Search

UNLABELLED: The global uptake of prostate cancer (PCa) active surveillance (AS) is steadily increasing. While prostate-specific antigen density (PSAD) is an important baseline predictor of PCa progression on AS, there is a scarcity of recommendations on its use in follow-up. In particular, the best way of measuring PSAD is unclear. One approach would be to use the baseline gland volume (BGV) as a denominator in all calculations throughout AS (nonadaptive PSAD, PSADNA), while another would be to remeasure gland volume at each new magnetic resonance imaging scan (adaptive PSAD, PSADA). In addition, little is known about the predictive value of serial PSAD in comparison to PSA. We applied a long short-term memory recurrent neural network to an AS cohort of 332 patients and found that serial PSADNA significantly outperformed both PSADA and PSA for follow-up prediction of PCa progression because of its high sensitivity. Importantly, while PSADNA was superior in patients with smaller glands (BGV ≤55 ml), serial PSA was better in men with larger prostates of >55 ml. PATIENT SUMMARY: Repeat measurements of prostate-specific antigen (PSA) and PSA density (PSAD) are the mainstay of active surveillance in prostate cancer. Our study suggests that in patients with a prostate gland of 55 ml or smaller, PSAD measurements are a better predictor of tumour progression, whereas men with a larger gland may benefit more from PSA monitoring.

In pectus excavatum, three-dimensional (3D) surface imaging provides an accurate and radiation-free alternative to computed tomography (CT) to determine severity. Yet, it does not allow for cardiac evaluation since 3D imaging solely captures the chest wall surface. The objective was to develop a 3D image-based prediction model for cardiac compression in patients evaluated for pectus excavatum. A prospective cohort study was conducted including consecutive patients referred for pectus excavatum who received a thoracic CT. Additionally, 3D images were acquired. The external pectus depth, its length, craniocaudal position, cranial slope, asymmetry, anteroposterior distance and chest width were calculated from 3D images. Together with baseline patient characteristics they were submitted to forward multivariable logistic regression to identify predictors for cardiac compression. Cardiac compression on CT was used as reference. The model's performance was depicted by the area under the receiver operating characteristic (AUROC) curve. Internal validation was performed using bootstrapping. Sixty-one patients were included of whom 41 had cardiac compression on CT. A combination of the 3D image derived external pectus depth and external anteroposterior distance was identified as predictive for cardiac compression, yielding an AUROC of 0.935 (95% confidence interval [CI]: 0.878-0.992) with an optimism of 0.006. In a second model for males alone, solely the external pectus depth was identified as predictor, yielding an AUROC of 0.947 (95% CI: 0.892-1.000) with an optimism of 0.0002. We have developed two 3D image-based prediction models for cardiac compression in patients evaluated for pectus excavatum which provide an outstanding discriminatory performance between the presence and absence of cardiac compression with negligible optimism.

Site preservation and associated terrain characteristics are among the major agents controlling the spatial distribution and discoverability of late Quaternary archaeological sites. However, site expectancy in extensive areas can be determined using a combination of geoarchaeological investigation and digital spatial analysis. In that sense a prediction model, respectively the result of a Multi Criteria Decision Analysis (MCDA), is presented to highlight areas with a high potential to contain Upper Palaeolithic sites in the Eastern Mediterranean and bordering arid margins. Based on detailed on-site analysis in the Wadi Sabra region, with its many sites and the location of other well-known sites in the southern Levant, eight individual parameters, aspect (cardinal point), elevation, geomorphology, hydrogeology, drainage network, slope inclination, vegetation, and a terrain ruggedness index (TRI) are evaluated and tested for significance. The statistical results confirm a significant divergence of the spatial distribution of Upper Palaeolithic sites from the distribution of natural terrain position factors. This allows for a definition of parameter classes with a high site expectancy and their use in a predictive model. For the results map, the site expectancy factors determined in this way were intersected with each other, thus enabling a distinction between areas with a high and low site occurrence probability. The accuracy of the presented result is assessed by the implementation of another MCDA pass based on an independent dataset of archaeological survey sites and compared to the original. Both datasets complement each other and can show the potential for further investigations with more numerous site datasets as well as more detailed spatial information. A local classification based on high-resolution elevation information and field data is evaluated for the Wadi Sabra, Southern Jordan, which can confirm yet further specify the result from the MCDA.

Machine learning may enhance prediction of outcomes after coronary artery bypass grafting (CABG). We sought to develop and validate a dynamic machine learning model to predict CABG outcomes at clinically relevant pre- and postoperative time points.The Society of Thoracic Surgeons (STS) registry data elements from 2086 isolated CABG patients were divided into training and testing datasets and input into Extreme Gradient Boosting decision-tree machine learning algorithms. Two prediction models were developed based on data from preoperative (80 parameters) and postoperative (125 parameters) phases of care. Outcomes included operative mortality, major morbidity or mortality, high cost, and 30-day readmission. Machine learning and STS model performance were assessed using accuracy and the area under the precision-recall curve (AUC-PR).Preoperative machine learning models predicted mortality (accuracy, 98%; AUC-PR = 0.16; F1 = 0.24), major morbidity or mortality (accuracy, 75%; AUC-PR = 0.33; F1 = 0.42), high cost (accuracy, 83%; AUC-PR = 0.51; F1 = 0.52), and 30-day readmission (accuracy, 70%; AUC-PR = 0.47; F1 = 0.49) with high accuracy. Preoperative machine learning models performed similarly to the STS for prediction of mortality (STS AUC-PR = 0.11; P = .409) and outperformed STS for prediction of mortality or major morbidity (STS AUC-PR = 0.28; P.001). Addition of intraoperative parameters further improved machine learning model performance for major morbidity or mortality (AUC-PR = 0.39; P.01) and high cost (AUC-PR = 0.64; P.01), with cross-clamp and bypass times emerging as important additive predictive parameters.Machine learning can predict mortality, major morbidity, high cost, and readmission after isolated CABG. Prediction based on the phase of care allows for dynamic risk assessment through the hospital course, which may benefit quality assessment and clinical decision-making.

Wax sedimentation in pipelines is a severe crude oil production and transportation challenge. Pipeline surface roughness is experienced at the early stages of the problem; with time, the effective pipe cross-sectional area is reduced due to pipeline wax plugging, causing pumping pressure increase, equipment failures, and blockages, resulting in unnecessary downtime costs and pipeline abandonments in the worst situation. This paper reviews mathematical and experimental loops models used for pipeline solid wax predictions and calculations as functions of pressure, temperature, and fluid composition; by assessing model's Assumptions, strengths and weaknesses. It is found that most mathematical models applied molecular-diffusion mechanisms in modeling and neglected shear effects; which resulted in wax over-prediction. Experimental loop was time-consuming due to mounting and dismounting of test section during wax deposition measurements; our modification has included sensor-integration to detect, measure, and analyze wax deposition; Reliable wax predictions models are essential to properly design pipelines and adopt cost-effective strategies for wax deposition prevention, control, and removal.

An accurate prediction of ship fuel consumption is critical for speed, trim, and voyage optimisation etc. While previous studies have focused on predicting ship fuel consumption with respect to a variety of factors, research on the impact of environmental factors on fuel consumption has been lacking. In addition, although recent research efforts have widely focused on machine learning methods to predict fuel consumption, studies on hyperparameter values that are suitable for these prediction models are limited. To compensate for this deficiency in existing literature, an adaptive hyperparameter tuning method is proposed, and the effects of maritime environmental factors on fuel consumption are taken into account. Through experimentation, the proposed adaptive hyperparameter tuning method was validated via artificial neural network (ANN), support vector regression (SVR), random forest (RF), and least absolute shrinkage and selection operator (Lasso). The hyperparameter tuning proportionally increased the amplitudes of the coefficients of determination (R2) of these algorithms. The increase of the amplitude demonstrated the following trend, in the order of the largest increase to the lowest increase: ANN, Lasso, SVM, and RF. The rates of increase were between 0.0773% and 2.1653%. Furthermore, after the environmental factors were considered, the prediction accuracies of the ANN and Lasso increased; however, the opposite was observed for the SVR and RF. As such, we confirmed that the use of Bayesian optimisation for hyperparameter tuning can effectively improve the fuel consumption prediction accuracy, and our proposed model can therefore serve as a significant reference for calculating fuel consumption.

Coronavirus disease 2019 (COVID-19) has endured constituting formidable economic, social, educational, and phycological challenges for the societies. Moreover, during pandemic outbreaks, the hospitals are overwhelmed with patients requiring more intensive care units and intubation equipment. Therein, to cope with these urgent healthcare demands, the state authorities seek ways to develop policies based on the estimated future casualties. These policies are mainly non-pharmacological policies including the restrictions, curfews, closures, and lockdowns. In this paper, we construct three model structures of the SpInItIbD-N (suspicious Sp, infected In, intensive care It, intubated Ib, and dead D together with the non-pharmacological policies N) holding two key targets. The first one is to predict the future COVID-19 casualties including the intensive care and intubated ones, which directly determine the need for urgent healthcare facilities, and the second one is to analyse the linear and non-linear dynamics of the COVID-19 pandemic under the non-pharmacological policies. In this respect, we have modified the non-pharmacological policies and incorporated them within the models whose parameters are learned from the available data. The trained models with the data released by the Turkish Health Ministry confirmed that the linear SpInItIbD-N model yields more accurate results under the imposed non-pharmacological policies. It is important to note that the non-pharmacological policies have a damping effect on the pandemic casualties and this can dominate the non-linear dynamics. Herein, a model without pharmacological or non-pharmacological policies might have more dominant non-linear dynamics. In addition, the paper considers two machine learning approaches to optimize the unknown parameters of the constructed models. The results show that the recursive neural network has superior performance for learning nonlinear dynamics. However, the batch least squares outperforms in the presence of linear dynamics and stochastic data. The estimated future pandemic casualties with the linear SpInItIbD-N model confirm that the suspicious, infected, and dead casualties converge to zero from 200000, 1400, 200 casualties, respectively. The convergences occur in 120 days under the current conditions.

Time-series data are widespread in real-world industrial scenarios. To recover and infer missing information in real-world applications, the problem of time-series prediction has been widely studied as a classical research topic in data mining. Deep learning architectures have been viewed as next-generation time-series prediction models. However, recent studies have shown that deep learning models are vulnerable to adversarial attacks. In this study, we prospectively examine the problem of time-series prediction adversarial attacks and propose an attack strategy for generating an adversarial time series by adding malicious perturbations to the original time series to deteriorate the performance of time-series prediction models. Specifically, a perturbation-based adversarial example generation algorithm is proposed using the gradient information of the prediction model. In practice, unlike the imperceptibility to humans in the field of image processing, time-series data are more sensitive to abnormal perturbations and there are more stringent requirements regarding the amount of perturbations. To address this challenge, we craft an adversarial time series based on the importance measurement to slightly perturb the original data. Based on comprehensive experiments conducted on real-world time-series datasets, we verify that the proposed adversarial attack methods not only effectively fool the target time-series prediction model LSTNet, they also attack state-of-the-art CNN-, RNN-, and MHANET-based models. Meanwhile, the results show that the proposed methods achieve a good transferability. That is, the adversarial examples generated for a specific prediction model can significantly affect the performance of the other methods. Moreover, through a comparison with existing adversarial attack approaches, we can see that much smaller perturbations are sufficient for the proposed importance-measurement based adversarial attack method. The methods described in this paper are significant in understanding the impact of adversarial attacks on a time-series prediction and promoting the robustness of such prediction technologies.

Background Risk prediction models that estimate patient probabilities of adverse events are commonly deployed in bariatric surgery. The objective was to validate a machine learning (Super Learner) prediction model of 30-day readmission after bariatric surgery in comparison with a traditional logistic regression. Methods This prognostic study for validation of risk prediction models used data from the Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program. Patients who underwent elective laparoscopic gastric bypass or laparoscopic sleeve gastrectomy between 2015 and 2018 were included. Models used 5-fold cross-validation and were evaluated using the area under the receiver operating characteristic curve, the net reclassification index, and the integrated discrimination improvement. Results The 30-day readmission rate among 393,833 patients was 3.9%. Super Learner area under the receiver operating characteristic curve was 0.674 (95% confidence interval 0.670–0.679), compared to 0.650 (95% confidence interval 0.645–0.654) for logistic regression. The net reclassification index was 0.239 (95% confidence interval 0.223–0.254), and 0.252 (95% confidence interval 0.249–0.255) for those who were and were not readmitted within 30 days. The integrated discrimination improvement was 0.0032 (95% confidence interval 0.0030–0.0033). Conclusion The Super Learner outperformed traditional logistic regression in predicting risk of 30-day readmission after bariatric surgery. Machine learning models may help target high-risk patients more optimally and prevent unnecessary readmissions.

Background Despite a better understanding of the epidemiology, pathogenesis, and management of patients with anaphylaxis, there remain knowledge gaps. Enumerating and prioritizing these gaps would allow limited scientific resources to be directed more effectively. Objective We sought to systematically describe and appraise anaphylaxis knowledge gaps and future research priorities based on their potential impact and feasibility. Methods We convened a 25-member multidisciplinary panel of anaphylaxis experts. Panelists formulated knowledge gaps/research priority statements in an anonymous electronic survey. Four anaphylaxis themed writing groups were formed to refine statements: (1) Population Science, (2) Basic and Translational Sciences, (3) Emergency Department Care/Acute Management, and (4) Long-Term Management Strategies and Prevention. Revised statements were incorporated into an anonymous electronic survey, and panelists were asked to rate the impact and feasibility of addressing statements on a continuous 0 to 100 scale. Results The panel generated 98 statements across the 4 anaphylaxis themes: Population Science (29), Basic and Translational Sciences (27), Emergency Department Care/Acute Management (24), and Long-Term Management Strategies and Prevention (18). Median scores for impact and feasibility ranged from 50.0 to 95.0 and from 40.0 to 90.0, respectively. Key statements based on median rating for impact/feasibility included the need to refine anaphylaxis diagnostic criteria, identify reliable diagnostic, predictive, and prognostic anaphylaxis bioassays, develop clinical prediction models to standardize postanaphylaxis observation periods and hospitalization criteria, and determine immunotherapy best practices. Conclusions We identified and systematically appraised anaphylaxis knowledge gaps and future research priorities. This study reinforces the need to harmonize scientific pursuits to optimize the outcomes of patients with and at risk of anaphylaxis.

Accurate diagnosis of a disease is essential in healthcare. Prediction models, based on classical regression techniques, are widely used in clinical practice. Machine Learning (ML) techniques might be preferred in case of a large amount of data per patient and relatively limited numbers of subjects. However, this increases the risk of overfitting, and external validation is imperative. However, in the field of ML, new and more efficient techniques are developed rapidly, and if recruiting patients for a validation study is time consuming, the ML technique used to develop the first model might have been surpassed by more efficient ML techniques, rendering this original model no longer relevant. We demonstrate a stepwise design for simultaneous development and validation of prediction models based on ML techniques. The design enables - in one study - evaluation of the stability and robustness of a prediction model over increasing sample size as well as assessment of the stability of sensitivity/specificity at a chosen cut-off. This will shorten the time to introduction of a new test in health care. We finally describe how to use regular clinical parameters in conjunction with ML based predictions, to further enhance differentiation between subjects with and without a disease.

Machine learning; Mechanical ventilation; Prognostic model Aprenentatge automàtic; Ventilació mecànica; Model pronòstic Aprendizaje automático; Ventilacion mecanica; Modelo pronóstico Purpose Existing expert systems have not improved the diagnostic accuracy of ventilator-associated pneumonia (VAP). The aim of this systematic literature review was to review and summarize state-of-the-art prediction models detecting or predicting VAP from exhaled breath, patient reports and demographic and clinical characteristics. Methods Both diagnostic and prognostic prediction models were searched from a representative list of multidisciplinary databases. An extensive list of validated search terms was added to the search to cover papers failing to mention predictive research in their title or abstract. Two authors independently selected studies, while three authors extracted data using predefined criteria and data extraction forms. The Prediction Model Risk of Bias Assessment Tool was used to assess both the risk of bias and the applicability of the prediction modelling studies. Technology readiness was also assessed. Results Out of 2052 identified studies, 20 were included. Fourteen (70%) studies reported the predictive performance of diagnostic models to detect VAP from exhaled human breath with a high degree of sensitivity and a moderate specificity. In addition, the majority of them were validated on a realistic dataset. The rest of the studies reported the predictive performance of diagnostic and prognostic prediction models to detect VAP from unstructured narratives [2 (10%)] as well as baseline demographics and clinical characteristics [4 (20%)]. All studies, however, had either a high or unclear risk of bias without significant improvements in applicability. Conclusions The development and deployment of prediction modelling studies are limited in VAP and related outcomes. More computational, translational, and clinical research is needed to bring these tools from the bench to the bedside. The project is supported by the Academy of Finland (project number 326291) and the University of Oulu.

In order to develop prediction models of one-year treatment response in lupus nephritis, an approach using machine learning to combine traditional clinical data and novel urine biomarkers was undertaken. Contemporary lupus nephritis biomarkers were identified through an unbiased PubMed search. Thirteen novel urine proteins contributed to the top 50% of ranked biomarkers and were selected for measurement at the time of lupus nephritis flare. These novel markers along with traditional clinical data were incorporated into a variety of machine learning algorithms to develop prediction models of one-year proteinuria and estimated glomerular filtration rate (eGFR). Models were trained on 246 individuals from four different sub-cohorts and validated on an independent cohort of 30 patients with lupus nephritis. Seven models were considered for each outcome. Three quarters of these models demonstrated good predictive value with areas under the receiver operating characteristic curve over 0.7. Overall, prediction performance was the best for models of eGFR response to treatment. Furthermore, the best performing models contained both traditional clinical data and novel urine biomarkers, including cytokines, chemokines, and markers of kidney damage. Thus, our study provides further evidence that a machine learning approach can predict lupus nephritis outcomes at one year using a set of traditional and novel biomarkers. However, further validation of the utility of machine learning as a clinical decision aid to improve outcomes will be necessary before it can be routinely used in clinical practice to guide therapy.

Accumulated studies have discovered that circular RNAs (CircRNAs) are closely related to many complex human diseases. Due to this close relationship, CircRNAs can be used as good biomarkers for disease diagnosis and therapeutic targets for treatments. However, the number of experimentally verified circRNA-disease associations are still fewer and also conducting wet-lab experiments are constrained by the small scale and cost of time and labour. Therefore, effective computational methods are required to predict associations between circRNAs and diseases which will be promising candidates for small scale biological and clinical experiments. In this paper, we propose novel computational models based on Graph Convolution Networks (GCN) for the potential circRNA-disease association prediction. Currently most of the existing prediction methods use shallow learning algorithms. Instead, the proposed models combine the strengths of deep learning and graphs for the computation. First, they integrate multi-source similarity information into the association network. Next, models predict potential associations using graph convolution which explore this important relational knowledge of that network structure. Two circRNA-disease association prediction models, GCN based Node Classification (GCN-NC) and GCN based Link Prediction (GCN-LP) are introduced in this work and they demonstrate promising results in various experiments and outperforms other existing methods. Further, a case study proves that some of the predicted results of the novel computational models were confirmed by published literature and all top results could be verified using gene-gene interaction networks.

Forecasting solar irradiance is of utmost importance in supplying renewable energy efficiently and timely. This paper aims to experiment five variants of recurrent neural networks (RNN), and develop effective and reliable 5-minute short term solar irradiance prediction models. The 5 RNN classes are long-short term memory (LSTM), gated recurrent unit (GRU), Simple RNN, bidirectional LSTM (Bi-LSTM), and bidirectional GRU (Bi-GRU); the first 3 classes are unidirectional and the last two are bidirectional RNN models. The 26 months data under consideration, exhibits extremely volatile weather conditions in Jinju city, South Korea. Therefore, after different experimental processes, 5 hyper-parameters were selected for each model cautiously. In each model, different levels of depth and width were tested; moreover, a 9-fold cross validation was applied to distinguish them against high variability in the seasonal time-series dataset. Generally the deeper architectures of the aforementioned models had significant outcomes; meanwhile, the Bi-LSTM and Bi-GRU provided more accurate predictions as compared to the unidirectional ones. The Bi-GRU model provided the lowest RMSE and highest R2 values of 46.1 and 0.958; additionally, it required 5.25*10-5 seconds per trainable parameter per epoch, the lowest incurred computational cost among the mentioned models. All 5 models performed differently over the four seasons in the 9-fold cross validation test. On average, the bidirectional RNNs and the simple RNN model showed high robustness with less data and high temporal data variability; although, the stronger architectures of the bidirectional models, deems their results more reliable.

To date, the accurate prediction of tunnel boring machine (TBM) performance remains a considerable challenge owing to the complex interactions between the TBM and ground. Using evolutionary polynomial regression (EPR) and random forest (RF), this study develops two novel prediction models for TBM performance. Both models can predict the TBM penetration rate and field penetration index as outputs with four input parameters: the uniaxial compressive strength, intact rock brittleness index, distance between planes of weakness, and angle between the tunnel axis and planes of weakness (α). First, the performances of both EPR- and RF-based models are examined by comparison with the conventional numerical regression method (i.e., multivariate linear regression). Subsequently, the performances of the RF- and EPR-based models are further investigated and compared, including the model robustness for unknown datasets, interior relationships between input and output parameters, and variable importance. The results indicate that the RF-based model has greater prediction accuracy, particularly in identifying outliers, whereas the EPR-based model is more convenient to use by field engineers owing to its explicit expression. Both EPR- and RF-based models can accurately identify the relationships between the input and output parameters. This ensures their excellent generalization ability and high prediction accuracy on unknown datasets.

As a powerful tool for data streams processing, the vast majority of existing evolving intelligent systems (EISs) learn prediction models from data in a supervised manner. However, high-quality labelled data can be difficult to obtain in many real-world classification applications concerning data streams, though unlabelled data is plentiful. To overcome the labelling bottleneck and construct a stronger classification model, a novel semi-supervised EIS is proposed in this paper. After being primed with a small amount of labelled data, the proposed method is capable of continuously self-developing its system structure and self-updating the meta-parameters from unlabelled data streams chunk-by-chunk in a non-iterative, exploratory manner by exploiting a novel pseudo-labelling strategy. Thanks to its transparent prototype-based structure and human-understandable reasoning process, the proposed method can provide users high explainability and interpretability while achieving great classification precision. Experimental investigation demonstrates the superior performance of the proposed method.

In this study, the warpage prediction models have been developed for an injection molded PVC component named a drip chamber. Two popular and widely used ensemble machine learning algorithms, namely random forest and gradient boosted regression tree have been used to develop the predictive models. 40 experiments were carried out for various input process parameters to develop the warpage prediction dataset. ANOVA was performed to identify the significant input process parameters. These process parameters were barrel temperature, holding pressure, holding time, mold temperature, and cooling time. The results shows that the mean absolute percentage errors of random forest and gradient boosted regression tree model are 3.25% and 9.37%, respectively. This indicates that random forest ensemble algorithm outperforms gradient boosted model in predicting the warpage of injection molded part. This model allows production managers to monitor injection molding method parameters and regulate the warpage before actual production and with minimum waste.

Accurate wind power forecasting is a vital factor in day-ahead dispatch and increasing the level of penetration of renewable energy. The feature extraction of meteorological factors related to wind power output is a big challenge to improve forecasting accuracy, and selecting key meteorological factors based on experience will decrease the prediction accuracy. Therefore, a day-ahead wind power combined forecasting approach is innovatively proposed through key meteorological factors selection, data decomposition and reconstruction, combined forecasting model generation, and optimization strategy. Correlated variables namely variational mode decomposition and weighted permutation entropy (VMD-WPE) decomposed historical wind power and key meteorological factors are used as the inputs. A forecasting model based on convolutional neural network (CNN) and long short-term memory network (LSTM) is used to forecast future wind power. Four optimizers with different optimization performances are used to find the best parameters of the forecasting model to obtain accurate prediction results. Multiple comparative experiments from regional wind farms in Ningxia and Jilin of China are utilized as case studies to evaluate the effectiveness of the proposed model. Results show that the proposed approach outperforms other benchmark prediction models, taking into account multiple-error metrics including error metrics, accuracy rate, and improvement percentages.

Eutrophication and excessive algal growth pose a threat on aquatic organisms and the health of the public, environment, and the economy. Understanding what drives excessive algal growth can inform mitigation measures and aid in advance planning to minimize impacts. We demonstrate how simulated data from weather, hydrological, and agroecosystem numerical prediction models can be combined with machine learning (ML) to assess and predict chlorophyll a (chl a) concentrations, a proxy for lake eutrophication and algal biomass. The study area is Lake Erie for a 16-year period, 2002–2017. A total of 20 environmental variables from linked and coupled physical models are used as input features to train the ML model with chl a observations from 16 measuring stations. Included are meteorological variables from the Weather Research and Forecasting (WRF) model, hydrological variables from the Variable Infiltration Capacity (VIC) model, and agricultural management practice variables from the Environmental Policy Integrated Climate (EPIC) agroecosystem model. The consolidation of these variables is conducive to a successful prediction of chl a. Aside from the synergistic effects that weather, hydrology, and fertilizers have on eutrophication and excessive algal growth, we found that the application of different forms of both P and N fertilizers are highly ranked for the prediction of chl a concentration. The developed ML model successfully predicts chl a with a coefficient of determination of 0.81, bias of −0.12 μg/l and RMSE of 4.97 μg/l. The developed ML-based modeling approach can be used for impact assessment of agriculture practices in a changing climate that affect chl a concentrations in Lake Erie.

Flow-accelerated corrosion (FAC) of carbon steel piping is a significant problem in nuclear power plants. The basic process of FAC is currently understood relatively well; however, the accuracy of prediction models of the wall-thinning rate under an FAC environment is not reliable. Herein, we propose a methodology to construct pipe wall-thinning rate prediction models using artificial neural networks and a convolutional neural network, which is confined to a straight pipe without geometric changes. Furthermore, a methodology to generate training data is proposed to efficiently train the neural network for the development of a machine learning-based FAC prediction model. Consequently, it is concluded that machine learning can be used to construct pipe wall thinning rate prediction models and optimize the number of training datasets for training the machine learning algorithm. The proposed methodology can be applied to efficiently generate a large dataset from an FAC test to develop a wall thinning rate prediction model for a real situation.

Objectives: No clear guidance exists on handling missing data at each stage of developing, validating and implementing a clinical prediction model (CPM). We aimed to review the approaches to handling missing data that underly the CPMs currently recommended for use in UK healthcare.Study Design and Setting: A descriptive cross-sectional meta-epidemiological study aiming to identify CPMs recommended by the National Institute for Health and Care Excellence (NICE), which summarized how missing data is handled across their pipelines.Results: A total of 23 CPMs were included through “sampling strategy.” Six missing data strategies were identified: complete case analysis (CCA), multiple imputation, imputation of mean values, k-nearest neighbours imputation, using an additional category for missingness, considering missing values as risk-factor-absent. 52% of the development articles and 48% of the validation articles did not report how missing data were handled. CCA was the most common approach used for development (40%) and validation (44%). At implementation, 57% of the CPMs required complete data entry, whilst 43% allowed missing values. Three CPMs had consistent paths in their pipelines.Conclusion: A broad variety of methods for handling missing data underly the CPMs currently recommended for use in UK healthcare. Missing data handling strategies were generally inconsistent. Better quality assurance of CPMs needs greater clarity and consistency in handling of missing data.

Background\ud The use of accurate prediction tools and early intervention are important for addressing severe coronavirus disease 2019 (COVID-19). However, the prediction models for severe COVID-19 available to date are subject to various biases. This study aimed to construct a nomogram to provide accurate, personalized predictions of the risk of severe COVID-19.\ud \ud Methods\ud This study was based on a large, multicenter retrospective derivation cohort and a validation cohort. The derivation cohort consisted of 496 patients from Jiangsu Province, China, between January 10, 2020, and March 15, 2020, and the validation cohort contained 105 patients from Huangshi, Hunan Province, China, between January 21, 2020, and February 29, 2020. A nomogram was developed with the selected predictors of severe COVID-19, which were identified by univariate and multivariate logistic regression analyses. We evaluated the discrimination of the nomogram with the area under the receiver operating characteristic curve (AUC) and the calibration of the nomogram with calibration plots and Hosmer-Lemeshow tests.\ud \ud Results\ud Three predictors, namely, age, lymphocyte count, and pulmonary opacity score, were selected to develop the nomogram. The nomogram exhibited good discrimination (AUC 0.93, 95% confidence interval [CI] 0.90–0.96 in the derivation cohort; AUC 0.85, 95% CI 0.76–0.93 in the validation cohort) and satisfactory agreement.\ud \ud Conclusions\ud The nomogram was a reliable tool for assessing the probability of severe COVID-19 and may facilitate clinicians stratifying patients and providing early and optimal therapies.

BACKGROUND Research on predictors and risk of recurrence after suicide attempt from China is lacking. This study aims to identify risk factors and develop prediction models for recurrent suicidal behavior among suicide attempters using Cox proportional hazard (CPH) and machine learning methods. METHODS The prospective cohort study included 1103 suicide attempters with a maximum follow-up of 10 years from rural China. Baseline characteristics, collected by face-to-face interviews at least 1 month later after index suicide attempt, were used to predict recurrent suicidal behavior. CPH and 3 machine learning algorithms, namely, the least absolute shrinkage and selection operator, random survival forest, and gradient boosting decision tree, were used to construct prediction models. Model performance was accessed by concordance index (C-index) and the time-dependent area under the receiver operating characteristic curve (AUC) value for discrimination, and time-dependent calibration curve along with Brier score for calibration. RESULTS The median follow-up time was 7.79 years, and 49 suicide attempters had recurrent suicidal behavior during the study period. Four models achieved comparably good discrimination and calibration performance, with all C-indexes larger than 0.70, AUC values larger than 0.65, and Brier scores smaller than 0.06. Mental disorder emerged as the most important predictor across all four models. Suicide attempters with mental disorders had a 3 times higher risk of recurrence than those without. History of suicide attempt (HR = 2.84, 95% CI: 1.34-6.02), unstable marital status (HR = 2.81, 95% CI: 1.38-5.71), and older age (HR = 1.51, 95% CI: 1.14-2.01) were also identified as independent predictors of recurrent suicidal behavior by CPH model. CONCLUSIONS We developed four models to predict recurrent suicidal behavior with comparable good prediction performance. Our findings potentially provided benefits in screening vulnerable individuals on a more precise scale.

To further explore the empirical seismic capacity of brick and wood (BW) structures and the vulnerability characteristics of the multiple seismic intensity regions, a vulnerability analysis was conducted based on historical observation data and a field survey of 154 typical destructive earthquakes occurring in 18 provinces (municipalities and autonomous regions) of China from 1990 to 2010. Based on the data collection, classification, and mathematical statistics of the empirical seismic damage of BW structures, a vulnerability probability matrix model based on 11,520.403 × 104 m2 and 48,456 BW structures was obtained and established. Considering the statistical analysis of the empirical data by year, frequency, survey area, and number, empirical vulnerability models (matrix, point cloud, function, and curve) based on the evaluation of the Chinese seismic intensity scale (CSIS) were established by employing the vulnerability probability model, nonlinear regression theory, and numerical analysis method. The failure ratio (FR) and exceedance probability (EP) were employed as vulnerability assessment parameters, and nonlinear vulnerability demand comparison models considering empirical damage of BW in different years were established. In addition, the mean damage index (MDI) parameter matrix model was proposed to evaluate the overall vulnerability of regional BW. Combined with creating the empirical seismic damage investigation sample database, the evaluation and prediction models of BW vulnerability based on MDI were conducted, considering the overall investigation area.

International audience; The use of CNN and segmentation to extract image features for the prediction of four postures for sows kept in crates was examined. The extracted features were used as input variables in an SVM classification method to estimate posture. The possibility of using a posture prediction model with images not necessarily obtained under the same conditions as those used for the training set was explored. As a reference case, the efficacy of the posture prediction model was explored when training and testing datasets were built using the same pool of images. In this case, all the models produced satisfactory results, with a maximum f1-score of 97.7% with CNNs and 93.3% with segmentation. To evaluate the impact of environmental variations, the models were trained and tested on different monitoring days. In this case, the best f1-score dropped to 86.7%. The impact of using the posture prediction model on animals that were not present in the training dataset was then explored. The best f1-score reduced to 63.4% when the posture prediction models were trained on one animal and tested on 11 other different animals. Conversely, when the models were tested on one animal and trained on the 11 others, the f1-score only decreased to 86% with the best model. On average, a decrease of around 17% caused by environmental and individual variations between training and testing was observed.; Nous avons comparé l’utilisation des CNN et de la segmentation pour extraire des caractéristiques d’intérêts dans les images (features), afin de prédire la posture de truies allaitantes. Une fois extraite, les features peuvent être utilisées comme variable d’entrée d’une méthode de classification, de type SVM, pour prédire la posture de la truie. Nous avons exploré l’impact de l’ensemble d’apprentissage sur la qualité de la prédiction, notamment lorsque l’ensemble d’apprentissage et de test ne sont pas obtenues dans les mêmes conditions. Nous avons d’abord considéré un cas de référence, où la méthode d’estimation de la posture était entrainée et testée sur des images prises dans les mêmes conditions. Dans ce cas, tous les modèles testés ont fournis des résultats satisfaisants, avec un f1-score de 97.7% pour le meilleur CNN et de 93.3% pour la méthode basée sur la segmentation d’images. Pour évaluer l’impact des changements environnementaux lors de la prise de vue des images, nous avons ensuite entrainé et testé les modèles sur des images provenants de deux jours différents. Dans ce cas, le meilleur f1-score est réduit à 86.7%. Enfin, nous avons considéré le cas où des animaux présent dans l’ensemble test, ne le sont pas dans l’ensemble d’apprentissage. Lorsque le modèle de prediction de posture est entrainé sur une truie, et testé sur les 11 truies, le meilleur f1-score obtenu est de 63.4%. Lorsque le modèle est entrainé sur 11 truies et testé sur 1 truie, le f1-score est de 86%. En moyenne, nous avons observé que les variations environnementales et individuelles faisaient baisser le f1-score de 17%.

Customer churn prediction models using machine learning classification have been developed predominantly by training and testing on one time slice of data. We train models on multiple time slices of data and refer to this approach as multi-slicing. Our results show that given the same time frame of data, multi-slicing significantly improves churn prediction performance compared to training on the entire data set as one time slice. We demonstrate that besides an increased training set size, the improvement is driven by training on samples from different time slices. For data from a convenience wholesaler, we show that multi-slicing addresses the rarity of churn samples and the risk of overfitting to the distinctive situation in a single training time slice. Multi-slicing makes a model more generalizable, which is particularly relevant whenever conditions change or fluctuate over time. We also discuss how to choose the number of time slices.

The scale of criminal networks (e.g. drug syndicates and terrorist networks) extends globally and poses national security threat to many nations as they also tend to be technologically advance (e.g. Dark Web and Silk Road cryptocurrency). Therefore, it is critical for law enforcement agencies to be equipped with the latest tools in criminal network analysis (CNA) to obtain key hidden links (relationships) within criminal networks to preempt and disrupt criminal network structures and activities. Current hidden or missing link predictive models that are based on Social Network Analysis models rely on ML techniques to improve the performance of the models in terms of predictive accuracy and computing power. Given the improvement in the recent performance of Deep Reinforcement Learning (DRL) techniques which could train ML models through self-generated dataset, DRL can be usefully applied to domains with relatively smaller dataset such as criminal networks. The objective of this study is to assess the comparative performance of a CNA hidden link prediction model developed using DRL techniques against classical ML models such as gradient boosting machine (GBM), random forest (RF) and support vector machine (SVM). The experiment results exhibit an improvement in the performance of the DRL model of about 7.4% over the next best performing classical RF model trained within 1500 iterations. The performance of these link prediction models can be scaled up with the parallel processing capabilities of graphical processing units (GPUs), to significantly improve the speed of training the model and the prediction of hidden links.

To achieve a high penetration of renewable energy integration, an effective solution is to explore the interdependence between numerical weather prediction (NWP) data and historical wind power to improve prediction accuracy. This paper proposes a novel combined approach for wind power prediction. The characteristics of NWP and historical wind power data are extracted by using the feature extraction technique, the predictor is designed based on extreme learning machine (ELM) and least squares support vector machine (LSSVM) model, and then key parameters of the prediction models are optimized by improving cuckoo search (ICS) to obtain a reliable value, which is defined as the pre-combined prediction value (PPA). To obtain a reliable result, a variance strategy is developed to allocate the weights of the pre-combined prediction model to obtain the final predicted values. Four seasons dataset collected from regional wind farms in China is utilized as a benchmark experiment to evaluate the effectiveness of the proposed approach. The results of comprehensive numerical cases with different seasons show that the proposed approach, which considers multiple-error metrics, including error metrics, accuracy rate, qualification rate, and improvement percentages, achieves higher accuracy than other benchmark prediction models.

To timely track and evaluate the impact of dynamic geopolitical risk (GPR) shocks, we use mixed-frequency GPR to explain and forecast stock market returns. Our empirical findings confirm that the real-time GPR shock has a lasting negative impact on stock returns. Moreover, the overall model fits, and the reliability of forecasts is sensitive to the selected sample frequency. We show that the mixed-frequency GPR estimates can be considered more significant and robust than those from an analogous common-frequency approach. The out-of-sample return forecasting tests further suggest that the prediction models incorporating mixed-frequency GPR provide substantial and statistically significant improvements in stock return forecast accuracy.

The brown planthopper (BPH), Nilaparvata lugens (Stal) is a migratory rice pest that periodically erupts across China, and mainly come from Indochina Peninsula in spring. The climate change in Indochina Peninsula has caused obvious impacts on the immigration of BPH. To further understand the influence of climate change on BPH immigration, the light-trap data of BPH in southern China from 1980 to 2016 were collected, and the best performance data of each meteorological factor (temperature, humidity, precipitation and wind) in Indochina Peninsula by simulating eight Global Climate Models (GCMs) were chosen. The key predicting factors were screened out by identifying correlations between the number of BPH and the meteorological factors. The support vector machine (SVM), back propagation (BP) neural network, and stepwise regression were used to establish medium long-term prediction models of immigration amount of BPH, and their advantages and disadvantages were compared, and prediction of the occurrence of BPH in the future was carried out by using three models based on future climate data provided by GCMs (under RCP4.5 or RCP8.5 scenario). The results showed that: (1) BNU-CSM1–1 model was the best in temperature simulation; CESM1-CAM5 model was the best in humidity simulation; HadGEM2-AO model was the best in precipitation and meridional wind simulation; BCC-CSM1–1 model was the best in zonal wind simulation. (2) The accuracies of retrospective testing and forward forecasting of three models indicated the SVM model was more suitable for medium long-term predicting the occurrence of BPH than BP neural networkmodel and stepwise regression model. (3) The immigration of BPH in Guangxi, Jiangxi, Hunan and Hubei would be greater than other provinces in southern China in the next 7 years, and should be taken seriously.

In chemical processes, grasping the changing trend of critical parameters can help field operators take appropriate adjustments to eliminate potential fluctuations. Thus, deep networks, renowned for its revolutionary feature representation capability, have been gradually exploited for building reliable prediction models from massive data embraced tremendously nonlinearities and dynamics. Because of the inherent complexity, the process trajectories over the whole running duration make distinctive contributions to the ultimate targets. Specifically, features extracted from different secondary variables at different previous instants have diverse impacts on the current state of primary variables. However, this spatiotemporal relevance discrepancy is rarely considered, which may lead to deterioration of prediction performance. Therefore, this paper seamlessly integrates the spatiotemporal attention (STA) mechanism with convolutional neural networks (CNN) and bi-directional long short-term memory (BiLSTM), and proposes a novel predictive model, namely STA-ConvBiLSTM. Using the deep framework composed of CNN and BiLSTM, the integrated model can, not only automatically explore the esoteric spatial correlations among high-dimensional variables at each time step, but also adaptively excavate beneficial temporal characteristics across all time steps. Meanwhile, STA is further introduced to assign corresponding weights to information with dissimilar importance, so as to prevent high target-relevant interactions from being discarded due to overlong sequences and excessive features. STA-ConvBiLSTM is applied in the case of furnace tube temperature prediction of a delayed coking unit, which exhibits a significant improvement of the prediction accuracy.

As Photovoltaic (PV) energy is impacted by various weather variables such as solar radiation and temperature, one of the key challenges facing solar energy forecasting is choosing the right inputs to achieve the most accurate prediction. Weather datasets, past power data sets, or both sets can be utilized to build different forecasting models. However, operators of grid-connected PV farms do not always have full sets of data available to them especially over an extended period of time as required by key techniques such as multiple regression (MR) or artificial neural network (ANN). Therefore, the research reported here considered these two main approaches of building prediction models and compared their performance when utilizing structural, time-series, and hybrid methods for data input. Three years of PV power generation data (of an actual farm) as well as historical weather data (of the same location) with several key variables were collected and utilized to build and test six prediction models. Models were built and designed to forecast the PV power for a 24-hour ahead horizon with 15 min resolutions. Results of comparative performance analysis show that different models have different prediction accuracy depending on the input method used to build the model: ANN models perform better than the MR regardless of the input method used. The hybrid input method results in better prediction accuracy for both MR and ANN techniques, while using the time-series method results in the least accurate forecasting models. Furthermore, sensitivity analysis shows that poor data quality does impact forecasting accuracy negatively especially for the structural approach.

Many conventional quality prediction models are directly developed based on the easy-to-measure variables, and thus the local information within individual unit may be buried by information of other units. In this study, a cascaded regression network (RegNet) is proposed to solve the aforementioned issue. Specifically, the features which are adopted to develop RegNet model are extracted in two steps, including variable-wise and unit-wise feature extractions. In variable-wise feature extraction, several adjacent variables and their corresponding time lags are integrated using convolutional filter. By this means, both local correlation and temporal information within each unit can be preserved. In the unit-wise feature extraction, the local information of each unit is adopted to further explore the global correlation between different operation units. Based on the obtained global features, a fully connected layer is designed to calculate the regression weight of the quality prediction model. It is noted that the architecture of RegNet can be readily generalized to many existing methods by replacing the convolutional filter and fully connected layer. The performance of the proposed method is illustrated using a simulated process and two real industrial processes, and the experimental results show that it can provide reliable prediction results for industrial applications.

Safety-critical situations in road traffic often result from incorrect estimation of the future behavior of other road users. Therefore, many Advanced Driver Assistance Systems (ADAS) need prediction models to ensure safety. Physical prediction models offer the advantage of general use and work quite well for short prediction horizons, while for longer periods of time, maneuver-based models offer better performance which, however, strongly depends on the data used to train them. An additional challenge for prediction is the fact that the surrounding traffic can change its path, i.e. for safety not only one maneuver should be considered but regular updates are required. Against this background, we propose a method that uses three physics-based predictions – corresponding to different prediction assumptions and models – combined with possible maneuver-based trajectories derived from environmental knowledge. Continuous monitoring is used to select the most likely of the three physics-based models. This choice then influences the environment-based prediction and the output of both models is fused afterwards. The output of the resulting Multiple Model Trajectory Prediction (MMTP) has been validated with measured data from two different scenarios – a city junction and a highway – with a good prediction performance and without the need for special measurements as commonly required for maneuver-based prediction.

Objective The study aimed to explore existing prediction models of functioning in spinal cord injury (SCI). Study Design and Setting The databases PubMed, EBSCOhost CINAHL Complete, and IEEE Xplore were searched for relevant literature. The search strategy included published search filters for prediction model and impact studies, index terms and keywords for SCI, and relevant outcome measures able to assess functioning as reflected in the International Classification of Functioning, Disability and Health (ICF). The search was completed in October 2020. Results We identified seven prediction model studies reporting twelve prediction models of functioning. The identified prediction models were mainly envisioned to be used for rehabilitation planning, however, also other possible applications were stated. The method predominantly used was regression analysis and the investigated predictors covered mainly the ICF components of body functions and activities and participation, next to characteristics of the health condition and health interventions. Conclusion Findings suggest that the development of prediction models of functioning for use in clinical practice remains to be fully exploited. By providing a comprehensive overview of what has been done, this review informs future research on prediction models of functioning in SCI and contributes to an efficient use of research evidence.

Background Few studies report on machine learning models for suicide risk prediction in adolescents and their utility in identifying those in need of further evaluation. This study examined whether a model trained and validated using data from all age groups works as well for adolescents or whether it could be improved. Methods We used healthcare data for 1.4 million specialty mental health and primary care outpatient visits among 256,823 adolescents across 7 health systems. The prediction target was 90-day risk of suicide attempt following a visit. We used logistic regression with least absolute shrinkage and selection operator (LASSO) and generalized estimating equations (GEE) to predict risk. We compared performance of three models: an existing model, a recalibrated version of that model, and a newly-learned model. Models were compared using area under the receiver operating curve (AUC), sensitivity, specificity, positive predictive value and negative predictive value. Results The AUC produced by the existing model for specialty mental health visits estimated in adolescents alone (0.796; [0.789, 0.802]) was not significantly different than the AUC of the recalibrated existing model (0.794; [0.787, 0.80]) or the newly-learned model (0.795; [0.789, 0.801]). Predicted risk following primary care visits was also similar: existing (0.855; [0.844, 0.866]), recalibrated (0.85 [0.839, 0.862]), newly-learned (0.842, [0.829, 0.854]). Limitations The models did not incorporate non-healthcare risk factors. The models relied on ICD9-CM codes for diagnoses and outcome measurement. Conclusions Prediction models already in operational use by health systems can be reliably employed for identifying adolescents in need of further evaluation.

Objective To reduce delays in referral to specialty mental health care, we evaluated clinical prediction models estimating the likelihood of response to primary care treatment of depression in the VA healthcare system. Methods We included patients with a primary care depression diagnosis between October 1, 2015 and December 31, 2017, an initial PHQ-9 score ≥ 10 within 30 days, a follow-up PHQ-9 score within 2-8 months, and no specialty mental health care within three months prior to depression diagnosis. We evaluated eight ordinary least squares regression models, each with a different procedure for selecting predictors of percentage change in PHQ-9 score from baseline to follow-up. Predictors included patient characteristics from electronic health records and neighborhood characteristics from US census data. We repeated each modeling procedure 1,000 times, using different training and validation sets of patients. We used R2, RMSE, and MAE to evaluate model performance. Results The final cohort included 3,464 patients. The two best performing models included multiple iterations of backwards stepwise variable selection with R2 of 0.07, RMSE of 41.45, MAE of 33.30; and R2 of 0.07, RMSE of 41.39, MAE of 33.28. Limitations Wide follow-up interval, possibility of misclassification error due to use of EHR data. Conclusions Model performance did not suggest its use as a guide in clinical decision-making. Future research should explore whether obtaining additional risk factor data from patients (e.g., duration of symptoms) or modeling PHQ-9 scores over a narrower time interval improves performance of clinical risk prediction tools for depression.

Objective Clinical risk prediction models are generally assessed on population level with a lack of measures that evaluate their stability at predicting risks of individual patients. This study evaluated the use of ranking as a measure to assess individual level stability between risk prediction models. Study Design and Setting A large patient cohort (3.66 million patients with 0.11 million cardiovascular events) extracted from the Clinical Practice Research Datalink was used in the exemplar of cardiovascular disease risk prediction. Results It was found that 15 models (including machine learning and statistical models) had similar population-level model performance (C statistics about 0.88). For patients with high absolute risks, the models were more consistent in ranking of risk predictions (interquartile range (IQR) of differences in rank percentiles -0.6 to 1.0), but inconsistent in absolute risk (IQR of differences in absolute risk -18.8 to 9.0). At low risk, the reverse was true with inconsistent ranking but more consistent absolute risk. Conclusion Consistency of ranking of individual risk predictions is a useful measure to assess risk prediction models providing complementary information to absolute risk stability. Model developing guidelines including “TRIPOD” and “PROBAST” should incorporate ranking to assess individual level stability between risk prediction models.

The clinical application of prediction models is increasing within the field of gynaecology and obstetrics. This is mostly due to the fact that clinicians and patients prefer individualized counselling and person specific, more objective outcome assessment. To prevent using inadequate models, it is important to construct and perform prediction model studies correctly. Therefore, the TRIPOD statement (the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) was developed. The aim of this review is to obtain an overview of the existing published prediction models for benign gynaecology and to investigate to what extent these studies meet the TRIPOD criteria. We performed a literature search in the databases PubMed, Embase and Cochrane Library from inception to August 2020. Searching the cross-references of the relevant studies within our search identified additional articles. Publications were included if the aim of the study was to develop a multivariable prediction model within the field of benign gynaecology. Two independent reviewers extracted the data. Analysis of the studies was performed by using a checklist derived from the TRIPOD criteria. Based on our search, 2487 studies were selected, including potential duplications. Eventually, a total of twenty-two studies were selected. 91% of these studies handled their predictors by univariable analysis before developing a multivariable prediction model. Fifteen studies described having missing data, but not all of them (9%) handled these missing data. Four different internal validation methods were used in twenty studies. Fifteen studies (68%) had prediction models with a C-index ≥ 0.7, which indicates a good model. Half of the studies (50%) did not measure the calibration, overall performance was described in two studies (9%). External validation was performed in 9% of the studies. The correct development of a prediction model within benign gynaecology and subsequent transparent reporting of the model development is important to facilitate clinical use. Without transparent reporting, wrong assumptions can be made leading to incorrect application of a specific prediction model. This overview shows that excepting carrying out an external validation, only one article met all the criteria. Therefore, we strongly recommend use of the TRIPOD criteria for developing and validating a prediction model (study). In addition, prior to publication, content experts should critically and statistically review the prediction model. If too many criteria are not met, refusing publication should be considered.

Moulds cause severe economic losses at different points of plant food commodities production, from the field to the final foodstuffs. Predictive modelling is an increasingly used tool applied to solve different issues in food production. In this opinion, we have dealt, in one hand, with the latest publications on predictive mycology used for early prediction of fungal spoilage of foods, as well as for assessing efficacy of antimicrobials in foods. Moreover, prediction models have been applied to assess the impact that climate change may have in the near future in terms of geographic fungal distribution and impact on mycotoxin occurrence. Finally, there is a growing interest on analysing fungal growth and mycotoxin contamination in cereals and nuts using infrared spectrometry models. All these cases exemplify the increasing interest of predictive modelling to assist decision making in different points of the food chain. This work was supported by the Spanish Ministry of Economy, Industry and Competitiveness (MINECO/AEI/FEDER, UE, project AGL2017- 87755-R). Antoni Femenias acknowledges the financial support of the University of Lleida (predoctoral grant). Luisa Freire acknowledges the financial support of São Paulo Research Foundation: Grant #2016/21041-5, São Paulo Research Foundation (FAPESP). This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001. A. S. Sant’Ana is thankful to the National Council for Scientific and Technological Development (CNPq): Grants #302763/2014-7 and #305804/2017-0.

Background The universal Surgical Risk Preoperative Assessment System (SURPAS) prediction models for postoperative adverse outcomes have good accuracy for estimating risk in broad surgical populations and for surgical specialties. The accuracy in individual operations has not yet been assessed. The objective of this study was to evaluate the Surgical Risk Preoperative Assessment System in predicting adverse outcomes for selected individual operations. Methods The SURPAS models were applied to the top 2 most frequent common procedural terminology codes in 9 surgical specialties and 5 additional common general surgical operations in the 2009 to 2018 database of the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP). Goodness of fit statistics were estimated, including c-indices for discrimination, Hosmer-Lemeshow graphs and P values for calibration, overall observed versus expected event rates, and Brier scores. Results The total sample size was 2,020,172, which represented 29% of the 6.9 million operations in the ACS NSQIP database. Average c-indices across 12 outcomes were acceptable (≥0.70) for 13 (56.5%) of the 23 operations. Overall observed-to-expected rates were similar for mortality and overall morbidity across the 23 operations. Hosmer-Lemeshow graphs over quintiles of risk comparing observed-to-expected rates of mortality and overall morbidity were similar for 52% and 70% of operations, respectively. Model performance was better in less complex operations and those done in patients with lower preoperative risk. Conclusion SURPAS displayed accuracy in estimating postoperative adverse events for some of the 23 operations studied, but not all. In the procedures where SURPAS was not accurate, developing disease or operation-specific risk models might be appropriate.

Available clinical methods for heart failure (HF) diagnosis are expensive and require a high-level of experts intervention. Recently, various machine learning models have been developed for the prediction of HF where most of them have an issue of over-fitting. Over-fitting occurs when machine learning based predictive models show better performance on the training data yet demonstrate a poor performance on the testing data and the other way around. Developing a machine learning model which is able to produce generalization capabilities (such that the model exhibits better performance on both the training and the testing data sets) could overall minimize the prediction errors. Hence, such prediction models could potentially be helpful to cardiologists for the effective diagnose of HF. This paper proposes a two-stage decision support system to overcome the over-fitting issue and to optimize the generalization factor. The first stage uses a mutual information based statistical model while the second stage uses a neural network. We applied our approach to the HF subset of publicly available Cleveland heart disease database. Our experimental results show that the proposed decision support system has optimized the generalization capabilities and has reduced the mean percent error (MPE) to 8.8% which is significantly less than the recently published studies. In addition, our model exhibits a 93.33% accuracy rate which is higher than twenty eight recently developed HF risk prediction models that achieved accuracy in the range of 57.85% to 92.31%. We can hope that our decision support system will be helpful to cardiologists if deployed in clinical setup.

Failure prediction is the task of forecasting whether a material system of interest will fail at a specific point of time in the future. This task attains significance for strategies of industrial maintenance, such as predictive maintenance. For solving the prediction task, machine learning (ML) technology is increasingly being used, and the literature provides evidence for the effectiveness of ML-based prediction models. However, the state of recent research and the lessons learned are not well documented. Therefore, the objective of this review is to assess the adoption of ML technology for failure prediction in industrial maintenance and synthesize the reported results. We conducted a systematic search for experimental studies in peer-reviewed outlets published from 2012 to 2020. We screened a total of 1,024 articles, of which 34 met the inclusion criteria. We focused on understanding the datasets analyzed, the preprocessing to generate features, and the training and evaluation of prediction models. The results reveal (1) a broad range of systems and domains addressed, (2) the adoption of up-to-date approaches to preprocessing and training, (3) some lack of performance evaluation mitigating the overfitting problem, and (4) considerable heterogeneity in the reporting of experimental designs and results. We identify opportunities for future research and suggest ways to facilitate the comparison and integration of evidence obtained from single studies.

Various prediction models have been formulated to predict the need for mechanical ventilation (MV). In this study, we compare the sensitivity and specificity of Erasmus GBS Respiratory Insufficiency Score (EGRIS) and Sharshar score with single breath count (SBC) at 2 Hz to assess their usefulness in predicting MV and functional outcome. The primary outcome was prediction of MV and relative usefulness of all three models based on sensitivity and specificity. The secondary outcome was functional recovery at 6 months using the cut off points for MV and functional outcome at 6months was assessed using Hughes scale and categorized as good (2) and poor (≥2). The median age was 30 years, and 64 (69.6%) were males. The sensitivity and specificity of EGRIS score was 78.1% and 67.2%, Sharshar score was 75.0% and 64.1%, and SBC at 2 Hz was 78.1% and 75.0%. The area under the receiver operating curves was maximum for SBC at 2 Hz (0.86) compared to other models, although insignificant. EGRIS score 5 had better functional recovery compared to those with score ≥ 5 (66% vs 34%, p = 0.001). Similarly, in Sharshar score, 23/59 (39%) with score ≥ 4 had good recovery compared to 36 (61%) with score 4 (p = 0.01). All the three models have comparable predictive values for MV. Outcome prediction is the best with EGRIS.

Manufacturing processes typically involves a number of inspections, including basic inspections for all products and advanced inspections for selected sampled products. The partial application of advanced inspections decreases processing time and cost, although it deteriorates the final quality of products. Recently, several studies have focused on using inspection data to train a prediction model that predicts faults in final products. The lack of advanced inspection data for some products limits the prediction accuracy of the model. Herein, we propose an active inspection framework, where products are intelligently sampled for advanced inspections to achieve high prediction accuracy in a cost-effective manner. Two prediction models are used in the framework: basic and advanced models. The basic model is trained with data from basic inspections, whereas the advanced model is trained with data from both basic and advanced inspections to predict whether a product is faulty. For a product that undergoes basic inspections, the basic model outputs a fault score and its uncertainty in terms of the expected prediction change. If the uncertainty is low, then the fault score of the product is finalized. If the uncertainty is high, then the product is subject to advanced inspections, and the fault score is updated using the advanced model. We demonstrate the effectiveness of the proposed active inspection framework through a case study using real-world data acquired from a semiconductor manufacturer.

Background The aim of this systematic review is to evaluate the current evidence in the context of clinical prediction model for post-hepatectomy liver failure (PHLF). Methods A systematic search of the English literature for a period from December 2005 to September 2020 was conducted. Primary outcome was defined using the three common PHLF criteria (50-50 criteria, peak bilirubin>7 mg/dl criteria, and ≥ grade B PHLF criteria by the International Study Group of Liver Surgery). Studies that reported the value of area under receiver operative characteristic curve (AUC) for the occurrence of PHLF were included. Results Twenty eight of 1327 screened articles were eligible for inclusion. Eighteen studies developed the prediction models. The median AUC was found to be 0.79 (0.65–0.933). The parameters related to the amount of future liver remnant volume were most commonly identified as significant predictors for PHLF in statistical analysis (24 studies) and were most frequently incorporated in the prediction models (18 studies). The parameters associated with portal hypertension were significant for predicting PHLF in 16 studies and were adopted in the prediction models in 14 studies. Conclusion Parameters related to future liver remnant volume and portal hypertension seem to be facilitating in predicting PHLF.

During 2015–2017, the yearly average of fatalities caused by roadway departure (RwD) crashes was 19,233 in the U.S. Roadway departure crashes, or crashes in which a vehicle crosses an edge line, centerline, or otherwise leaves the traveled way, account for 52 percent of all traffic fatalities in the U.S. during this time. A majority of RwD crashes are run-off-road (ROR) crashes; these are crashes that result in a vehicle crossing the edge line on either side of the roadway. In this study, the research team analyzed seven years (2010–2016) of rural two-lane ROR crash data from Louisiana to better comprehend ROR crashes in refining safety predictions for rural two-lane highways. Statistical model (negative binomial model) and three separate machine learning models (random forest, support vector machine, and Cubist) were applied to determine the best fit models. Overall, Cubist is characterized by a better performance in estimating ROR crashes on rural two-lane roadways. The Cubist approach introduced rules-based safety performance functions (SPFs) for total and fatal and injury crashes. This approach will be beneficial for the safety practitioners in tackling localized issues in crash data with an emphasis on prediction accuracy.

Non-destructive quality assessment of the inner content of potatoes is a key challenge in automatic grading of agricultural quality, especially when potatoes have surface impurities. This study compares different partial least-square regression (PLSR) models for the prediction of soluble solid content (SSC) of potatoes under conditions of surface cleanliness and surface impurities. Smoothing and spectral preprocessing with first-order derivatives and variable sorting for normalization (VSN) can effectively eliminate spectral noise. Variable selection algorithms were used to extract effective variables in order to further optimise the prediction models. The results showed that the method of the variable combination population analysis—iteratively retains informative variables (VCPA-IRIV) is the best method for selecting valid variables, and that the 35-variable VCPA-IRIV-PLSR prediction model could predict the potato SSC with a predictive correlation coefficient (Rp), root-mean-square error of prediction (RMSEP), and residual predictive deviation (RPD) values of 0.831, 0.461○Brix and 1.798, respectively. Therefore, the experimental results show the feasibility and applicability of the proposed SSC prediction method for potatoes with surface impurities using near-infrared spectroscopy.

Highlights • Biochemical failure rates after focal salvage high-dose-rate brachytherapy are high. • Evidence on predictors for failure is limited to whole-gland salvage radiotherapy. • Two prediction models for biochemical failure were developed and internally validated. • Based on pre- and post-salvage characteristics we can identify high-risk patients. • These models can be used for patient counselling at baseline and during follow-up., Background and purpose Magnetic resonance-guided focal salvage high-dose-rate brachytherapy (FS-HDR-BT) for radiorecurrent prostate cancer (PCa) shows low toxicity rates. However, biochemical failure (BF) after treatment occurs frequently. We developed two prediction models for BF (Phoenix definition) with the aim of enhancing patient counselling before FS-HDR-BT and during follow-up. Materials and methods A prospective cohort of 150 radiorecurrent PCa patients treated with FS-HDR-BT between 2013 and 2020 was used for model development and internal validation. Multivariable Cox Proportional Hazards regression was applied. For model 1, only pre-salvage variables were included as candidate predictors. For model 2, additional (post-)salvage characteristics were tested. After calibration, nomograms and webtools were constructed. Finally, three risk groups were identified. Results Sixty-one patients (41%) experienced BF. At baseline (model 1), age, gross tumour volume, pre-salvage PSA, and pre-salvage PSA doubling time (PSADT) were predictive of BF. During follow-up (model 2), age, pre-salvage PSA and PSADT, seminal vesicle involvement, post-salvage time to PSA nadir, and percentage PSA reduction were predictive of BF. The adjusted C-statistics were 0.73 (95% CI: 0.66–0.81) and 0.84 (95% CI: 0.78–0.90), respectively, with acceptable calibration. Estimated 2-year biochemical disease-free survival for the low-, intermediate-, and high-risk groups were 84%, 70%, and 31% (model 1), and 100%, 71%, and 5% (model 2). Conclusion Two models are provided for prediction of BF in patients with radiorecurrent PCa treated with FS-HDR-BT. Based on pre- and post-salvage characteristics, we are able to identify patients with a high risk of BF. These findings can aid patient counselling for FS-HDR-BT.