Ability Of Machine Learning Models Research Articles

On failure prediction and failure identification modeling in a gas turbine system: a survey of classification approaches in a three-class problem

Rapid developments in sensor technology, data processing tools and data storage capability have helped fuel an increased appetite for equipment health monitoring in mechanical systems. As a result, the number of sensors and amount of data collected for health monitoring has grown tremendously. It is hoped that by collecting large quantities of operational data, predictive tools can be developed that will provide operational, maintenance and safety benefits. Data mining and machine learning techniques are important tools in addressing the ensuing challenge of extracting useful results from the data collected.
 In this work, the sensor data from a gas turbine system was analyzed with the objective of failure modeling and prediction. Previous efforts had used a two-class approach for this problem, to distinguish healthy and failed states of the system. In this work, a third class labelled as deteriorated data is added prior to each failure event to explore the ability of machine learning models to provide early warning of upcoming incidents. Several maintenance incidents were recorded by the sensor system in two separate vehicles. Three approaches to selecting training data were used. The first followed a traditional method of randomly selecting data points from all data according to a desired percentage of failed data to include in training, target ratios between failed and healthy data in each data set, as well as target ratios between training and testing data. The second data selection strategy was to consider data related to failure incidents as a whole and select certain incidents to include in training, and the remaining ones to be unseen in testing. The third approach was cross-validation which is typically used as a technique to evaluate how a classifier will perform on unseen data while still using the entirety of the data to train the final classifier. In addition to investigating training and data selection strategies, the effect of hyperparameter optimization was explored as well as the effect of varying the time period of the deteriorated class.
 Using the gas turbine data, which included 7 failure incidents and 76 predictor variables, a variety of classifier models of the system were developed in a three-class problem to differentiate healthy, deteriorated and failed system states. The classifier methods included support vector machines, Gaussian Naïve Bayes, random forest, adaboost, multilayer perceptron, k-nearest neighbor, and XG boost. Ensemble models were also created to leverage all the individual classifier models that were developed. This paper will describe the comprehensive results that were obtained using the various approaches and combinations, highlighting the respective benefits and limitations.

Machine-learning fusion approach for the prediction of atrial fibrillation onset using photoplethysmographic-based smart device

Abstract Background The mobile and wearable technology and artificial technology are rapidly developing for atrial fibrillation (AF) detection. Nonetheless, how these smart technologies combine to improve AF prediction, rather than detection, remains unclear. Objective To investigate a machine-learning fusion approach beyond photoplethysmographic (PPG)-based AF detection for predicting AF onset. Methods There are 2,120,210 individuals who used PPG-based smart devices (Huawei Technologies Co., Ltd., Shenzhen, China) to monitor their pulse rhythm from pre-mAFA study (Huawei heart study) for AF screening across China between October 26, 2018 and March 26, 2021. Two machine-learning (ML) models, short-term model and long-term mode, have been developed based on PPG continuing monitoring signals to predict AF episode at 4h or 8h before AF onset respectively, with supervised ML (XGBoost and LightGBM) techniques, and further optimized with feature extension and hyperparameter optimization, respectively. The probabilities of AF onset calculated by both short-term and long-term models with timely PPG monitoring signals would be further input into Warning Decision Module, combined with prior predicting risk of AF onset, then the real risk ratio of predicting AF onset in advance would be output by the fusion module (Figure). The predictive ability of ML models, classified by time to last AF onset, has been validated among individuals involving in AF screening from pre-mAFA study between March 19 to March 26, compared to identified AF by PPG monitoring. Results There were 6294 (mean age ± standard deviation, 51.6±16.0 years old, 5439 male) individuals with identified AF, with total 142,518 identified AF episodes by PPG. The 107,864 identified AF were utilized to validate ML fusion approach for the prediction of AF, given effective PPG monitoring signals before identified AF episodes for ML models. There were 443,630 PPG monitoring signals before identified AF episodes and 563,309 non-AF PPG monitoring signals of ML models. The sensitivity of ML fusion approach for the prediction of AF onset was 94.04%, the specificity of 96.35%, and the recall of 94.04%, respectively (Table). When time to last AF onset over 24 hours, the sensitivity of ML fusion approach for AF onset at 8h in advance was 85.73%, specificity of 96.62%, and recall of 85.73%, respectively (Table). Conclusion The ML fusion approach based on PPG-smart devices provided the effective early 'alters' of AF onset in advance in this large cohort, which may help the early 'upstream' management of AF. Funding Acknowledgement Type of funding sources: Public grant(s) – National budget only. Main funding source(s): National HealthCare of China (20BJZ26)

Machine learning for the life-time risk prediction of Alzheimer's disease: a systematic review.

Alzheimer’s disease is a neurodegenerative disorder and the most common form of dementia. Early diagnosis may assist interventions to delay onset and reduce the progression rate of the disease. We systematically reviewed the use of machine learning algorithms for predicting Alzheimer’s disease using single nucleotide polymorphisms and instances where these were combined with other types of data. We evaluated the ability of machine learning models to distinguish between controls and cases, while also assessing their implementation and potential biases. Articles published between December 2009 and June 2020 were collected using Scopus, PubMed and Google Scholar. These were systematically screened for inclusion leading to a final set of 12 publications. Eighty-five per cent of the included studies used the Alzheimer's Disease Neuroimaging Initiative dataset. In studies which reported area under the curve, discrimination varied (0.49–0.97). However, more than half of the included manuscripts used other forms of measurement, such as accuracy, sensitivity and specificity. Model calibration statistics were also found to be reported inconsistently across all studies. The most frequent limitation in the assessed studies was sample size, with the total number of participants often numbering less than a thousand, whilst the number of predictors usually ran into the many thousands. In addition, key steps in model implementation and validation were often not performed or unreported, making it difficult to assess the capability of machine learning models.

Predicting Power Conversion Efficiency of Organic Photovoltaics: Models and Data Analysis.

In this paper, the ability of three selected machine learning neural and baseline models in predicting the power conversion efficiency (PCE) of organic photovoltaics (OPVs) using molecular structure information as an input is assessed. The bidirectional long short-term memory (gFSI/BiLSTM), attentive fingerprints (attentive FP), and simple graph neural networks (simple GNN) as well as baseline support vector regression (SVR), random forests (RF), and high-dimensional model representation (HDMR) methods are trained to both the large and computational Harvard clean energy project database (CEPDB) and the much smaller experimental Harvard organic photovoltaic 15 dataset (HOPV15). It was found that the neural-based models generally performed better on the computational dataset with the attentive FP model reaching a state-of-the-art performance with the test set mean squared error of 0.071. The experimental dataset proved much harder to fit, with all of the models exhibiting a rather poor performance. Contrary to the computational dataset, the baseline models were found to perform better than the neural models. To improve the ability of machine learning models to predict PCEs for OPVs, either better computational results that correlate well with experiments or more experimental data at well-controlled conditions are likely required.

Consensus of Feature Selection Methods and Reduced Generalization Gap Model to Improve Diagnosis of Heart Disease

Enhancing the diagnostic ability of Machine Learning models for acceptable prediction in the healthcare community is still a concern. There are critical care disease datasets available online on which researchers have experimented with a different number of instances and features for similar disease prediction. Further, different Machine Learning (ML) models have different preprocessing requirements. Framingham heart disease data is multicollinear and has missing values. Thus, the proposed model aims to explore the differential preprocessing needs of ML models followed by feature selection in consensus with domain experts and feature extraction to resolve multicollinearity issues. Missing values have been imputed differently for each feature. The work also identifies optimal train set size by plotting a learning curve that provides a minimum generalization gap. When testing is done on this hyperparameter tuned model, performance is enhanced with respect to the F score weighted by support and stratification since the data is imbalanced. Experimental results demonstrate improvement in performance metrics, i.e., weighted F score, precision, recall, accuracy up to 3 %, and F1 score by 8 % for Logistic Regression Classifier with the proposed model. Further, the time required for hyperparameter tuning is reduced by 50% for tree-based models, particularly Classification and Regression Tree (CART).

A Machine Learning Model to Successfully Predict Future Diagnosis of Chronic Myelogenous Leukemia With Retrospective Electronic Health Records Data.

Chronic myelogenous leukemia (CML) is a clonal stem cell disorder accounting for 15% of adult leukemias. We aimed to determine if machine learning models could predict CML using blood cell counts prior to diagnosis. We identified patients with a diagnostic test for CML (BCR-ABL1) and at least 6 consecutive prior years of differential blood cell counts between 1999 and 2020 in the largest integrated health care system in the United States. Blood cell counts from different time periods prior to CML diagnostic testing were used to train, validate, and test machine learning models. The sample included 1,623 patients with BCR-ABL1 positivity rate 6.2%. The predictive ability of machine learning models improved when trained with blood cell counts closer to time of diagnosis: 2 to 5 years area under the curve (AUC), 0.59 to 0.67, 0.5 to 1 years AUC, 0.75 to 0.80, at diagnosis AUC, 0.87 to 0.92. Blood cell counts collected up to 5 years prior to diagnostic workup of CML successfully predicted the BCR-ABL1 test result. These findings suggest a machine learning model trained with blood cell counts could lead to diagnosis of CML earlier in the disease course compared to usual medical care.

CPT Data Interpretation Employing Different Machine Learning Techniques

The classification of soils into categories with a similar range of properties is a fundamental geotechnical engineering procedure. At present, this classification is based on various types of cost- and time-intensive laboratory and/or in situ tests. These soil investigations are essential for each individual construction site and have to be performed prior to the design of a project. Since Machine Learning could play a key role in reducing the costs and time needed for a suitable site investigation program, the basic ability of Machine Learning models to classify soils from Cone Penetration Tests (CPT) is evaluated. To find an appropriate classification model, 24 different Machine Learning models, based on three different algorithms, are built and trained on a dataset consisting of 1339 CPT. The applied algorithms are a Support Vector Machine, an Artificial Neural Network and a Random Forest. As input features, different combinations of direct cone penetration test data (tip resistance qc, sleeve friction fs, friction ratio Rf, depth d), combined with “defined”, thus, not directly measured data (total vertical stresses σv, effective vertical stresses σ’v and hydrostatic pore pressure u0), are used. Standard soil classes based on grain size distributions and soil classes based on soil behavior types according to Robertson are applied as targets. The different models are compared with respect to their prediction performance and the required learning time. The best results for all targets were obtained with models using a Random Forest classifier. For the soil classes based on grain size distribution, an accuracy of about 75%, and for soil classes according to Robertson, an accuracy of about 97–99%, was reached.

1088-P: Can Machine Learning Support Decision for Insulin Initiation in Patients with Type 2 Diabetes?

We examined the ability of machine learning models to predict insulin initiation by specialists and examined whether the machine learning approach could support decision-making by general physicians for insulin initiation in patients with type 2 diabetes. Prescription data from diabetes specialists’ patient registries on antidiabetic agents from 2009 to 2015 were used for 4,860 patients who received initial monotherapy with either insulin (293 patients) or non-insulin (4,567 patients) and had laboratory data. Neural network output was insulin initiation ranging from 0 to 1 with a cutoff >0.5 for the dichotomous classification. Accuracy, recall and area under the receiver operation characteristic curves (AUCs) were calculated to compare the ability of machine-learning decisions on insulin initiation to decisions by logistic regression and 22 general physicians. Then, in comparing the ability of machine-learning and logistic regression decisions to decisions made by general physicians, the gold standard was defined as 80% agreement among 9 diabetes specialists based on patients’ information. AUCs and accuracy according to logistic regression were higher than those of machine learning (AUC of logistic regression, 0.89-0.90; AUC of machine learning, 0.67-0.74). Recall (accuracy of the cases receiving insulin) in machine learning was similar to that of logistic regression analysis (recall of logistic regression, 0.05-0.68; recall of machine learning, 0.11-0.52). The accuracy of logistic regression and the machine learning model was 1.00 and 0.86, respectively, for down sampling ratios of 1:8, which was higher than that of general physicians (0.43) for 7 gold standard cases. In conclusion, machine learning had higher accuracy in prediction of insulin initiation compared with that of general physicians defined by diabetes specialists’ choice of the gold standard. Assistance by machine learning may be beneficial to any general physician in deciding upon insulin initiation. Disclosure K. Fujihara: None. H. Sone: Research Support; Self; Astellas Pharma Inc., Eisai Co., Ltd., Kyowa Kirin Co., Ltd., Novo Nordisk, Ono Pharmaceutical Co., Ltd., Taisho Pharmaceutical Co., Ltd., Takeda Pharmaceutical Co. Funding Japan Society for the Promotion of Science (19H04028)

Advances in soil moisture retrieval from multispectral remote sensing using unoccupied aircraft systems and machine learning techniques

Abstract. This study investigates the ability of machine learning models to retrieve the surface soil moisture of a grassland area from multispectral remote sensing carried out using an unoccupied aircraft system (UAS). In addition to multispectral images, we use terrain attributes derived from a digital elevation model and hydrological variables of precipitation and potential evapotranspiration as covariates to predict surface soil moisture. We tested four different machine learning algorithms and interrogated the models to rank the importance of different variables and to understand their relationship with surface soil moisture. All the machine learning algorithms we tested were able to predict soil moisture with good accuracy. The boosted regression tree algorithm was marginally the best, with a mean absolute error of 3.8 % volumetric moisture content. Variable importance analysis revealed that the four most important variables were precipitation, reflectance in the red wavelengths, potential evapotranspiration, and topographic position indices (TPI). Our results demonstrate that the dynamics of soil water status across heterogeneous terrain may be adequately described and predicted by UAS remote sensing and machine learning. Our modeling approach and the variable importance and relationships we have assessed in this study should be useful for management and environmental modeling tasks where spatially explicit soil moisture information is important.

Combination of machine learning and VIRS for predicting soil organic matter

Visible-near-infrared spectroscopy (VIRS) is one of the most promising alternative techniques for soil organic matter (SOM) due to its direct response. In this study, partial least squares regression (PLSR), support vector machine (SVM), artificial neural networks (ANNs), and Cubist combined with VIRS were utilized to develop the calibration model and evaluate the ability of machine learning models to predict soil organic matter content. A total of 190 surface soil samples (earth-cumulic-orthic anthrosols) were collected from the Weihe Plain of Shaanxi Province, China. The Kennard–Stone (KS) algorithm was employed to divide them into calibration and validation data. Moreover, the successive projections algorithm (SPA), competitive adaptive weight weighting algorithm (CARS), and their combination (SPA + CARS) were utilized to select characteristic wavelengths and improve the predictive ability of the model. Different evaluation indices, including root mean square error (RMSE), coefficient of determination (R2), the ratio of the performance to deviation (RPD), and the ratio of performance to interquartile range (RPIQ), were adopted to evaluate the accuracy of the model. In all cases, the AFS-SPA + CARS-Cubist method outperformed the PLSR, SVM, and ANN. For the Cubist model, the Rv2, RPD, and RPIQ ranged from 0.8629 to 0.9782, 0.8720 to 3.0203, and 2.005 to 4.4164, respectively. According to the results, combining VIRS with Cubist could accurately determine the SOM of earth-cumulic-orthic anthrosol soils of the Weihe Plain, China. Furthermore, SPA + CARS provided more precise calibration–validation models than SPA and CARS.

Accurate and Interpretable Machine Learning for Transparent Pricing of Health Insurance Plans

Health insurance companies cover half of the United States population through commercial employer-sponsored health plans and pay 1.2 trillion US dollars every year to cover medical expenses for their members. The actuary and underwriter roles at a health insurance company serve to assess which risks to take on and how to price those risks to ensure profitability of the organization. While Bayesian hierarchical models are the current standard in the industry to estimate risk, interest in machine learning as a way to improve upon these existing methods is increasing. Lumiata, a healthcare analytics company, ran a study with a large health insurance company in the United States. We evaluated the ability of machine learning models to predict the per member per month cost of employer groups in their next renewal period, especially those groups who will cost less than 95\% of what an actuarial model predicts (groups with "concession opportunities"). We developed a sequence of two models, an individual patient-level and an employer-group-level model, to predict the annual per member per month allowed amount for employer groups, based on a population of 14 million patients. Our models performed 20\% better than the insurance carrier's existing pricing model, and identified 84\% of the concession opportunities. This study demonstrates the application of a machine learning system to compute an accurate and fair price for health insurance products and analyzes how explainable machine learning models can exceed actuarial models' predictive accuracy while maintaining interpretability.

A first-principles and machine learning combined method to investigate the interfacial friction between corrugated graphene

Simulating the frictional properties of complex interfaces is computational resource consuming. In this paper, we propose a density functional theory (DFT) calculation combined machine learning (ML) strategy to investigate the sliding potential energy corrugation between geometrical corrugated graphene (Gr) sheets. By the aid of few DFT calculations and geometrical descriptors Σr −n (n = 1, 2, 6, 12), the trained ML models can accurately predict the sliding potential evolutions of Gr/Pt and Gr/Re systems. To be specific, based on DFT calculations of sliding along [110] direction, the trained linear regression (LIN) models can properly give out the potential energy evolution along the [100] direction with deviation less than 5%. By the dataset of given distances (9.3 Å, 9.65 Å and 10 Å) between two Re monolayers in Gr/Re systems, LIN and Bayesian ridge regression (BR) models can quantitatively predict the potential energy evolution of unknown distances (9.2 Å, 9.4 Å, 9.5 Å and 9.6 Å). The predicted magnitudes of potential energy corrugations by BR model divert less than 3 meV Å−2 from DFT calculations. The prediction results for extrapolated distances (9.0 Å and 9.1 Å) deviate notably, but the extension of training dataset effectively improves the predictive ability of ML models, especially for the LIN model. Thus, the supposed strategy could become an effective method to investigate the frictional characteristics of complex interfaces.

Predicting Australian Adults at High Risk of Cardiovascular Disease Mortality Using Standard Risk Factors and Machine Learning.

Effective cardiovascular disease (CVD) prevention relies on timely identification and intervention for individuals at risk. Conventional formula-based techniques have been demonstrated to over- or under-predict the risk of CVD in the Australian population. This study assessed the ability of machine learning models to predict CVD mortality risk in the Australian population and compare performance with the well-established Framingham model. Data is drawn from three Australian cohort studies: the North West Adelaide Health Study (NWAHS), the Australian Diabetes, Obesity, and Lifestyle study, and the Melbourne Collaborative Cohort Study (MCCS). Four machine learning models for predicting 15-year CVD mortality risk were developed and compared to the 2008 Framingham model. Machine learning models performed significantly better compared to the Framingham model when applied to the three Australian cohorts. Machine learning based models improved prediction by 2.7% to 5.2% across three Australian cohorts. In an aggregated cohort, machine learning models improved prediction by up to 5.1% (area-under-curve (AUC) 0.852, 95% CI 0.837–0.867). Net reclassification improvement (NRI) was up to 26% with machine learning models. Machine learning based models also showed improved performance when stratified by sex and diabetes status. Results suggest a potential for improving CVD risk prediction in the Australian population using machine learning models.

Machine Learning Approach to Decision Making for Insulin Initiation in Japanese Patients With Type 2 Diabetes (JDDM 58): Model Development and Validation Study.

BackgroundApplications of machine learning for the early detection of diseases for which a clear-cut diagnostic gold standard exists have been evaluated. However, little is known about the usefulness of machine learning approaches in the decision-making process for decisions such as insulin initiation by diabetes specialists for which no absolute standards exist in clinical settings.ObjectiveThe objectives of this study were to examine the ability of machine learning models to predict insulin initiation by specialists and whether the machine learning approach could support decision making by general physicians for insulin initiation in patients with type 2 diabetes.MethodsData from patients prescribed hypoglycemic agents from December 2009 to March 2015 were extracted from diabetes specialists’ registries, resulting in a sample size of 4860 patients who had received initial monotherapy with either insulin (n=293) or noninsulin (n=4567). Neural network output was insulin initiation ranging from 0 to 1 with a cutoff of >0.5 for the dichotomous classification. Accuracy, recall, and area under the receiver operating characteristic curve (AUC) were calculated to compare the ability of machine learning models to make decisions regarding insulin initiation to the decision-making ability of logistic regression and general physicians. By comparing the decision-making ability of machine learning and logistic regression to that of general physicians, 7 cases were chosen based on patient information as the gold standard based on the agreement of 8 of the 9 specialists.ResultsThe AUCs, accuracy, and recall of logistic regression were higher than those of machine learning (AUCs of 0.89-0.90 for logistic regression versus 0.67-0.74 for machine learning). When the examination was limited to cases receiving insulin, discrimination by machine learning was similar to that of logistic regression analysis (recall of 0.05-0.68 for logistic regression versus 0.11-0.52 for machine learning). Accuracies of logistic regression, a machine learning model (downsampling ratio of 1:8), and general physicians were 0.80, 0.70, and 0.66, respectively, for 43 randomly selected cases. For the 7 gold standard cases, the accuracies of logistic regression and the machine learning model were 1.00 and 0.86, respectively, with a downsampling ratio of 1:8, which were higher than the accuracy of general physicians (ie, 0.43).ConclusionsAlthough we found no superior performance of machine learning over logistic regression, machine learning had higher accuracy in prediction of insulin initiation than general physicians, defined by diabetes specialists’ choice of the gold standard. Further study is needed before the use of machine learning–based decision support systems for insulin initiation can be incorporated into clinical practice.

Machine learning approaches to predicting amyloid status using data from an online research and recruitment registry: The Brain Health Registry.

IntroductionThis study investigated the extent to which subjective and objective data from an online registry can be analyzed using machine learning methodologies to predict the current brain amyloid beta (Aβ) status of registry participants.MethodsWe developed and optimized machine learning models using data from up to 664 registry participants. Models were assessed on their ability to predict Aβ positivity using the results of positron emission tomography as ground truth.ResultsStudy partner–assessed Everyday Cognition score was preferentially selected for inclusion in the models by a feature selection algorithm during optimization.DiscussionOur results suggest that inclusion of study partner assessments would increase the ability of machine learning models to predict Aβ positivity.

Machine Learning Based Predictions of Dissolved Oxygen in a Small Coastal Embayment

Coastal dissolved oxygen (DO) concentrations have a profound impact on nearshore ecosystems and, in recent years, there has been an increased prevalance of low DO hypoxic events that negatively impact nearshore organisms. Even with advanced numerical models, accurate prediction of coastal DO variability is challenging and computationally expensive. Here, we apply machine learning techniques in order to reconstruct and predict nearshore DO concentrations in a small coastal embayment while using a comprehensive set of nearshore and offshore measurements and easily measured input (training) parameters. We show that both random forest regression (RFR) and support vector regression (SVR) models accurately reproduce both the offshore DO and nearshore DO with extremely high accuracy. In general, RFR consistently peformed slightly better than SVR, the latter of which was more difficult to tune and took longer to train. Although each of the nearshore datasets were able to accurately predict DO values using training data from the same site, the model only had moderate success when using training data from one site to predict DO at another site, which was likely due to the the complexities in the underlying dynamics across the sites. We also show that high accuracy can be achieved with relatively little training data, highlighting a potential application for correcting time series with missing DO data due to quality control or sensor issues. This work establishes the ability of machine learning models to accurately reproduce DO concentrations in both offshore and nearshore coastal waters, with important implications for the ability to detect and indirectly measure coastal hypoxic events in near real-time. Future work should explore the ability of machine learning models in order to accurately forecast hypoxic events.

Machine Learning Models for Predicting the Ammonium Concentration in Alluvial Groundwaters

Considering the great importance of groundwater quality for water supply, in the last decade, significant scientific attention has been devoted to nitrate reduction transformation pathways and nitrogen conservation in groundwaters in the form of ammonium. To evaluate and assess the ability of machine learning models to predict the ammonium concentration, four machine learning models were applied: a three-layer neural network (NN), a deep neural network (DNN), and two variants of support vector regression (SVR) models: with linear and with Gaussian radial basis function kernel. A dataset of 322 samples with 13 predictor variables representing selected parameters responsible for oxidative/reductive nitrogen transformations in shallow alluvial groundwater was acquired from measurements in 55 monitoring wells during a 6-year monitoring period (2011–2016) in Serbia. Applied principal component analysis and cluster analysis gave an insight into conditionality and relations between the selected parameters, distinguishing four main factors, which explained 70.97% of total variance, and classifying examined objects by similarity. Extracted factors correlated the concentration patterns, implying the main nitrogen transformations in examined groundwater. The machine learning models were successfully applied for predicting the ammonium concentration with high determination coefficients (R2) in tests: 0.84 for DNN and 0.64 for NN, while the SVR did not prove to be adequate with the best R2 of 0.24.

A review of machine learning applications in wildfire science and management

Artificial intelligence has been applied in wildfire science and management since the 1990s, with early applications including neural networks and expert systems. Since then, the field has rapidly progressed congruently with the wide adoption of machine learning (ML) methods in the environmental sciences. Here, we present a scoping review of ML applications in wildfire science and management. Our overall objective is to improve awareness of ML methods among wildfire researchers and managers, as well as illustrate the diverse and challenging range of problems in wildfire science available to ML data scientists. To that end, we first present an overview of popular ML approaches used in wildfire science to date and then review the use of ML in wildfire science as broadly categorized into six problem domains, including (i) fuels characterization, fire detection, and mapping; (ii) fire weather and climate change; (iii) fire occurrence, susceptibility, and risk; (iv) fire behavior prediction; (v) fire effects; and (vi) fire management. Furthermore, we discuss the advantages and limitations of various ML approaches relating to data size, computational requirements, generalizability, and interpretability, as well as identify opportunities for future advances in the science and management of wildfires within a data science context. In total, to the end of 2019, we identified 300 relevant publications in which the most frequently used ML methods across problem domains included random forests, MaxEnt, artificial neural networks, decision trees, support vector machines, and genetic algorithms. As such, there exists opportunities to apply more current ML methods — including deep learning and agent-based learning — in the wildfire sciences, especially in instances involving very large multivariate datasets. We must recognize, however, that despite the ability of ML models to learn on their own, expertise in wildfire science is necessary to ensure realistic modelling of fire processes across multiple scales, while the complexity of some ML methods such as deep learning requires a dedicated and sophisticated knowledge of their application. Finally, we stress that the wildfire research and management communities play an active role in providing relevant, high-quality, and freely available wildfire data for use by practitioners of ML methods.

Machine Learning Models with Preoperative Risk Factors and Intraoperative Hypotension Parameters Predict Mortality After Cardiac Surgery

Machine learning models used to predict postoperative mortality rarely include intraoperative factors. Several intraoperative factors like hypotension (IOH), vasopressor-inotropes, and cardiopulmonary bypass (CPB) time are significantly associated with postoperative outcomes. The authors explored the ability of machine learning models incorporating intraoperative risk factors to predict mortality after cardiac surgery. Retrospective study. Tertiary hospital. A total of 5,015 adults who underwent cardiac surgery from 2008 to 2016. None. The intraoperative phase was divided into the following: (1) CPB, (2) outside CPB, and (3) total surgery for quantifying IOH only. Phase-specific IOH parameters (area under the curve for mean arterial pressure <65 mmHg), vasopressor-inotropes (norepinephrine equivalents), duration, and cross-clamp time, along with preoperative risk factors ,were incorporated into the models. The primary outcome was mortality. The following 5 models were applied to 3 intraoperative phases separately: (1) logistic regression, (2) random forests, (3) neural networks, (4) support vector machines, and (5) extreme gradient boosting (XGB). Mortality was predicted using area under the receiver operating characteristic curve. Of 5,015 patients included, 112 (2.2%) died. XGB model from the outside-CPB phase predicted mortality better with area under the receiver operating characteristic curve, 95% confidence interval (CI): 0.88(0.83-0.94); positive predictive value, 0.10(0.06-0.15); specificity 0.85 (0.83-0.87) and sensitivity 0.75 (0.57-0.90). XGB machine learning model from IOH outside the CPB phase seemed to offer a better discrimination, sensitivity, specificity, and positive predictive value compared with other models. Machine learning models incorporating intraoperative adverse factors might offer better predictive ability for risk stratification and triaging of patients after cardiac surgery.

A Machine Learning Model to Detect Flow Disturbances during Manufacturing of Composites by Liquid Moulding

In this work, a supervised machine learning (ML) model was developed to detect flow disturbances caused by the presence of a dissimilar material region in liquid moulding manufacturing of composites. The machine learning model was designed to predict the position, size and relative permeability of an embedded rectangular dissimilar material region through use of only the signals corresponding to an array of pressure sensors evenly distributed on the mould surface. The burden of experimental tests required to train in an efficient manner such predictive models is so high that favours its substitution with synthetically-generated simulation datasets. A regression model based on the use of convolutional neural networks (CNN) was developed and trained with data generated from mould-filling simulations carried out through use of OpenFoam as numerical solver. The evolution of the pressure sensors through the filling time was stored and used as grey-level images containing information regarding the pressure, the sensor location within the mould and filling time. The trained CNN model was able to recognise the presence of a dissimilar material region from the data used as inputs, meeting accuracy expectation in terms of detection. The purpose of this work was to establish a general framework for fully-synthetic-trained machine learning models to address the occurrence of manufacturing disturbances without placing emphasis on its performance, robustness and optimization. Accuracy and model robustness were also addressed in the paper. The effect of noise signals, pressure sensor network size, presence of different shape dissimilar regions, among others, were analysed in detail. The ability of ML models to examine and overcome complex physical and engineering problems such as defects produced during manufacturing of materials and parts is particularly innovative and highly aligned with Industry 4.0 concepts.