHCA survivors. The TIMECARD score was established and demonstrated good discriminatory performance in the development cohort (area under the receiver operating characteristic curve [AUC] = 0.855) and validation cohorts (AUC = 0.918 and 0.877, respectively).Conclusion: In emergency settings, the TIMECARD score is a practical and simple-to-calculate tool for predicting neurological prognosis in OHCA survivors, and may help determine whether to initiate TTM in indicated patients. [ABSTRACT FROM AUTHOR]

The strategies of the extrapolation adjusted by model prediction (ExAMP) blending scheme, which trusts the field pattern predicted by extrapolation and allows the field intensity to be adjusted by numerical weather prediction (NWP), for rainfall nowcasting are analyzed in this study. The McGill algorithm for precipitation nowcasting by Lagrangian extrapolation (MAPLE) and the Weather Research and Forecasting (WRF) model serve as the extrapolation and NWP models, respectively. Seven 150-min rainfall nowcasting experiments with different strategies are carried out for 37 sampled periods from seven heavy rainfall events in Taiwan in 2019. The results of the overall statistics indicate that, for the extrapolation component, extrapolating the current rainfall rate estimated from the lowest dual-polarimetric radar observations is a superior strategy. The ExAMP scheme that blends the MAPLE and WRF forecasts can surpass both components in 150-min rainfall nowcasting, and an empirical limitation on the innovation of intensity during the blending procedure is found unnecessary in this study. Moreover, the spatial performance for two contrasting events reveals the ability of ExAMP in grasping the rainfall strengthening and weakening in different areas. The skill statistics separately at rainfall strengthening gauges and weakening gauges further prove the effectiveness of ExAMP even though it is effective in intensity correction instead of pattern correction. [ABSTRACT FROM AUTHOR]

With the rapid development of medicine and technology, machine learning (ML) techniques are extensively applied to medical informatics and the suboptimal health field to identify critical predictor variables and risk factors. Metabolic syndrome (MetS) and chronic kidney disease (CKD) are important risk factors for many comorbidities and complications. Existing studies that utilize different statistical or ML algorithms to perform CKD data analysis mostly analyze the early-stage subjects directly, but few studies have discussed the predictive models and important risk factors for the stage-III CKD high-risk health screening population. The middle stages 3a and 3b of CKD indicate moderate renal failure. This study aims to construct an effective hybrid important risk factor evaluation scheme for subjects with MetS and CKD stages III based on ML predictive models. The six well-known ML techniques, namely random forest (RF), logistic regression (LGR), multivariate adaptive regression splines (MARS), extreme gradient boosting (XGBoost), gradient boosting with categorical features support (CatBoost), and a light gradient boosting machine (LightGBM), were used in the proposed scheme. The data were sourced from the Taiwan health examination indicators and the questionnaire responses of 71,108 members between 2005 and 2017. In total, 375 stage 3a CKD and 50 CKD stage 3b CKD patients were enrolled, and 33 different variables were used to evaluate potential risk factors. Based on the results, the top five important variables, namely BUN, SBP, Right Intraocular Pressure (R-IOP), RBCs, and T-Cho/HDL-C (C/H), were identified as significant variables for evaluating the subjects with MetS and CKD stage 3a or 3b. [ABSTRACT FROM AUTHOR]

Simple Summary: Previous survival-prediction studies have had several limitations, such as a lack of comprehensive clinical data types, testing only in limited machine-learning algorithms, or a lack of a sufficient external testing set. This lung-cancer-survival-prediction model is based on multiple data types, multiple novel machine-learning algorithms, and external testing. This predicted model demonstrated a higher performance (ANN, AUC, 0.89; accuracy, 0.82; precision, 0.91) than previous similar studies. A well-established lung-cancer-survival-prediction model that relies on multiple data types, multiple novel machine-learning algorithms, and external testing is absent in the literature. This study aims to address this gap and determine the critical factors of lung cancer survival. We selected non-small-cell lung cancer patients from a retrospective dataset of the Taipei Medical University Clinical Research Database and Taiwan Cancer Registry between January 2008 and December 2018. All patients were monitored from the index date of cancer diagnosis until the event of death. Variables, including demographics, comorbidities, medications, laboratories, and patient gene tests, were used. Nine machine-learning algorithms with various modes were used. The performance of the algorithms was measured by the area under the receiver operating characteristic curve (AUC). In total, 3714 patients were included. The best performance of the artificial neural network (ANN) model was achieved when integrating all variables with the AUC, accuracy, precision, recall, and F1-score of 0.89, 0.82, 0.91, 0.75, and 0.65, respectively. The most important features were cancer stage, cancer size, age of diagnosis, smoking, drinking status, EGFR gene, and body mass index. Overall, the ANN model improved predictive performance when integrating different data types. [ABSTRACT FROM AUTHOR]

Background: Weaning from mechanical ventilation (MV) is an essential issue in critically ill patients, and we used an explainable machine learning (ML) approach to establish an extubation prediction model. Methods: We enrolled patients who were admitted to intensive care units during 2015–2019 at Taichung Veterans General Hospital, a referral hospital in central Taiwan. We used five ML models, including extreme gradient boosting (XGBoost), categorical boosting (CatBoost), light gradient boosting machine (LightGBM), random forest (RF) and logistic regression (LR), to establish the extubation prediction model, and the feature window as well as prediction window was 48 h and 24 h, respectively. We further employed feature importance, Shapley additive explanations (SHAP) plot, partial dependence plot (PDP) and local interpretable model-agnostic explanations (LIME) for interpretation of the model at the domain, feature, and individual levels. Results: We enrolled 5,940 patients and found the accuracy was comparable among XGBoost, LightGBM, CatBoost and RF, with the area under the receiver operating characteristic curve using XGBoost to predict extubation was 0.921. The calibration and decision curve analysis showed well applicability of models. We also used the SHAP summary plot and PDP plot to demonstrate discriminative points of six key features in predicting extubation. Moreover, we employed LIME and SHAP force plots to show predicted probabilities of extubation and the rationale of the prediction at the individual level. Conclusions: We developed an extubation prediction model with high accuracy and visualised explanations aligned with clinical workflow, and the model may serve as an autonomous screen tool for timely weaning. [ABSTRACT FROM AUTHOR]

Simple Summary: Epidemiological evidence has suggested that modifiable factors play an essential role in the risk of head and neck cancer (HNC). However, few studies have established HNC prediction models based on sex and tumor subsites. In this study, we establish sex- and subsite-specific HNC risk prediction models for the general Taiwanese population. Our study draws from a large sample size of 14,423 participants. The HNC prediction models may be applied in clinical risk assessments for health promotion programs. Epidemiological evidence has suggested that modifiable lifestyle factors play a significant role in the risk of head and neck cancer (HNC). However, few studies have established risk prediction models of HNC based on sex and tumor subsites. Therefore, we predicted HNC risk by creating a risk prediction model based on sex- and tumor subsites for the general Taiwanese population. This study adopted a case-control study design, including 2961 patients with HNC and 11,462 healthy controls. Multivariate logistic regression and nomograms were used to establish HNC risk prediction models, which were internally validated using bootstrap sampling. The multivariate logistic regression model indicated that age, education level, alcohol consumption, cigarette smoking, passive smoking, coffee consumption, and body mass index are common HNC predictors in both sexes, while the father's ethnicity, betel-nut-chewing habits, and tea consumption were male-specific HNC predictors. The risk factors of the prediction model for the HNC tumor subsite among men were the same as those for all patients with HNC. Additionally, the risks of alcohol consumption, cigarette smoking, and betel nut chewing varied, based on the tumor subsite. A c-index ranging from 0.93 to 0.98 indicated that all prediction models had excellent predictive ability. We developed several HNC risk prediction models that may be useful in health promotion programs. [ABSTRACT FROM AUTHOR]

Aims: Medication‐taking behaviours of breast cancer survivors undergoing adjuvant hormone therapy have received considerable attention. This study aimed to determine factors affecting medication‐taking behaviours in people with breast cancer using data mining. Design: A longitudinal observational retrospective cohort study with a hospital‐based survey. Methods: A total of 385 subjects were surveyed, analysing existing data from January 2010 to December 2017 in Taiwan. Three data mining approaches—multiple logistic regression, decision tree and artificial neural network—were used to build the prediction models and rank the importance of influencing factors. Accuracy, specificity and sensitivity were used as assessment indicators for the prediction models. Results: Multiple logistic regression was the most effective approach, achieving an accuracy of 96.37%, specificity of 96.75% and sensitivity of 96.12%. The duration of adjuvant hormone therapy discontinuation, duration of adjuvant hormone therapy use and age at diagnosis by data mining were the three most critical factors influencing the medication‐taking behaviours of people with breast cancer. [ABSTRACT FROM AUTHOR]

This study assessed the feasibility of five separate machine learning (ML) classifiers for predicting disease progression in patients with pre-dialysis chronic kidney disease (CKD). The study enrolled 858 patients with CKD treated at a veteran's hospital in Taiwan. After classification into early and advanced stages, patient demographics and laboratory data were processed and used to predict progression to renal failure and important features for optimal prediction were identified. The random forest (RF) classifier with synthetic minority over-sampling technique (SMOTE) had the best predictive performances among patients with early-stage CKD who progressed within 3 and 5 years and among patients with advanced-stage CKD who progressed within 1 and 3 years. Important features identified for predicting progression from early- and advanced-stage CKD were urine creatinine and serum creatinine levels, respectively. The RF classifier demonstrated the optimal performance, with an area under the receiver operating characteristic curve values of 0.96 for predicting progression within 5 years in patients with early-stage CKD and 0.97 for predicting progression within 1 year in patients with advanced-stage CKD. The proposed method resulted in the optimal prediction of CKD progression, especially within 1 year of advanced-stage CKD. These results will be useful for predicting prognosis among patients with CKD. [ABSTRACT FROM AUTHOR]

The progress of industrial development naturally leads to the demand for more electrical power. Unfortunately, due to the fear of the safety of nuclear power plants, many countries have relied on thermal power plants, which will cause more air pollutants during the process of coal burning. This phenomenon as well as increased vehicle emissions around us, have constituted the primary factors of serious air pollution. Inhaling too much particulate air pollution may lead to respiratory diseases and even death, especially PM $_{2.5}$ 2. 5 . By predicting the air pollutant concentration, people can take precautions to avoid overexposure to air pollutants. Consequently, accurate PM $_{2.5}$ 2. 5 prediction becomes more important. In this study, we propose a PM $_{2.5}$ 2. 5 prediction system, which utilizes the dataset from EdiGreen Airbox and Taiwan EPA. Autoencoder and Linear interpolation are adopted for solving the missing value problem. Spearman’s correlation coefficient is used to identify the most relevant features for PM $_{2.5}$ 2. 5 . Two prediction models (i.e., LSTM and LSTM based on K-means) are implemented which predict PM $_{2.5}$ 2. 5 value for each Airbox device. To assess the performance of the model prediction, the daily average error and the hourly average accuracy for the duration of a week are calculated. The experimental results show that LSTM based on K-means has the best performance among all methods. Therefore, LSTM based on K-means is chosen to provide real-time PM $_{2.5}$ 2. 5 prediction through the Linebot. [ABSTRACT FROM AUTHOR]

Background: Tuberculosis (TB) remained one of the world's most deadly chronic communicable diseases. Future TB incidence prediction is a benefit for intervention options and resource-allocation planning. We aimed to develop rapid univariate prediction models for epidemics forecasting employment. Methods: The surveillance data regarding Taiwan monthly TB incidence rates which from January 2005 to June 2017 were utilized for simulation modelling and from July 2017 to December 2020 for model validation. The modeling approaches including the Seasonal Autoregressive Integrated Moving Average (SARIMA), the Exponential Smoothing (ETS), and SARIMA-ETS hybrid algorithms were constructed and compared. The modeling performance of in-sample simulating training sets and pseudo-out-of-sample validating sets were evaluated by metrics of the root mean square error (RMSE), mean absolute percentage error (MAPE), mean absolute error (MAE), and mean absolute scaled error (MASE). Results: A total of 191,526 TB cases with a highest incidence rate in 2005 (72.5 per 100,000 person-year) and lowest in 2020 (33.2 per 100,000 person-year), from January-2005 to December-2020 showed a seasonality and steadily declining trend in Taiwan. The monthly incidence rates data were utilized to formulate these forecasting models. Through stepwise screening and assessing of the accuracy metrics, the optimized SARIMA(3,0,0)(2,1,0)12, ETS(A, A,A) and SARIMA-ETShybrid models were respectively selected as the candidate models. Regarding the outcome assessment of model performance, the SARIMA-ETS-hybrid model outperformed the ARIMA and ETS in the short term prediction with metrics of RMSE, MAE MAPE, and MASE of 0.084%, 0.067%, 0.646%, and 0.870%, during the pseudo-out-of-sample forecasting period. After projecting ahead to the long term forecasting TB incidence rates, ETS model showed the best performance resulting as a 41.69% (range: 22.1-56.38%) reduction of TB epidemics in 2025 and a 54.48% (range: 33.7-68.7%) reduction in 2030 compared with the 2015 levels. Conclusion: This time series modeling might offer us a rapid surveillance tool for facilitating WHO's future TB elimination milestone. Our proposed SARIMA-ETS or ETS model outperformed the SARIMA in predicting less or 12-30 months ahead of epidemics, and all models showed better in short or medium-term forecasting than long-term forecasting. [ABSTRACT FROM AUTHOR]

Internationally, the impact of venous leg ulcers (VLUs) on the quality of life is well recognised; however, in Taiwan, the focus is only on chronic wound management. This cross‐sectional correlational study conducted at the cardiovascular and plastic surgery clinics of a regional teaching hospital between August 2019 and June 2020 investigates venous clinical severity, pain, fatigue, depression, sleep quality, quality of life, and related factors among 167 patients with VLUs. The potential predictors of the quality of life in terms of activities were venous clinical severity (P < 0.001), pain (P = 0.004), and fatigue (P < 0.001) after adjusting for covariates. The potential predictors of the quality of life in terms of the psychological domain were marital status (single/divorced) (P = 0.016), marital status (widowed) (P = 0.027), venous clinical severity (P < 0.001), pain (P = 0.001), and fatigue (P = 0.002). The potential predictors of the quality of life with regard to symptoms were venous clinical severity (P < 0.001), pain (P < 0.001), fatigue (P = 0.001), and depression (P = 0.038). These potential predictors can serve as the basis of interventions for patients with VLUs, such as those related to nutrition or training in wound dressing. [ABSTRACT FROM AUTHOR]

Effectively handling the limited number of surgery operating rooms equipped with expensive equipment is a challenging task for hospital management such as reducing the case-time duration and reducing idle time. Improving the efficiency of operating room usage via reducing the idle time with better scheduling would rely on accurate estimation of surgery duration. Our model can achieve a good prediction result on surgery duration with a dozen of features. We have found the result of our best performing department-specific XGBoost model with the values 31.6 min, 18.71 min, 0.71, 28% and 27% for the metrics of root-mean-square error (RMSE), mean absolute error (MAE), coefficient of determination ( R 2 ), mean absolute percentage error (MAPE) and proportion of estimated result within 10% variation, respectively. We have presented each department-specific result with our estimated results between 5 and 10 min deviation would be more informative to the users in the real application. Our study shows comparable performance with previous studies, and the machine learning methods use fewer features that are better suited for universal usability. [ABSTRACT FROM AUTHOR]

The emergency department (ED) is at the forefront of medical care, and the medical team needs to make outright judgments and treatment decisions under time constraints. Thus, knowing how to make personalized and precise predictions is a very challenging task. With the advancement of artificial intelligence (AI) technology, Chi Mei Medical Center (CMMC) adopted AI, the Internet of Things (IoT), and interaction technologies to establish diverse prognosis prediction models for eight diseases based on the ED electronic medical records of three branch hospitals. CMMC integrated these predictive models to form a digital AI dashboard, showing the risk status of all ED patients diagnosed with any of these eight diseases. This study first explored the methodology of CMMC's AI development and proposed a four-tier AI dashboard architecture for ED implementation. The AI dashboard's ease of use, usefulness, and acceptance was also strongly affirmed by the ED medical staff. The ED AI dashboard is an effective tool in the implementation of real-time risk monitoring of patients in the ED and could improve the quality of care as a part of best practice. Based on the results of this study, it is suggested that healthcare institutions thoughtfully consider tailoring their ED dashboard designs to adapt to their unique workflows and environments. [ABSTRACT FROM AUTHOR]

Recently, annual agricultural data have been highly volatile as a result of climate change and national economic trends. Therefore, such data might not be enough to develop good agricultural policies for stabilizing agricultural output. A good agricultural output prediction model to assist agricultural policymaking has thus become essential. However, the highly volatile data would affect the prediction model's performance. For this reason, this study proposes a marriage in honey bees optimization/support vector regression (MBO/SVR) model to minimize the effects of highly volatile data (outliers) and enhance prediction accuracy. We verified the performance of the MBO/SVR model by using the annual total agricultural output collected from the official Agricultural Statistics Yearbook of the Council of Agriculture, Taiwan. Taiwan's annual total agricultural output integrates agricultural, livestock and poultry, fishery, and forest products. The results indicated that the MBO/SVR model had a lower mean absolute percentage error (MAPE), root mean square percentage error (RMSPE), and relative root mean squared error (r-RMSE) than those of the models it was compared to. Furthermore, the MBO/SVR model predicted long-term agricultural output more accurately and achieved higher directional symmetry (DS) than the other models. Accordingly, the MBO/SVR model is a robust, high-prediction-accuracy model for predicting long-term agricultural output to assist agricultural policymaking. [ABSTRACT FROM AUTHOR]

Introduction: The coronavirus disease 2019 (COVID-19) has brought excessive patients in emergency departments. Several COVID-19 prediction scores have been developed to aid in the patient disposition of emergency physicians. This study aimed to validate different COVID-19 prediction scores. Method: "DynaMed" was used to retrieve high-quality COVID-19 prediction scores for the evaluation of in-hospital mortality rate. SEIMC score, 4C-Mortality score, SOARS score, and Veterans Health Administration COVID-19 (VACO) Index were selected. A retrospective, single-center study was done on elderly patients hospitalized for COVID-19 from May 2021 to July 2021 in MacKay Memorial Hospital. Patients who were (I) negative for COVID-19 examination, (II) aged < 65 years old, (III) previously infected with COVID-19 and de-isolated (IV) hospital-acquired COVID-19 infection, (V) not admitted for hospitalization, and (VI) with missing of demographic characteristics were excluded. The area under the receiver operating characteristic curves (AUC) was computed to predict the in-hospital mortality rate. Result: Of 66,090 patients who underwent COVID-19 examination in MacKay Memorial Hospital, 133 patients were included in this study, with 26 deceased patients (19.5%). Among included patients, the median age was 74.38 years and 53% patients were male. Of the selected COVID-19 prediction scores, 4C-Mortality Score (AUC = 0.8), SEIMC score (AUC = 0.75), and SOARS score (AUC = 0.72) contained a good prognostic value, with an AUC > 0.70. VACOindex demonstrated less predictive value (AUC = 0.61). Conclusion: COVID-19 prediction scores were validated, and it was found that 4C-Mortality Score, SEIMC score, and SOARS score performed well in predicting the in-hospital mortality rate of elderly patients with COVID-19, and 4C-Mortality score is best appreciated. [ABSTRACT FROM AUTHOR]

In recent years, the development and application of artificial intelligence have both been topics of concern. In the medical field, an important direction of medical technology development is the extraction and use of applicable information from existing medical records to provide more accurate and helpful diagnosis suggestions. Therefore, this paper proposes using the development of diseases with easily discernible symptoms to predict the development of other medically related but distinct diseases that lack similar data. The aim of this study is to improve the ease of assessing the development of diseases in which symptoms are difficult to detect, and to improve the utilization of medical data. First, a time series model was used to capture the continuous manifestations of diseases with symptoms that could be easily found at different time intervals. Then, through transfer learning and attention mechanism, the general features captured were applied to the predictive model of the development of diseases with insufficient data and symptoms that are difficult to detect. Finally, we conducted a comprehensive experimental study based on a dataset collected from the National Health Insurance Research Database in Taiwan. The results demonstrate that the effectiveness of our transfer learning approach outperforms state-of-the-art deep learning prediction models for disease course prediction. [ABSTRACT FROM AUTHOR]

Purpose. Urinary tract infections (UTIs) are the most common infections among hospitalized patients. Cystoscopy is a minimally invasive procedure to check bladder disease, among the patients receiving procedure, approximately 10% of patients may experience UTI. In this study, a neural network model with high accuracy, sensitivity, and specificity was developed to predict the probability of UTIs caused by cystoscopic procedures. To reduce antibiotic overuse during cystoscopic procedures, the model can provide clinicians with a rapid assessment of whether patients require prophylactic antibiotics. Materials and Methods. Patients who underwent cystoscopic procedures at China Medical University Beigang Hospital from 2016 to 2019 were retrospectively reviewed. A total of 1647 patients were enrolled, and 147 cases of urinary tract infection occurred. An artificial neural network (ANN) and logistic regression analysis were used to develop the prediction models, and the two models were compared. Results. The logistic regression analysis model had an accuracy of 91%, sensitivity of 2%, and specificity of 99%, indicating that the logistic regression model predicted that most patients had a low risk of infection. The neural network model had a high accuracy of 85%, sensitivity of 80%, and specificity of 88%. Conclusions. Because the logistic regression model had low sensitivity and missed most cases of UTI, the logistic regression model is inappropriate for clinical application. The neural network model has superior predictive ability and can be considered a tool in clinical practice. [ABSTRACT FROM AUTHOR]

Background: Renal cell carcinoma (RCC) is the most common neoplasm in kidneys, and surgical resection remains the mainstay treatment. Few studies have investigated how the postoperative pain changes over time and what has affected its trajectory. This study aimed to characterize the variations in postoperative pain over time and investigate associated factors after RCC surgery. Methods: This retrospective study was conducted in a single medical center in Taiwan, where maximal pain scores in a numeric rating scale were recorded daily in the first five postoperative days (PODs) after RCC surgery. Latent curve models were developed, using two latent variables, intercept and slope, which represented the baseline pain and rate of pain resolution. These models explain the variations in postoperative pain scores over time. A predictive model for postoperative pain trajectories was also constructed. Results: There were 861 patients with 3850 pain observations included in the analysis. Latent curve analysis identified that female patients and those with advanced cancer (stage III and IV) tended to have increased baseline pain scores (p = 0.028 and 0.012, respectively). Furthermore, patients over 60 years, without PCA use (both p < 0.001), and with more surgical blood loss (p = 0.001) tended to have slower pain resolution. The final predictive model fit the collected data acceptably (RMSEA = 0.06, CFI = 0.95). Conclusion: Latent curve analysis identified influential factors of acute pain trajectories after RCC surgery. This study may also help elucidate the complex relationships between the variations in pain intensity over time and their determinants, and guide personalized pain management after surgery for RCC. [ABSTRACT FROM AUTHOR]

Forecasting of PM2.5 concentration is a global concern. Evidence has shown that the ambient PM2.5 concentrations are harmful to human health, climate change, plant species mortality, etc. PM2.5 concentrations are caused by natural and anthropogenic activities, and it is challenging to predict them due to many uncertain factors. Current research has focused on developing a new model while overlooking the fact that every single model for PM2.5 prediction has its own strengths and weaknesses. This paper proposes an ensemble framework which combines four diverse learning models for PM2.5 forecasting in Puli, Taiwan. It explores the synergy between parametric and non-parametric learning, and short-term and long-term learning. The feature set covers periodic, meteorological, and autoregression variables which are selected by a spiral validation process. The experimental dataset, spanning from 1 January 2008 to 31 December 2019, from Puli Township in central Taiwan, is used in this study. The experimental results show the proposed multi-model framework can synergize the advantages of the embedded models and obtain an improved forecasting result. Further, the benefit obtained by blending short-term learning with long-term learning is validated, in surpassing the performance obtained by using just single type of learning. Our multi-model framework compares favorably with deep-learning models on Puli dataset. It also shows high adaptivity, such that our multi-model framework is comparable to the leading methods for PM2.5 forecasting in Delhi, India. [ABSTRACT FROM AUTHOR]

Background and purpose -- Predicted survival may influence the treatment decision for patients with skeletal extremity metastasis, and PATHFx was designed to predict the likelihood of a patient dying in the next 24 months. However, the performance of prediction models could have ethnogeographical variations. We asked if PATHFx generalized well to our Taiwanese cohort consisting of 356 surgically treated patients with extremity metastasis. Patients and methods -- We included 356 patients who underwent surgery for skeletal extremity metastasis in a tertiary center in Taiwan between 2014 and 2019 to validate PATHFx's survival predictions at 6 different time points. Model performance was assessed by concordance index (c-index), calibration analysis, decision curve analysis (DCA), Brier score, and model consistency (MC). Results -- The c-indexes for the 1-, 3-, 6-, 12-, 18-, and 24-month survival estimations were 0.71, 0.66, 0.65, 0.69, 0.68, and 0.67, respectively. The calibration analysis demonstrated positive calibration intercepts for survival predictions at all 6 timepoints, indicating PATHFx tended to underestimate the actual survival. The Brier scores for the 6 models were all less than their respective null model's. DCA demonstrated that only the 6-, 12-, 18-, and 24-month predictions appeared useful for clinical decision-making across a wide range of threshold probabilities. The MC was < 0.9 when the 6- and 12-month models were compared with the 12-month and 18-month models, respectively. Interpretation -- In this Asian cohort, PATHFx's performance was not as encouraging as those of prior validation studies. Clinicians should be cognizant of the potential decline in validity of any tools designed using data outside their particular patient population. Developers of survival prediction tools such as PATHFx might refine their algorithms using data from diverse, contemporary patients that is more reflective of the world's population. [ABSTRACT FROM AUTHOR]

The Central Weather Bureau in Taiwan successfully implemented a version of WRF coined as TWRF (Typhoon WRF) as the operational typhoon prediction system. The TWRF has two nested domains with 15/3 km resolution covering large areas over the western North Pacific, which has led to significant improvements of typhoon predictions over the previous version with coarser resolutions. Built upon this success, the WRF Four-Dimensional Data Assimilation (FDDA) system has been implemented in TWRF. This study investigates the impact of assimilating dual-Doppler radar retrieval winds from eight sets of dual-Doppler radars using FDDA on the prediction of Typhoon Nesat (2017) that passed over Taiwan. The wind field retrieved with dual-Doppler radars has a vertical extent from 1 to 10 km and the horizontal resolution is 1 km. After quality control and data thinning, the radar retrieval winds are assimilated using FDDA for two update cycles in addition to the existing hybrid 3DEnVar for all other observations in TWRF. Furthermore, we assimilate the additional radar data that become available during the 4-hour waiting period for the completion of the global model prediction to be used as the lateral boundary condition for TWRF. For Typhoon Nesat, the wind structure and rainfall forecasts are improved with the assimilation of radar-retrieval winds. The overall improvements demonstrated by this case study suggest potentially high impacts for improving the prediction of typhoon-related rainfall with assimilation of dual-Doppler radar data. [ABSTRACT FROM AUTHOR]

The displacement of synchronous generators by converter interfaced energy renewable resource essentially reduces the system inertia, which can trigger under-frequency load shedding relay. To support the frequency containment, virtual inertia control (VIC) and fast frequency control (FFC) based on energy storage system (ESS) have been developed. Most of the existing VIC did not take into account the power constraints of the ESS. Although some of the FFC methods account for the ESS constraints, they ignore the intermediate steady-state (ISS) operating point of the primary frequency response (PFR). To counteract these problems, a model predictive control (MPC) based FFC, which considers the power limitations of the ESS, the frequency deviation constraint, and the ISS frequency tracking capability, is proposed in this paper. The proposed MPC only requires few local measurements, and the system parameters including inertia constants, onset time of inertia response and the ISS time of the PFR can be accurately estimated. To achieve the optimal control action and reduce computational burden, quadratic programming (QP) is incorporated in the proposed MPC. The proposed MPC is tested on a real Taiwan power system. The fast response and robustness of the proposed MPC has been justified by comparing with the state-of-the-art method. [ABSTRACT FROM AUTHOR]

Estimating short-term exposure to PM 2.5 has been achieved for population health studies using the land use regression with machine learning (LUR_ML) and microenvironmental exposure (ME) models. However, there is a lack of clarity regarding the performance of these models in predicting PM 2.5 exposure for individuals residing in diverse environments, and the factors influencing the variations in accuracy between these models. This study performed the LUR_ML and ME models to estimate daily exposure concentrations of PM 2.5 for elders residing in urban, suburban, rural, and industrial regions in Taiwan. The accuracy of the model predictions was assessed by comparing them with personal PM 2.5 monitoring for both overall and regional assessments. The LUR_ML model demonstrated reasonably moderate agreement (R2 = 0.516) overall with personal exposure to PM 2.5 , while the ME models exhibited relatively higher predictions (R2 = 0.535–0.575) and lower biases. The agreement of PM 2.5 predictions varies across regions, particularly in areas with higher exposure contrast. The ME model 1, utilizing region-specific microenvironmental measurements rather than generic data, highlights the potential for accurate prediction of personal PM 2.5 exposure. This study contributed to the understanding of variations in prediction accuracy across different regions and support the need for improved exposure models of air pollution. [Display omitted] • The land use regression with machine learning moderately agreed (R2 = 0.516) with PM 2.5 exposure. • Microenvironmental exposure models had a higher prediction (R2 = 0.535–0.575) and lower bias. • Exposure predictions of daily PM 2.5 concentrations vary by regions. • Indoor exposure is a stronger predictor of personal PM 2.5 exposure than outdoor exposure. • Higher PM 2.5 exposure agreement is likely due to greater exposure contrast in the population. [ABSTRACT FROM AUTHOR]

This paper presents a performance analysis of predictive models for the generator module which can be used as a reference for improvement in the condition monitoring system using wind turbines in a wind farm in Taiwan. With the generator being a critical component prone to failures, it is important to perform data analysis on its parameters that could be used for condition monitoring. The main innovative feature in this framework is the conduct of performance analysis before the development of the condition monitoring system. Also, the consistency of the performance between the different wind turbines in the wind farm is evaluated. The predictive models are generated using the neural network algorithm with a different combination of parameters from the SCADA system. The correlation of the parameters as well as the mean square error of the predictive models were then computed for analysis. Results showed that pairing of input parameters with a higher correlation to the output parameter would give better performance for the predictive model. Furthermore, the performance of the different models was consistent throughout the different wind turbines in the wind farm which indicates that the same model can be developed and used for wind turbines belonging to the same wind farm. Employing a preliminary performance analysis of different combinations of component parameters could help in optimizing predictive models for condition monitoring. [ABSTRACT FROM AUTHOR]

The deferral of graduation rate in Taiwan's universities is estimated 16%, which will affect the scheduling of school resources. Therefore, if we can expect to take notice of students' academic performance and provide guidance to students who cannot pass the threshold as expected, the waste of school resources can effectively be reduced. In this research, the recent years' student data and course results are used as training data to construct student performance prediction models. The K-Means algorithm was used to classify all courses from the freshman to the senior. The related courses will be grouped in the same cluster, which will more likely to find similar features and improve the accuracy of the prediction. Then, this study constructs independent neural networks for each course according to the different academic year. Each model will be pre-trained by using Denoising Auto-encoder. After pre-training, the corresponding structure and weights are taken as the initial value of the neural network model. Each neural network is treated as a base predictor. All predictors will be integrated into an Ensemble predictor according to different years' weights to predict the current student's course performance. As the students finish the course at the end of each semester, the prediction model will continue track and update to enhance model accuracy through online learning. [ABSTRACT FROM AUTHOR]

Recently, micro-perforated panels (MPP) have become a popular sound absorbing material in the field of architectural acoustics. However, the cost of MPP is still high for the commercial market in Taiwan, and MPP is still not very popular compared to other sound absorbing materials and devices. The objective of this study is to develop a prediction model for MEMM via a machine learning approach. An experiment including 14 types of MEMM was first carried out in a reverberation room based on ISO 354. To predict the sound absorption coefficient of the MEMM, the capability of three conventional models and three machine learning (ML) models of the supervised learning method were studied for the development of the prediction model. The results showed that in most conventional models, the sound absorption coefficient of using an equivalent perimeter had the best agreement compared with other parameters, and the root mean square error (RMSE) between prediction models and experimental data were around 0.2~0.3. However, the RMSE of all ML models was less than 0.1, and the RMSE of the gradient boost model was 0.033 in the training sets and 0.062 in the testing sets, which showed the best agreement with the experiment data. [ABSTRACT FROM AUTHOR]

Prediction model mainly focused on specific diseases, such as diabetes, hypertension, cardiovascular disease, or patients with cancer, or populations of Europe and America, thereby limiting its generalization. This study aimed to develop and validate a 10-year mortality risk score by using data from a population-representative sample of adults. Data were collected from 2,221 Taichung Community Health study participants aged ≥40 years. The baseline period of the study was 2004, and all participants were followed up until death or in 2016. Cox's proportional hazards regression analyses were used to develop the prediction model. A total of 262 deaths were ascertained during the 10-year follow-up. The all-cause mortality prediction model calculated the significant risk factors, namely, age, sex, marital status, physical activity, tobacco use, estimated glomerular filtration rate, and albumin-to-creatinine ratio, among the baseline risk factors. The expanded cardiovascular disease (CVD) mortality prediction model consisted of six variables: age, sex, body mass index, heart disease plus heart disease medication use, stroke plus medication use, and ankle–brachial index. The areas under receiver operating curves of the 3-, 5- and 10-year predictive models varied between 0.97, 0.96, and 0.88 for all-cause mortality, and between 0.97, 0.98, and 0.84 for expanded CVD mortality. These mortality prediction models are valid and can be used as tools to identify the increased risk for mortality. [ABSTRACT FROM AUTHOR]