Search

Backgrounds: The in-hospital mortality in lung cancer patients admitted to intensive care unit (ICU) is extremely high. This study intended to adopt machine learning algorithm models to predict in-hospital mortality of critically ill lung cancer for providing relative information in clinical decision-making. Methods: Data were extracted from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) for a training cohort and data extracted from the Medical Information Mart for eICU Collaborative Research Database (eICU-CRD) database for a validation cohort. Logistic regression, random forest, decision tree, light gradient boosting machine (LightGBM), eXtreme gradient boosting (XGBoost), and an ensemble (random forest+LightGBM+XGBoost) model were used for prediction of in-hospital mortality and important feature extraction. The AUC (area under receiver operating curve), accuracy, F1 score and recall were used to evaluate the predictive performance of each model. Shapley Additive exPlanations (SHAP) values were calculated to evaluate feature importance of each feature. Results: Overall, there were 653 (24.8%) in-hospital mortality in the training cohort, and 523 (21.7%) in-hospital mortality in the validation cohort. Among the six machine learning models, the ensemble model achieved the best performance. The top 5 most influential features were the sequential organ failure assessment (SOFA) score, albumin, the oxford acute severity of illness score (OASIS) score, anion gap and bilirubin in random forest and XGBoost model. The SHAP summary plot was used to illustrate the positive or negative effects of the top 15 features attributed to the XGBoost model. Conclusion: The ensemble model performed best and might be applied to forecast in-hospital mortality of critically ill lung cancer patients, and the SOFA score was the most important feature in all models. These results might offer valuable and significant reference for ICU clinicians' decision-making in advance. [ABSTRACT FROM AUTHOR]

Unoccupied aerial systems (UAS) based high throughput phenotyping studies require further investigation to combine different environments and planting times into one model. Here 100 elite breeding hybrids of maize (Zea mays L.) were evaluated in two environment trials–one with optimal planting and irrigation (IHOT), and one dryland with delayed planting (DHOT). RGB (Red-Green-Blue) based canopy height measurement (CHM) and vegetation indices (VIs) were estimated from a UAS platform. Time series and cumulative VIs, by both summation (ΣVI-SUMs) and area under the curve (ΣVI-AUCs), were fit via machine learning regression modeling (random forest, linear, ridge, lasso, elastic net regressions) to estimate grain yield. VIs were more valuable predictors of yield to combine different environments than CHM. Time series VIs and CHM produced high accuracies (~68–72%), but inconsistent models. A little sacrifice in accuracy (~60–65%) produced consistent models using ΣVI-SUMs and CHM during pre-reproductive vegetative growth. Absence of VIs produced poorer accuracies (by about ~5–10%). Normalized difference type VIs produced maximum accuracies, and flowering times were the best times for UAS data acquisition. This study suggests that the best yielding varieties can be accurately predicted in new environments at or before flowering when combining multiple temporal flights and predictors. [ABSTRACT FROM AUTHOR]

Machine learning method has become a popular, convenient and efficient computing tool applied to many industries at present. Multi-hole pressure probe is an important technique widely used in flow vector measurement. It is a new attempt to integrate machine learning method into multi-hole probe measurement. In this work, six typical supervised learning methods in scikit-learn library are selected for parameter adjustment at first. Based on the optimal parameters, a comprehensive evaluation is conducted from four aspects: prediction accuracy, prediction efficiency, feature sensitivity and robustness on the failure of some hole port. As results, random forests and K-nearest neighbors' algorithms have the better comprehensive prediction performance. Compared with the in-house traditional algorithm, the machine learning algorithms have the great advantages in the computational efficiency and the convenience of writing code. Multi-layer perceptron and support vector machines are the most time-consuming algorithms among the six algorithms. The prediction accuracy of all the algorithms is very sensitive to the features. Using the features based on the physical knowledge can obtain a high accuracy predicted results. Finally, KNN algorithm is successfully applied to field measurements on the angle of attack of a wind turbine blades. These findings provided a new reference for the application of machine learning method in multi-hole probe calibration and measurement. [ABSTRACT FROM AUTHOR]

Our aim was to predict future high-cost patients with machine learning using healthcare claims data. We applied a random forest (RF), a gradient boosting machine (GBM), an artificial neural network (ANN) and a logistic regression (LR) to predict high-cost patients in the following year. Therefore, we exploited routinely collected sickness funds claims and cost data of the years 2016, 2017 and 2018. Various specifications of each algorithm were trained and cross-validated on training data (n = 20,984) with claims and cost data from 2016 and outcomes from 2017. The best performing specifications of each algorithm were selected based on validation dataset performance. For performance comparison, selected models were applied to unforeseen data with features of the year 2017 and outcomes of the year 2018 (n = 21,146). The RF was the best performing algorithm measured by the area under the receiver operating curve (AUC) with a value of 0.883 (95% confidence interval (CI): 0.872–0.893) on test data, followed by the GBM (AUC = 0.878; 95% CI: 0.867–0.889). The ANN (AUC = 0.846; 95% CI: 0.834–0.857) and LR (AUC = 0.839; 95% CI: 0.826–0.852) were significantly outperformed by the GBM and the RF. All ML algorithms and the LR performed ´good´ (i.e. 0.9 > AUC ≥ 0.8). We were able to develop machine learning models that predict high-cost patients with 'good' performance facilitating routinely collected sickness fund claims and cost data. We found that tree-based models performed best and outperformed the ANN and LR. [ABSTRACT FROM AUTHOR]

This study constructs machine learning algorithms that are trained to classify samples using sound symbolism, and then it reports on an experiment designed to measure their understanding against human participants. Random forests are trained using the names of Pokémon, which are fictional video game characters, and their evolutionary status. Pokémon undergo evolution when certain in-game conditions are met. Evolution changes the appearance, abilities, and names of Pokémon. In the first experiment, we train three random forests using the sounds that make up the names of Japanese, Chinese, and Korean Pokémon to classify Pokémon into pre-evolution and post-evolution categories. We then train a fourth random forest using the results of an elicitation experiment whereby Japanese participants named previously unseen Pokémon. In Experiment 2, we reproduce those random forests with name length as a feature and compare the performance of the random forests against humans in a classification experiment whereby Japanese participants classified the names elicited in Experiment 1 into pre-and post-evolution categories. Experiment 2 reveals an issue pertaining to overfitting in Experiment 1 which we resolve using a novel cross-validation method. The results show that the random forests are efficient learners of systematic sound-meaning correspondence patterns and can classify samples with greater accuracy than the human participants. [ABSTRACT FROM AUTHOR]

Objective: Ischemic stroke (IS) with subsequent cerebrocardiac syndrome (CCS) has a poor prognosis. We aimed to investigate electrocardiogram (ECG) changes after IS with artificial intelligence (AI). Methods: We collected ECGs from a healthy population and patients with IS, and then analyzed participant demographics and ECG parameters to identify abnormal features in post-IS ECGs. Next, we trained the convolutional neural network (CNN), random forest (RF) and support vector machine (SVM) models to automatically detect the changes in the ECGs; Additionally, We compared the CNN scores of good prognosis (mRS ≤ 2) and poor prognosis (mRS > 2) to assess the prognostic value of CNN model. Finally, we used gradient class activation map (Grad-CAM) to localize the key abnormalities. Results: Among the 3506 ECGs of the IS patients, 2764 ECGs (78.84%) led to an abnormal diagnosis. Then we divided ECGs in the primary cohort into three groups, normal ECGs (N-Ns), abnormal ECGs after the first ischemic stroke (A-ISs), and normal ECGs after the first ischemic stroke (N-ISs). Basic demographic and ECG parameter analyses showed that heart rate, QT interval, and P-R interval were significantly different between 673 N-ISs and 3546 N-Ns (p < 0.05). The CNN has the best performance among the three models in distinguishing A-ISs and N-Ns (AUC: 0.88, 95%CI = 0.86–0.90). The prediction scores of the A-ISs and N-ISs obtained from the all three models are statistically different from the N-Ns (p < 0.001). Futhermore, the CNN scores of the two groups (mRS > 2 and mRS ≤ 2) were significantly different (p < 0.05). Finally, Grad-CAM revealed that the V4 lead may harbor the highest probability of abnormality. Conclusion: Our study showed that a high proportion of post-IS ECGs harbored abnormal changes. Our CNN model can systematically assess anomalies in and prognosticate post-IS ECGs. [ABSTRACT FROM AUTHOR]

Objective: The objective of this study is twofold. First, we seek to understand the characteristics of the multimorbid population that needs hospital care by using all diagnoses information (ICD-10 codes) and two aggregated multimorbidity and frailty scores. Second, we use machine learning prediction models on these multimorbid patients characteristics to predict rehospitalization within 30 and 365 days and their length of stay. Methods: This study was conducted on 8 882 anonymized patients hospitalized at the University Hospital of Saint-Étienne. A descriptive statistical analysis was performed to better understand the characteristics of the patient population. Multimorbidity was measured using raw diagnoses information and two specific scores based on clusters of diagnoses: the Hospital Frailty Risk Score and the Calderon-Larrañaga index. Based on these variables different machine learning models (Decision Tree, Random forest and k-nearest Neighbors) were used to predict near future rehospitalization and length of stay (LoS). Results: The use of random forest algorithms yielded better performance to predict both 365 and 30 days rehospitalization and using the diagnoses ICD-10 codes directly was significantly more efficient. However, using the Calderon-Larrañaga's clusters of diagnoses can be used as an efficient substitute for diagnoses information for predicting readmission. The predictive power of the algorithms is quite low on length of stay indicator. Conclusion: Using machine learning techniques using patients' diagnoses information and Calderon-Larrañaga's score yielded efficient results to predict hospital readmission of multimorbid patients. These methods could help improve the management of care of multimorbid patients in hospitals. [ABSTRACT FROM AUTHOR]

Market making is a high-frequency trading problem for which solutions based on reinforcement learning (RL) are being explored increasingly. This paper presents an approach to market making using deep reinforcement learning, with the novelty that, rather than to set the bid and ask prices directly, the neural network output is used to tweak the risk aversion parameter and the output of the Avellaneda-Stoikov procedure to obtain bid and ask prices that minimise inventory risk. Two further contributions are, first, that the initial parameters for the Avellaneda-Stoikov equations are optimised with a genetic algorithm, which parameters are also used to create a baseline Avellaneda-Stoikov agent (Gen-AS); and second, that state-defining features forming the RL agent's neural network input are selected based on their relative importance by means of a random forest. Two variants of the deep RL model (Alpha-AS-1 and Alpha-AS-2) were backtested on real data (L2 tick data from 30 days of bitcoin–dollar pair trading) alongside the Gen-AS model and two other baselines. The performance of the five models was recorded through four indicators (the Sharpe, Sortino and P&L-to-MAP ratios, and the maximum drawdown). Gen-AS outperformed the two other baseline models on all indicators, and in turn the two Alpha-AS models substantially outperformed Gen-AS on Sharpe, Sortino and P&L-to-MAP. Localised excessive risk-taking by the Alpha-AS models, as reflected in a few heavy dropdowns, is a source of concern for which possible solutions are discussed. [ABSTRACT FROM AUTHOR]

Yellow fever virus (YFV) is the agent of the most severe mosquito-borne disease in the tropics. Recently, Brazil suffered major YFV outbreaks with a high fatality rate affecting areas where the virus has not been reported for decades, consisting of urban areas where a large number of unvaccinated people live. We developed a machine learning framework combining three different algorithms (XGBoost, random forest and regularized logistic regression) to analyze YFV genomic sequences. This method was applied to 56 YFV sequences from human infections and 27 from non-human primate (NHPs) infections to investigate the presence of genetic signatures possibly related to disease severity (in human related sequences) and differences in PCR cycle threshold (Ct) values (in NHP related sequences). Our analyses reveal four non-synonymous single nucleotide variations (SNVs) on sequences from human infections, in proteins NS3 (E614D), NS4a (I69V), NS5 (R727G, V643A) and six non-synonymous SNVs on NHP sequences, in proteins E (L385F), NS1 (A171V), NS3 (I184V) and NS5 (N11S, I374V, E641D). We performed comparative protein structural analysis on these SNVs, describing possible impacts on protein function. Despite the fact that the dataset is limited in size and that this study does not consider virus-host interactions, our work highlights the use of machine learning as a versatile and fast initial approach to genomic data exploration. [ABSTRACT FROM AUTHOR]

Acute coronary syndrome (ACS) is a serious cardiovascular disease that can lead to cardiac arrest if not diagnosed promptly. However, in the actual diagnosis and treatment of ACS, there will be a large number of redundant related features that interfere with the judgment of professionals. Further, existing methods have difficulty identifying high-quality ACS features from these data, and the interpretability work is insufficient. In response to this problem, this paper uses a hybrid feature selection method based on gradient boosting trees and recursive feature elimination with cross-validation (RFECV) to reduce ACS feature redundancy and uses interpretable feature learning for feature selection to retain the most discriminative features. While reducing the feature set search space, this method can balance model simplicity and learning performance to select the best feature subset. We leverage the interpretability of gradient boosting trees to aid in understanding key features of ACS, linking the eigenvalue meaning of instances to model risk predictions to provide interpretability for the classifier. The data set used in this paper is patient records after percutaneous coronary intervention (PCI) in a tertiary hospital in Fujian Province, China from 2016 to 2021. In this paper, we experimentally explored the impact of our method on ACS risk prediction. We extracted 25 key variables from 430 complex ACS medical features, with a feature reduction rate of 94.19%, and identified 5 key ACS factors. Compared with different baseline methods (Logistic Regression, Random Forest, Gradient Boosting, Extreme Gradient Boosting, Multilayer Perceptron, and 1D Convolutional Networks), the results show that our method achieves the highest Accuracy of 98.8%. [ABSTRACT FROM AUTHOR]

Objective: To explore if random forest (RF) model can predict the prognosis of hospital-acquired Klebsiella pneumoniae infection as well as traditional logistic regression(LR) model. Methods: A total of 254 cases of hospital-acquired Klebsiella pneumoniae infection in a tertiary hospital in Beijing from January 2016 to December 2020 were retrospectively collected. Appropriate influencing factors were selected by referring to relevant articles from the aspects of basic clinical information and contact history before infection, and divided into a training set and a test set. Both the RF and LR models were trained by the training set, and using testing set to compare these two models. Results: The prediction accuracy of the LR model was 87.0%, the true positive rate of the LR model was 94.7%; the false negative rate of the LR model was 5.3%; the false positive rate of the LR model was 35%; the true negative rate of the LR model was 65%; the sensitivity of the LR model for the prognosis prediction of hospital-acquired Klebsiella pneumoniae infection was 94.7%; and the specificity was 65%. The prediction accuracy of the RF model was 89.6%; the true positive rate of the RF model was 92.1%; the false negative rate of the RF model was 7.9%; the false positive rate of the RF model was 21.4%; the true negative rate of the RF model was 78.6%; the sensitivity of the RF model for the prognosis prediction of hospital-acquired Klebsiella pneumoniae infection was 92.1%; and the specificity was 78.6%. ROC curve shows that the area under curve(AUC) of the LR model was 0.91, and that of the RF model was 0.95. Conclusion: The RF model has higher specificity, sensitivity, and accuracy for the prognostic prediction of hospital-acquired Klebsiella pneumoniae infection than the LR model and has greater clinical application prospects. [ABSTRACT FROM AUTHOR]

Remote sensing can be used to map tillage practices at large spatial and temporal scales. However, detecting such management practices in smallholder systems is challenging given that the size of fields is smaller than historical readily-available satellite imagery. In this study we used newer, higher-resolution satellite data from Sentinel-1, Sentinel-2, and Planet to map tillage practices in the Eastern Indo-Gangetic Plains in India. We specifically tested the classification performance of single sensor and multiple sensor random forest models, and the impact of spatial, temporal, or spectral resolution on classification accuracy. We found that when considering a single sensor, the model that used Planet imagery (3 m) had the highest classification accuracy (86.55%) while the model that used Sentinel-1 data (10 m) had the lowest classification accuracy (62.28%). When considering sensor combinations, the model that used data from all three sensors achieved the highest classification accuracy (87.71%), though this model was not statistically different from the Planet only model when considering 95% confidence intervals from bootstrap analyses. We also found that high levels of accuracy could be achieved by only using imagery from the sowing period. Considering the impact of spatial, temporal, and spectral resolution on classification accuracy, we found that improved spatial resolution from Planet contributed the most to improved classification accuracy. Overall, it is possible to use readily-available, high spatial resolution satellite data to map tillage practices of smallholder farms, even in heterogeneous systems with small field sizes. [ABSTRACT FROM AUTHOR]

An important step to achieve greater financial inclusion is to increase the acceptance and usage of digital payments. Although consumer adoption of digital payments has improved dramatically globally, the acceptance and usage of digital payments for micro, small, and medium-sized retailers (MSMRs) remain challenging. Using random forest estimation, we identify 14 key predictors out of 190 variables with the largest predictive power for MSMR adoption and usage of digital payments. Using conditional inference trees, we study the importance of sequencing and interactions of various factors such as public policy initiatives, technological advancements, and private sector incentives. We find that in countries with low POS terminal adoption, killer applications such as mobile phone payment apps increase the likelihood of P2B digital transactions. We also find the likelihood of digital P2B payments at MSMRs increases when MSMRs pay their employees and suppliers digitally. The level of ownership of basic financial accounts by consumers and the size of the shadow economy are also important predictors of greater adoption and usage of digital payments. Using causal forest estimation, we find a positive and economically significant marginal effect for merchant and consumer fiscal incentives on POS terminal adoption on average. When countries implement financial inclusion initiatives, POS terminal adoption increases significantly and MSMRs' share of P2B digital payments also increases. Merchant and consumer fiscal incentives also increase MSMRs' share of P2B electronic payments. [ABSTRACT FROM AUTHOR]

Successive plantings of Prunus species produce suboptimal growth and yield in many California soils due to a poorly understood soilborne disease complex, Prunus replant disease (PRD). We explored the hypothesis that PRD is mediated by microbial taxa in roots of Nemaguard peach, a rootstock for almond and other stone fruits. In a greenhouse bioassay, portions of 10 replant soils were treated with fumigation or pasteurization or left untreated as a control before being planted with peach seedlings. Ten weeks after planting, seedlings were considered PRD-affected if their top fresh weights in the control were significantly reduced, compared to the weights in pasteurization and fumigation treatments; plants with equivalent top weights in all treatments were considered to be non-affected. The roots were washed from the soil, frozen, extracted for total DNA, and used for metabarcoding of rRNA gene amplicons from bacteria, fungi, and oomycetes. High-throughput amplicon sequencing revealed that root microbial community shifts resulted from preplant treatments, and specific taxa were associated with PRD induction among controls. Random forest (RF) analysis discriminated effectively between PRD-affected and non-affected root communities. Among the 30 RF top-ranked amplicon sequence variant (ASV) predictors, 26 were bacteria, two were oomycetes, and two were fungi. Among them, only Streptomyces scabiei, Steroidobacter denitrificans, Streptomyces bobili, and Pythium mamillatum had root abundances ≥5% that were either associated positively (former two ASVs) or negatively (latter two) with PRD. Thus, our findings were consistent with microbial mediation of PRD in roots and suggested taxa that may be involved in the mediation. [ABSTRACT FROM AUTHOR]

Background: Multimorbidity is a worldwide concern related to greater disability, worse quality of life, and mortality. The early prediction is crucial for preventive strategies design and integrative medical practice. However, knowledge about how to predict multimorbidity is limited, possibly due to the complexity involved in predicting multiple chronic diseases. Methods: In this study, we present the use of a machine learning approach to build cost-effective multimorbidity prediction models. Based on predictors easily obtainable in clinical practice (sociodemographic, clinical, family disease history and lifestyle), we build and compared the performance of seven multilabel classifiers (multivariate random forest, and classifier chain, binary relevance and binary dependence, with random forest and support vector machine as base classifiers), using a sample of 15105 participants from the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). We developed a web application for the building and use of prediction models. Results: Classifier chain with random forest as base classifier performed better (accuracy = 0.34, subset accuracy = 0.15, and Hamming Loss = 0.16). For different feature sets, random forest based classifiers outperformed those based on support vector machine. BMI, blood pressure, sex, and age were the features most relevant to multimorbidity prediction. Conclusions: Our results support the choice of random forest based classifiers for multimorbidity prediction. [ABSTRACT FROM AUTHOR]

With the advent of single cell/nucleus RNA sequencing (sc/snRNA-seq), the field of cell phenotyping is now a data-driven exercise providing statistical evidence to support cell type/state categorization. However, the task of classifying cells into specific, well-defined categories with the empirical data provided by sc/snRNA-seq remains nontrivial due to the difficulty in determining specific differences between related cell types with close transcriptional similarities, resulting in challenges with matching cell types identified in separate experiments. To investigate possible approaches to overcome these obstacles, we explored the use of supervised machine learning methods—logistic regression, support vector machines, random forests, neural networks, and light gradient boosting machine (LightGBM)–as approaches to classify cell types using snRNA-seq datasets from human brain middle temporal gyrus (MTG) and human kidney. Classification accuracy was evaluated using an F-beta score weighted in favor of precision to account for technical artifacts of gene expression dropout. We examined the impact of hyperparameter optimization and feature selection methods on F-beta score performance. We found that the best performing model for granular cell type classification in both datasets is a multinomial logistic regression classifier and that an effective feature selection step was the most influential factor in optimizing the performance of the machine learning pipelines. [ABSTRACT FROM AUTHOR]

The objective of this work is to examine disparities in the completion of substance use disorder treatment in the U.S. Our data is from the Treatment Episode Dataset Discharge (TEDS-D) datasets from the U.S. Substance Abuse and Mental Health Services Administration (SAMHSA) for 2017–2019. We apply a two-stage virtual twins model (random forest + decision tree) where, in the first stage (random forest), we determine differences in treatment completion probability associated with race/ethnicity, income source, no co-occurrence of mental health disorders, gender (biological), no health insurance, veteran status, age, and primary substance (alcohol or opioid). In the second stage (decision tree), we identify subgroups associated with probability differences, where such subgroups are more or less likely to complete treatment. We find the subgroups most likely to complete substance use disorder treatment, when the subgroup represents more than 1% of the sample, are those with no mental health condition co-occurrence (4.8% more likely when discharged from an ambulatory outpatient treatment program, representing 62% of the sample; and 10% more likely for one of the more specifically defined subgroups representing 10% of the sample), an income source of job-related wages/salary (4.3% more likely when not having used in the 30 days primary to discharge and when primary substance is not alcohol only, representing 28% of the sample), and white non-Hispanics (2.7% more likely when discharged from residential long-term treatment, representing 9% of the sample). Important implications are that: 1) those without a co-occurring mental health condition are the most likely to complete treatment, 2) those with job related wages or income are more likely to complete treatment, and 3) racial/ethnicity disparities persist in favor of white non-Hispanic individuals seeking to complete treatment. Thus, additional resources may be needed to combat such disparities. [ABSTRACT FROM AUTHOR]

Background: The increasing burden of hypertension in low- to middle-income countries necessitates the assessment of care coverage to monitor progress and guide future policies. This study uses an ensemble learning approach to evaluate hypertension care coverage in a nationally representative Iranian survey. Methods: The data source was the cross-sectional 2016 Iranian STEPwise approach to risk factor surveillance (STEPs). Hypertension was based on blood pressure ≥140/90 mmHg, reported use of anti-hypertensive medications, or a previous hypertension diagnosis. The four steps of care were screening (irrespective of blood pressure value), diagnosis, treatment, and control. The proportion of patients reaching each step was calculated, and a random forest model was used to identify features associated with progression to each step. After model optimization, the six most important variables at each step were considered to demonstrate population-based marginal effects. Results: The total number of participants was 30541 (52.3% female, median age: 42 years). Overall, 9420 (30.8%) had hypertension, among which 89.7% had screening, 62.3% received diagnosis, 49.3% were treated, and 7.9% achieved control. The random forest model indicated that younger age, male sex, lower wealth, and being unmarried/divorced were consistently associated with a lower probability of receiving care in different levels. Dyslipidemia was associated with reaching diagnosis and treatment steps; however, patients with other cardiovascular comorbidities were not likely to receive more intensive blood pressure management. Conclusion: Hypertension care was mostly missing the treatment and control stages. The random forest model identified features associated with receiving care, indicating opportunities to improve effective coverage. [ABSTRACT FROM AUTHOR]

The NICHD BPD Outcome Estimator uses clinical and demographic data to stratify respiratory outcomes of extremely preterm infants by risk. However, the Estimator does not have an option in its pull-down menu for infants of Asian descent. We hypothesize that respiratory outcomes in extreme prematurity among various racial/ethnic groups are interconnected and therefore the Estimator can still be used to predict outcomes in infants of Asian descent. Our goal was to apply a machine learning approach to assess whether outcome prediction for infants of Asian descent is possible with information hidden in the prediction results using White, Black, and Hispanic racial/ethnic groups as surrogates. We used the three racial/ethnic options in the Estimator to obtain the probabilities of BPD outcomes for each severity category. We then combined the probability results and developed three respiratory outcome prediction models at various postmenstrual age (PMA) by a random forest algorithm. We showed satisfactory model performance, with receiver operating characteristics area under the curve of 0.934, 0.850, and 0.757 for respiratory outcomes at PMA 36, 37, and 40 weeks, respectively, in the testing data set. This study suggested an interrelationship among racial/ethnic groups for respiratory outcomes among extremely preterm infants and showed the feasibility of extending the use of the Estimator to the Asian population. [ABSTRACT FROM AUTHOR]

Detrimental to individuals and society, online hateful messages have recently become a major social issue. Among them, one new type of hateful message, "hateful meme", has emerged and brought difficulties in traditional deep learning-based detection. Because hateful memes were formatted with both text captions and images to express users' intents, they cannot be accurately identified by singularly analyzing embedded text captions or images. In order to effectively detect a hateful meme, the algorithm must possess strong vision and language fusion capability. In this study, we move closer to this goal by feeding a triplet by stacking the visual features, object tags, and text features of memes generated by the object detection model named Visual features in Vision-Language (VinVl) and the optical character recognition (OCR) technology into a Transformer-based Vision-Language Pre-Training Model (VL-PTM) OSCAR+ to perform the cross-modal learning of memes. After fine-tuning and connecting to a random forest (RF) classifier, our model (OSCAR+RF) achieved an average accuracy and AUROC of 0.684 and 0.768, respectively, on the hateful meme detection task in a public test set, which was higher than the other eleven (11) published baselines. In conclusion, this study has demonstrated that VL-PTMs with the addition of anchor points can improve the performance of deep learning-based detection of hateful memes by involving a more substantial alignment between the text caption and visual information. [ABSTRACT FROM AUTHOR]

Detection of Premature Ventricular Contractions (PVC) is of crucial importance in the cardiology field, not only to improve the health system but also to reduce the workload of experts who analyze electrocardiograms (ECG) manually. PVC is a non-harmful common occurrence represented by extra heartbeats, whose diagnosis is not always easily identifiable, especially when done by long-term manual ECG analysis. In some cases, it may lead to disastrous consequences when associated with other pathologies. This work introduces an approach to identify PVCs using machine learning techniques without feature extraction and cross-validation techniques. In particular, a group of six classifiers has been used: Decision Tree, Random Forest, Long-Short Term Memory (LSTM), Bidirectional LSTM, ResNet-18, MobileNetv2, and ShuffleNet. Two types of experiments have been performed on data extracted from the MIT-BIH Arrhythmia database: (i) the original dataset and (ii) the balanced dataset. MobileNetv2 came in first in both experiments with high performance and promising results for PVCs' final diagnosis. The final results showed 99.90% of accuracy in the first experiment and 99.00% in the second one, despite no feature detection techniques were used. The approach we used, which was focused on classification without using feature extraction and cross-validation techniques, allowed us to provide excellent performance and obtain better results. Finally, this research defines as first step toward understanding the explanations for deep learning models' incorrect classifications. [ABSTRACT FROM AUTHOR]

Background: Adolescent Idiopathic Scoliosis (AIS) is a three-dimensional (3D) spinal deformity characterized by coronal curvature and rotational deformity. Predicting curve progression is important for the selection and timing of treatment. Although there is a consensus in the literature regarding prognostic factors associated with curve progression, the order of importance, as well as the combination of factors that are most predictive of curve progression is unknown. Objectives: (1) create an ordered list of prognostic factors that most contribute to curve progression, and (2) develop and validate a Machine Learning (ML) model to predict the final major Cobb angle in AIS patients. Methods: 193 AIS patients were selected for the current study. Preoperative PA, lateral and lateral bending radiographs were retrospectively obtained from the Shriners Hospitals for Children. Demographic and radiographic features, previously reported to be associated with curve progression, were collected. Sequential Backward Floating Selection (SBFS) was used to select a subset of the most predictive features. Based on the performance of several machine learning methods, a Random Forest (RF) regressor model was used to provide the importance rank of prognostic features and to predict the final major Cobb angle. Results: The seven most predictive prognostic features in the order of importance were initial major Cobb angle, flexibility, initial lumbar lordosis angle, initial thoracic kyphosis angle, age at last visit, number of levels involved, and Risser "+" stage at the first visit. The RF model predicted the final major Cobb angle with a Mean Absolute Error (MAE) of 4.64 degrees. Conclusion: A RF model was developed and validated to identify the most important prognostic features for curve progression and predict the final major Cobb angle. It is possible to predict the final major Cobb angle value within 5 degrees error from 2D radiographic features. Such methods could be directly applied to guide intervention timing and optimization for AIS treatment. [ABSTRACT FROM AUTHOR]

Collaborative modelling of the Internet of Things (IoT) with Artificial Intelligence (AI) has merged into the Intelligence of Things concept. This recent trend enables sensors to track required parameters and store accumulated data in cloud storage, which can be further utilized by AI based predictive models for automatic decision making. In a smart and sustainable environment, effective waste management is a concern. Poor regulation of waste in surrounding areas leads to rapid spread of contagious disease risks. Traditional waste object management requires more working staff, increases effort, consumes time and is relatively ineffective. In this research, an Intelligence of Things Enabled Smart Waste Management (IoT-SWM) model with predictive capabilities is developed. Here, local sinks (LS) are deployed in specified locations. At every instant, the current status of smart bins in each LS is notified to users to determine the priority level of LS to be emptied. Based on aggregated sensor values for the three smart bins, LS weight and poison gas value, the priority order of emptying LS is computed, and decision is made whether to notify the users with an alert message or not. It also helps in predicting the LS, which is likely to be filled up at a faster rate based on assigned timestamp. This model is implemented in real time with many LS and it was observed that bins, which were close to more crowded sites filled up faster compared to sparse populated areas. Random forest algorithm was used to predict whether an alert notification is to be sent or not. An average mean of 95.8% accuracy was noted while using 60 decision trees in random forest algorithm. The average mean execution latency recorded for training and testing sets is 13.06 sec and 14.39 sec respectively. Observed accuracy rate, precision, recall and f1-score parameters were 95.8%, 96.5%, 98.5% and 97.2% respectively. Model buildup and the validation time computed were 3.26 sec and 4.25 sec respectively. It is also noted that at a threshold value of 0.93 in LS level, the maximum accuracy rate reached was 95.8%. Thus, based on the prediction of random forest approach, a decision to notify the users is taken. Obtained outcome indicates that the waste level can be efficiently determined, and the overflow of dustbins can be easily checked in time [ABSTRACT FROM AUTHOR]

Annual monitoring of the spatial distribution of cultivated land is important for maintaining the ecological environment, achieving a status quo of land resource management, and guaranteeing agricultural production. With the gradual development of remote sensing technology, it has become a common practice to obtain cultivated land boundary information on a large scale with the help of satellite Earth observation images. Traditional land use classification methods are affected by multiple types of land cover, which leads to a decrease in the accuracy of cultivated land mapping. In contrast, although the current advanced methods (such as deep learning) can obtain more accurate cultivated land mapping results than traditional methods, such methods often require the use of a massive amount of training samples, large computing power, and highly complex model tuning processes, increasing the cost of mapping and requiring the involvement of more professionals. This has hindered the promotion of related methods in mapping institutions. This paper proposes a method based on time series vector features (MTVF), which uses vector thinking to establish the features. The advantage of this method is that the introduction of vector features enlarges the differences between the different land cover types, which overcomes the loss of mapping accuracy caused by the influences of the spectra of different ground objects and ensures the calculation efficiency. Moreover, the MTVF uses a traditional method (random forest) as the classification core, which makes the MTVF less demanding than advanced methods in terms of the number of training samples. Sentinel-2 satellite images were used to carry out cultivated land mapping for 2020 in northern Henan Province, China. The results show that the MTVF has the potential to accurately identify cultivated land. Furthermore, the overall accuracy, producer accuracy, and user accuracy of the overall study area and four sub-study areas were all greater than 90%. In addition, the cultivated land mapping accuracy of the MTVF is significantly better than that of the maximum likelihood, support vector machine, and artificial neural network methods. [ABSTRACT FROM AUTHOR]

With 36 players on the field, congestion in Australian football is an important consideration in identifying passing capacity, assessing fan enjoyment, and evaluating the effect of rule changes. However, no current method of objectively measuring congestion has been reported. This study developed two methods to measure congestion in Australian football. The first continuously determined the number of players situated within various regions of density at successive time intervals during a match using density-based clustering to group players as 'primary', 'secondary', or 'outside'. The second method aimed to classify the level of congestion a player experiences (high, nearby, or low) when disposing of the ball using the Random Forest algorithm. Both approaches were developed using data from the 2019 and 2021 Australian Football League (AFL) regular seasons, considering contextual variables, such as field position and quarter. Player tracking data and match event data from professional male players were collected from 56 matches performed at a single stadium. The random forest model correctly classified disposals in high congestion (0.89 precision, 0.86 recall, 0.96 AUC) and low congestion (0.98 precision, 0.86 recall, 0.96 AUC) at a higher rate compared to disposals nearby congestion (0.72 precision, 0.88 recall, 0.88 AUC). Overall, both approaches enable a more efficient method to quantify the characteristics of congestion more effectively, thereby eliminating manual input from human coders and allowing for a future comparison between additional contextual variables, such as, seasons, rounds, and teams. [ABSTRACT FROM AUTHOR]

Spatial patterns of stable isotopes in animal tissues or "isoscapes" can be used to investigate animal origins in a range of ecological and forensic investigations. Here, we developed a feather hydrogen isotope (δ2Hf) isoscape for Brazil based on 192 samples of feathers from the family Thraupidae from scientific collections. Raw values of δ2Hf ranged from -107.3 to +5.0‰, with higher values at the Caatinga biome (northeast Brazil) and lower values at the Amazon and Pantanal. A Random Forest (RF) method was used to model the spatial surface, using a range of environmental data as auxiliary variables. The RF model indicated a negative relationship between δ2Hf and Mean Annual Precipitation, Precipitation in the Warmest Quarter, and Annual Temperature Range and positive relationships for amount-weighted February-April precipitation δ2H (δ2Hp(Feb-April)) and Mean Annual Solar Radiation. Modelled δ2Hf values ranged from -85.7 to -13.6‰. Ours is the first δ2Hf isoscape for Brazil that can greatly assist our understanding of both ecological and biogeochemical processes controlling spatial variation in δ2H for this region. This isoscape can be used with caution, due to its poor predictive power (as found in other tropical regions) and can benefit from new sample input, new GNIP data, ecological and physiological studies, and keratin standard material better encompassing the range in feather samples from Brazil. So, we encourage new attempts to build more precise feather H isoscapes, as well as isoscapes based on other elements. [ABSTRACT FROM AUTHOR]

Fluctuations in a person's arousal accompany mental states such as drowsiness, mental effort, or motivation, and have a profound effect on task performance. Here, we investigated the link between two central instances affected by arousal levels, heart rate and eye movements. In contrast to heart rate, eye movements can be inferred remotely and unobtrusively, and there is evidence that oculomotor metrics (i.e., fixations and saccades) are indicators for aspects of arousal going hand in hand with changes in mental effort, motivation, or task type. Gaze data and heart rate of 14 participants during film viewing were used in Random Forest models, the results of which show that blink rate and duration, and the movement aspect of oculomotor metrics (i.e., velocities and amplitudes) link to heart rate–more so than the amount or duration of fixations and saccades. We discuss that eye movements are not only linked to heart rate, but they may both be similarly influenced by the common underlying arousal system. These findings provide new pathways for the remote measurement of arousal, and its link to psychophysiological features. [ABSTRACT FROM AUTHOR]

Background: Approaching epidemiological data with flexible machine learning algorithms is of great value for understanding disease-specific association patterns. However, it can be difficult to correctly extract and understand those patterns due to the lack of model interpretability. Method: We here propose a machine learning workflow that combines random forests with Bayesian network surrogate models to allow for a deeper level of interpretation of complex association patterns. We first evaluate the proposed workflow on synthetic data. We then apply it to data from the large population-based Study of Health in Pomerania (SHIP). Based on this combination, we discover and interpret broad patterns of individual serum TSH concentrations, an important marker of thyroid functionality. Results: Evaluations using simulated data show that feature associations can be correctly recovered by combining random forests and Bayesian networks. The presented model achieves predictive accuracy that is similar to state-of-the-art models (root mean square error of 0.66, mean absolute error of 0.55, coefficient of determination of R2 = 0.15). We identify 62 relevant features from the final random forest model, ranging from general health variables over dietary and genetic factors to physiological, hematological and hemostasis parameters. The Bayesian network model is used to put these features into context and make the black-box random forest model more understandable. Conclusion: We demonstrate that the combination of random forest and Bayesian network analysis is helpful to reveal and interpret broad association patterns of individual TSH concentrations. The discovered patterns are in line with state-of-the-art literature. They may be useful for future thyroid research and improved dosing of therapeutics. [ABSTRACT FROM AUTHOR]

Mangrove forests are the most important ecosystems on Pohnpei Island, Federated States of Micronesia, as the island communities of the central Pacific rely on the forests for many essential services including protection from sea-level rise that is occurring at a greater pace than the global average. As part of a multi-component assessment to evaluate vulnerabilities of mangrove forests on Pohnpei, mangrove forests were mapped at two points in time: 1983 and 2018. In 2018, the island had 6,426 ha of mangrove forest. Change analysis indicated a slight (0.76%) increase of mangrove area between 1983 and 2018, contrasting with global mangrove area declines. Forest structure and aboveground carbon (AGC) stocks were inventoried using a systematic sampling of field survey plots and extrapolated to the island using k-nearest neighbor and random forest species models. A gridded or wall to wall approach is suggested when possible for defining carbon stocks of a large area due to high variability seen in our data. The k-nearest neighbor model performed better than random forest models to map species dominance in these forests. Mean AGC was 167 ± 11 MgC ha-1, which is greater than the global average of mangroves (115 ± 7 MgC ha-1) but within their global range (37–255 MgC ha-1) Kauffman et al. (2020). In 2018, Pohnpei mangroves contained over 1.07 million MgC in AGC pools. By assigning the mean AGC stock per species per area to the map, carbon stock distributions were visualized spatially, allowing future conservation efforts to be directed to carbon dense stands. [ABSTRACT FROM AUTHOR]

Fasciola hepatica and Ostertagia ostertagi are internal parasites of cattle compromising physiology, productivity, and well-being. Parasites are complex in their effect on hosts, sometimes making it difficult to identify clear directions of associations between infection and production parameters. Therefore, unsupervised approaches not assuming a structure reduce the risk of introducing bias to the analysis. They may provide insights which cannot be obtained with conventional, supervised methodology. An unsupervised, exploratory cluster analysis approach using the k–mode algorithm and partitioning around medoids detected two distinct clusters in a cross-sectional data set of milk yield, milk fat content, milk protein content as well as F. hepatica or O. ostertagi bulk tank milk antibody status from 606 dairy farms in three structurally different dairying regions in Germany. Parasite–positive farms grouped together with their respective production parameters to form separate clusters. A random forests algorithm characterised clusters with regard to external variables. Across all study regions, co–infections with F. hepatica or O. ostertagi, respectively, farming type, and pasture access appeared to be the most important factors discriminating clusters (i.e. farms). Furthermore, farm level lameness prevalence, herd size, BCS, stage of lactation, and somatic cell count were relevant criteria distinguishing clusters. This study is among the first to apply a cluster analysis approach in this context and potentially the first to implement a k–medoids algorithm and partitioning around medoids in the veterinary field. The results demonstrated that biologically relevant patterns of parasite status and milk parameters exist between farms positive for F. hepatica or O. ostertagi, respectively, and negative farms. Moreover, the machine learning approach confirmed results of previous work and shed further light on the complex setting of associations a between parasitic diseases, milk yield and milk constituents, and management practices. [ABSTRACT FROM AUTHOR]

Viral-host protein protein interaction (PPI) analysis is essential to decode the molecular mechanism of viral pathogen and host immunity processes which eventually help to control viral diseases and optimize therapeutics. The state-of-the-art viral-host PPI predictor leverages unsupervised embedding learning technique (doc2vec) to generate statistical representations of viral-host protein sequences and a Random Forest classifier for interaction prediction. However, doc2vec approach generates the statistical representations of viral-host protein sequences by merely modelling the local context of residues which only partially captures residue semantics. The paper in hand proposes a novel technique for generating better statistical representations of viral and host protein sequences based on the infusion of comprehensive local and global contextual information of the residues. While local residue context aware encoding captures semantic relatedness and short range dependencies of residues. Global residue context aware encoding captures comprehensive long-range residues dependencies, positional invariance of residues, and unique residue combination distribution important for interaction prediction. Using concatenated rich statistical representations of viral and host protein sequences, a robust machine learning framework "LGCA-VHPPI" is developed which makes use of a deep forest model to effectively model complex non-linearity of viral-host PPI sequences. An in-depth performance comparison of the proposed LGCA-VHPPI framework with existing diverse sequence encoding schemes based viral-host PPI predictors reveals that LGCA-VHPPI outperforms state-of-the-art predictor by 6%, 2%, and 2% in terms of matthews correlation coefficient over 3 different benchmark viral-host PPI prediction datasets. [ABSTRACT FROM AUTHOR]

Terrorism is a major problem worldwide, causing thousands of fatalities and billions of dollars in damage every year. To address this threat, we propose a novel feature representation method and evaluate machine learning models that learn from localized news data in order to predict whether a terrorist attack will occur on a given calendar date and in a given state. The best model (a Random Forest aided by a novel variable-length moving average method) achieved area under the receiver operating characteristic (AUROC) of ≥ 0.667 (statistically significant w.r.t. random guessing with p ≤.0001) on four of the five states that were impacted most by terrorism between 2015 and 2018. These results demonstrate that treating terrorism as a set of independent events, rather than as a continuous process, is a fruitful approach—especially when historical events are sparse and dissimilar—and that large-scale news data contains information that is useful for terrorism prediction. Our analysis also suggests that predictive models should be localized (i.e., state models should be independently designed, trained, and evaluated) and that the characteristics of individual attacks (e.g., responsible group or weapon type) were not correlated with prediction success. These contributions provide a foundation for the use of machine learning in efforts against terrorism in the United States and beyond. [ABSTRACT FROM AUTHOR]

There remains a limited understanding of the HIV prevention and treatment needs among female sex workers in many parts of the world. Systematic reviews of existing literature can help fill this gap; however, well-done systematic reviews are time-demanding and labor-intensive. Here, we propose an automatic document classification approach to a systematic review to significantly reduce the effort in reviewing documents and optimizing empiric decision making. We first describe a manual document classification procedure that is used to curate a pertinent training dataset and then propose three classifiers: a keyword-guided method, a cluster analysis-based method, and a random forest approach that utilizes a large set of feature tokens. This approach is used to identify documents studying female sex workers that contain content relevant to either HIV or experienced violence. We compare the performance of the three classifiers by cross-validation in terms of area under the curve of the receiver operating characteristic and precision and recall plot, and found random forest approach reduces the amount of manual reading for our example by 80%; in sensitivity analysis, we found that even trained with only 10% of data, the classifier can still avoid reading 75% of future documents (68% of total) while retaining 80% of relevant documents. In sum, the automated procedure of document classification presented here could improve both the precision and efficiency of systematic reviews and facilitate live reviews, where reviews are updated regularly. We expect to obtain a reasonable classifier by taking 20% of retrieved documents as training samples. The proposed classifier could also be used for more meaningfully assembling literature in other research areas and for rapid documents screening with a tight schedule, such as COVID-related work during the crisis. [ABSTRACT FROM AUTHOR]

Background and objectives: Sleep disorders related to Parkinson's disease (PD) have recently attracted increasing attention, but there are few clinical reports on the correlation of Parkinson's disease patients with rapid eye movement (REM) sleep behavior disorder (RBD). Therefore, this study conducted a cognitive function examination for Parkinson's disease patients and discussed the application effect of three algorithms in the screening of influencing factors and risk prediction effects. Methods: Three algorithms (logistic regression, machine learning-based regression trees and random forest) were used to establish a prediction model for PD-RBD patients, and the application effects of the three algorithms in the screening of influencing factors and the risk prediction of PD-RBD were discussed. Results: The subjects included 169 patients with Parkinson's disease (Parkinson's disease with RBD [PD-RBD] = 69 subjects; Parkinson's disease without RBD [PD-nRBD] = 100 subjects). This study compared the predictive performance of RF, decision tree and logistic regression, selected a final model with the best model performance and proposed the importance of variables in the final model. After the analysis, the accuracy of RF (83.05%) was better than that of the other models (decision tree = 75.10%, logistic regression = 71.62%). PQSI, Scopa-AUT score, MoCA score, MMSE score, AGE, LEDD, PD-course, UPDRS total score, ESS score, NMSQ, disease type, RLSRS, HAMD, UPDRS III and PDOnsetage are the main variables for predicting RBD, along with increased weight. Among them, PQSI is the most important factor. The prediction model of Parkinson's disease RBD that was established in this study will help in screening out predictive factors and in providing a reference for the prognosis and preventive treatment of PD-RBD patients. Conclusions: The random forest model had good performance in the prediction and evaluation of PD-RBD influencing factors and was superior to decision tree and traditional logistic regression models in many aspects, which can provide a reference for the prognosis and preventive treatment of PD-RBD patients. [ABSTRACT FROM AUTHOR]

We use a rich set of transaction data from a large retailer in India and a dataset on bribe payments to train random forest and XGBoost models using empirical measures guided by Benford's Law, a commonly used tool in forensic analytics. We evaluate the performance around the 2016 Indian Demonetization, which affects the distribution of legal tender notes in India, and find that models using only pre-2016 data or post-2016 data for both training and testing data had F1 score ranges around 90%, suggesting that these models and Benford's law criteria contain meaningful information for detecting bribe payments. However, the performance for models trained in one regime and tested in another falls dramatically to less than 10%, highlighting the role of the institutional setting when using financial data analytics in an environment subject to regime shifts. [ABSTRACT FROM AUTHOR]

Introduction: Assisted reproductive technology has been proposed for women with infertility. Moreover, in vitro fertilization (IVF) cycles are increasing. Factors contributing to successful pregnancy have been widely explored. In this study, we used machine learning algorithms to construct prediction models for clinical pregnancies in IVF. Materials and methods: A total of 24,730 patients entered IVF and intracytoplasmic sperm injection cycles with clinical pregnancy outcomes at Taipei Medical University Hospital. Data used included patient characteristics and treatment. We used machine learning methods to develop prediction models for clinical pregnancy and explored how each variable affects the outcome of interest using partial dependence plots. Results: Experimental results showed that the random forest algorithm outperforms logistic regression in terms of areas under the receiver operating characteristics curve. The ovarian stimulation protocol is the most important factor affecting pregnancy outcomes. Long and ultra-long protocols have shown positive effects on clinical pregnancy among all protocols. Furthermore, total frozen and transferred embryos are positive for a clinical pregnancy, but female age and duration of infertility have negative effects on clinical pregnancy. Conclusion: Our findings show the importance of variables and propensity of each variable by random forest algorithm for clinical pregnancy in the assisted reproductive technology cycle. This study provides a ranking of variables affecting clinical pregnancy and explores the effects of each treatment on successful pregnancy. Our study has the potential to help clinicians evaluate the success of IVF in patients. [ABSTRACT FROM AUTHOR]

Evaluation of surgical skills during minimally invasive surgeries is needed when recruiting new surgeons. Although surgeons' differentiation by skill level is highly complex, performance in specific clinical tasks such as pegboard transfer and knot tying could be determined using wearable EMG and accelerometer sensors. A wireless wearable platform has made it feasible to collect movement and muscle activation signals for quick skill evaluation during surgical tasks. However, it is challenging since the placement of multiple wireless wearable sensors may interfere with their performance in the assessment. This study utilizes machine learning techniques to identify optimal muscles and features critical for accurate skill evaluation. This study enrolled a total of twenty-six surgeons of different skill levels: novice (n = 11), intermediaries (n = 12), and experts (n = 3). Twelve wireless wearable sensors consisting of surface EMGs and accelerometers were placed bilaterally on bicep brachii, tricep brachii, anterior deltoid, flexor carpi ulnaris (FCU), extensor carpi ulnaris (ECU), and thenar eminence (TE) muscles to assess muscle activations and movement variability profiles. We found features related to movement complexity such as approximate entropy, sample entropy, and multiscale entropy played a critical role in skill level identification. We found that skill level was classified with highest accuracy by i) ECU for Random Forest Classifier (RFC), ii) deltoid for Support Vector Machines (SVM) and iii) biceps for Naïve Bayes Classifier with classification accuracies 61%, 57% and 47%. We found RFC classifier performed best with highest classification accuracy when muscles are combined i) ECU and deltoid (58%), ii) ECU and biceps (53%), and iii) ECU, biceps and deltoid (52%). Our findings suggest that quick surgical skill evaluation is possible using wearables sensors, and features from ECU, deltoid, and biceps muscles contribute an important role in surgical skill evaluation. [ABSTRACT FROM AUTHOR]

As the rate of percutaneous coronary intervention increases, in-stent restenosis (ISR) has become a burden. Random forest (RF) could be superior to logistic regression (LR) for predicting ISR due to its robustness. We developed an RF model and compared its performance with the LR one for predicting ISR. We retrospectively included 1501 patients (age: 64.0 ± 10.3; male: 76.7%; ISR events: 279) who underwent coronary angiography at 9 to 18 months after implantation of 2nd generation drug-eluting stents. The data were randomly split into a pair of train and test datasets for model development and validation with 50 repeats. The predictive performance was assessed by the area under the curve (AUC) of the receiver operating characteristic (ROC). The RF models predicted ISR with larger AUC-ROCs of 0.829 ± 0.025 compared to 0.784 ± 0.027 of the LR models. The difference was statistically significant in 29 of the 50 repeats. The RF and LR models had similar sensitivity using the same cutoff threshold, but the specificity was significantly higher in the RF models, reducing 25% of the false positives. By removing the high leverage outliers, the LR models had comparable AUC-ROC to the RF models. Compared to the LR, the RF was more robust and significantly improved the performance for predicting ISR. It could cost-effectively identify patients with high ISR risk and help the clinical decision of coronary stenting. [ABSTRACT FROM AUTHOR]

Brazil presents large yield gaps in garlic crops partly due to nutrient mismanagement at local scale. Machine learning (ML) provides powerful tools to handle numerous combinations of yield-impacting factors that help reducing the number of assumptions about nutrient management. The aim of the current study is to customize fertilizer recommendations to reach high garlic marketable yield at local scale in a pilot study. Thus, collected 15 nitrogen (N), 24 phosphorus (P), and 27 potassium (K) field experiments conducted during the 2015 to 2017 period in Santa Catarina state, Brazil. In addition, 61 growers' observational data were collected in the same region in 2018 and 2019. The data set was split into 979 experimental and observational data for model calibration and into 45 experimental data (2016) to test ML models and compare the results to state recommendations. Random Forest (RF) was the most accurate ML to predict marketable yield after cropping system (cultivar, preceding crops), climatic indices, soil test and fertilization were included features as predictor (R2 = 0.886). Random Forest remained the most accurate ML model (R2 = 0.882) after excluding cultivar and climatic features from the prediction-making process. The model suggested the application of 200 kg N ha-1 to reach maximum marketable yield in a test site in comparison to the 300 kg N ha-1 set as state recommendation. P and K fertilization also seemed to be excessive, and it highlights the great potential to reduce production costs and environmental footprint without agronomic loss. Garlic root colonization by arbuscular mycorrhizal fungi likely contributed to P and K uptake. Well-documented data sets and machine learning models could support technology transfer, reduce costs with fertilizers and yield gaps, and sustain the Brazilian garlic production. [ABSTRACT FROM AUTHOR]

The Alzheimer′s Disease Prediction Of Longitudinal Evolution (TADPOLE) Challenge is the most comprehensive challenge to date with regard to the number of subjects, considered features, and challenge participants. The initial objective of TADPOLE was the identification of the most predictive data, features, and methods for the progression of subjects at risk of developing Alzheimer′s. The challenge was successful in recognizing tree-based ensemble methods such as gradient boosting and random forest as the best methods for the prognosis of the clinical status in Alzheimer's disease (AD). However, the challenge outcome was limited to which combination of data processing and methods exhibits the best accuracy; hence, it is difficult to determine the contribution of the methods to the accuracy. The quantification of feature importance was globally approached by all the challenge participant methods. In addition, TADPOLE provided general answers that focused on improving performance while ignoring important issues such as interpretability. The purpose of this study is to intensively explore the models of the top three TADPOLE Challenge methods in a common framework for fair comparison. In addition, for these models, the most meaningful features for the prognosis of the clinical status of AD are studied and the contribution of each feature to the accuracy of the methods is quantified. We provide plausible explanations as to why the methods achieve such accuracy, and we investigate whether the methods use information coherent with clinical knowledge. Finally, we approach these issues through the analysis of SHapley Additive exPlanations (SHAP) values, a technique that has recently attracted increasing attention in the field of explainable artificial intelligence (XAI). [ABSTRACT FROM AUTHOR]

Background: Revision endoscopic sinus surgery (ESS) is often considered for chronic rhinosinusitis (CRS) if maximal conservative treatment and baseline ESS prove insufficient. Emerging research outlines the risk factors of revision ESS. However, accurately predicting revision ESS at the individual level remains uncertain. This study aims to examine the prediction accuracy of revision ESS and to identify the effects of risk factors at the individual level. Methods: We collected demographic and clinical variables from the electronic health records of 767 surgical CRS patients ≥16 years of age. Revision ESS was performed on 111 (14.5%) patients. The prediction accuracy of revision ESS was examined by training and validating different machine learning models, while the effects of variables were analysed using the Shapley values and partial dependence plots. Results: The logistic regression, gradient boosting and random forest classifiers performed similarly in predicting revision ESS. Area under the receiving operating characteristic curve (AUROC) values were 0.744, 0.741 and 0.730, respectively, using data collected from the baseline visit until six months after baseline ESS. The length of time during which data were collected improved the prediction performance. For data collection times of 0, 3, 6 and 12 months after baseline ESS, AUROC values for the logistic regression were 0.682, 0.715, 0.744 and 0.784, respectively. The number of visits before or after baseline ESS, the number of days from the baseline visit to the baseline ESS, patient age, CRS with nasal polyps (CRSwNP), asthma, non-steroidal anti-inflammatory drug exacerbated respiratory disease and immunodeficiency or suspicion of it all associated with revision ESS. Patient age and number of visits before baseline ESS carried non-linear effects for predictions. Conclusions: Intelligent data analysis identified important predictors of revision ESS at the individual level, such as the frequency of clinical visits, patient age, Type 2 high diseases and immunodeficiency or a suspicion of it. [ABSTRACT FROM AUTHOR]

The paper investigates potentials and challenges during the interpretation of prehistoric settlement dynamics based on large archaeological datasets. Exemplarily, this is carried out using a database of 1365 Neolithic sites in the Weiße Elster river catchment in Central Germany located between the southernmost part of the Northern German Plain and the Central Uplands. The recorded sites are systematically pre-processed with regard to their chronology, functional interpretation and spatial delineation. The quality of the dataset is reviewed by analyzing site distributions with respect to field surveys and modern land use. The Random Forests machine learning algorithm is used to examine the impact of terrain covariates on the depth of sites and pottery preservation. Neolithic settlement dynamics are studied using Site Exploitation Territories, and site frequencies per century are used to compare the intensity of land use with adjacent landscapes. The results show that the main trends of the Neolithic settlement dynamics can be derived from the dataset. However, Random Forests analyses indicate poor pottery preservation in the Central Uplands and a superimposition of Neolithic sites in the southernmost part of the Northern German Plain. Throughout the Neolithic the margins between soils on loess and the Weiße Elster floodplain were continuously settled, whereas only Early and Late Neolithic land use also extended into the Central Uplands. These settlement patterns are reflected in the results of the Site Exploitation Territories analyses and explained with environmental economic factors. Similar with adjacent landscapes the Middle Neolithic site frequency is lower compared to earlier and later periods. [ABSTRACT FROM AUTHOR]

Background: The ability to accurately distinguish bacterial from viral infection would help clinicians better target antimicrobial therapy during suspected lower respiratory tract infections (LRTI). Although technological developments make it feasible to rapidly generate patient-specific microbiota profiles, evidence is required to show the clinical value of using microbiota data for infection diagnosis. In this study, we investigated whether adding nasal cavity microbiota profiles to readily available clinical information could improve machine learning classifiers to distinguish bacterial from viral infection in patients with LRTI. Results: Various multi-parametric Random Forests classifiers were evaluated on the clinical and microbiota data of 293 LRTI patients for their prediction accuracies to differentiate bacterial from viral infection. The most predictive variable was C-reactive protein (CRP). We observed a marginal prediction improvement when 7 most prevalent nasal microbiota genera were added to the CRP model. In contrast, adding three clinical variables, absolute neutrophil count, consolidation on X-ray, and age group to the CRP model significantly improved the prediction. The best model correctly predicted 85% of the 'bacterial' patients and 82% of the 'viral' patients using 13 clinical and 3 nasal cavity microbiota genera (Staphylococcus, Moraxella, and Streptococcus). Conclusions: We developed high-accuracy multi-parametric machine learning classifiers to differentiate bacterial from viral infections in LRTI patients of various ages. We demonstrated the predictive value of four easy-to-collect clinical variables which facilitate personalized and accurate clinical decision-making. We observed that nasal cavity microbiota correlate with the clinical variables and thus may not add significant value to diagnostic algorithms that aim to differentiate bacterial from viral infections. [ABSTRACT FROM AUTHOR]

Introduction: Machine learning (ML) accelerometer data processing methods have potential to improve the accuracy of device-based assessments of physical activity (PA) in young children. Yet the uptake of ML methods by health researchers has been minimal and the use of cut-points (CP) continues to be the norm, despite evidence of significant misclassification error. The lack of studies demonstrating a relative advantage for ML approaches over CP methods maybe a key contributing factor. Purpose: The current study compared the accuracy of PA intensity predictions provided by ML classification models and previously published CPs for preschool-aged children. Methods: In a free-living study, 31 preschool-aged children (mean age = 4.0 ± 0.9 y) wore wrist and hip ActiGraph GT3X+ accelerometers while completing a video recorded 20-minute free play session. Ground truth PA intensity was coded continuously using the Children's Activity Rating Scale (CARS). Accelerometer data was classified as sedentary (SED), light intensity (LPA), or moderate-to-vigorous intensity (MVPA) using ML random forest PA classifiers and published CPs for preschool-aged children. Performance differences were evaluated in a hold-out sample by comparing weighted kappa statistics, classification accuracy for each intensity band, and equivalence testing. Results: ML classification models (hip: κ = 0.76; wrist: κ = 0.72) exhibited significantly higher agreement with ground truth PA intensity than CP methods (hip: κ = 0.38–0.49; wrist: κ = 0.31–0.44). For the ML models, classification accuracy for SED and LPA ranged from 83% - 88%, while classification accuracy for MVPA ranged from 68% - 78%. For the CP's, classification accuracy ranged from 50% - 94% for SED, 19% - 75% for LPA, and 44% - 76.1% for MVPA. ML classification models showed equivalence (within ± 0.5 SD) with directly observed time in SED, LPA, and MVPA. None of the CP's exhibited evidence of equivalence. Conclusions: Under free living conditions, ML classification models for hip or wrist accelerometer data provide more accurate assessments of PA intensity in young children than CP methods. The results demonstrate the relative advantage of ML methods over threshold-based approaches and adds to a growing evidence base supporting the feasibility and accuracy of ML accelerometer data processing methods. [ABSTRACT FROM AUTHOR]

Activity recognition modelling using smartphone Inertial Measurement Units (IMUs) is an underutilized resource defining and assessing work efficiency for a wide range of natural resource management tasks. This study focused on the initial development and validation of a smartphone-based activity recognition system for excavator-based mastication equipment working in Ponderosa pine (Pinus ponderosa) plantations in North Idaho, USA. During mastication treatments, sensor data from smartphone gyroscopes, accelerometers, and sound pressure meters (decibel meters) were collected at three sampling frequencies (10, 20, and 50 hertz (Hz)). These data were then separated into 9 time domain features using 4 sliding window widths (1, 5, 7.5 and 10 seconds) and two levels of window overlap (50% and 90%). Random forest machine learning algorithms were trained and evaluated for 40 combinations of model parameters to determine the best combination of parameters. 5 work elements (masticate, clear, move, travel, and delay) were classified with the performance metrics for individual elements of the best model (50 Hz, 10 second window, 90% window overlap) falling within the following ranges: area under the curve (AUC) (95.0% - 99.9%); sensitivity (74.9% - 95.6%); specificity (90.8% - 99.9%); precision (81.1% - 98.3%); F1-score (81.9% - 96.9%); balanced accuracy (87.4% - 97.7%). Smartphone sensors effectively characterized individual work elements of mechanical fuel treatments. This study is the first example of developing a smartphone-based activity recognition model for ground-based forest equipment. The continued development and dissemination of smartphone-based activity recognition models may assist land managers and operators with ubiquitous, manufacturer-independent systems for continuous and automated time study and production analysis for mechanized forest operations. [ABSTRACT FROM AUTHOR]

Mapping population distribution at a fine spatial scale is essential for urban studies and planning. Numerous studies, mainly supported by geospatial and statistical methods, have focused primarily on predicting population counts. However, estimating their socio-economic characteristics beyond population counts, such as average age, income, and gender ratio, remains unattended. We enhance traditional population estimation by predicting not only the number of residents in an area, but also their demographic characteristics: average age and the proportion of seniors. By implementing and comparing different machine learning techniques (Random Forest, Support Vector Machines, and Linear Regression) in administrative areas in Singapore, we investigate the use of point of interest (POI) and real estate data for this purpose. The developed regression model predicts the average age of residents in a neighbourhood with a mean error of about 1.5 years (the range of average resident age across Singaporean districts spans approx. 14 years). The results reveal that age patterns of residents can be predicted using real estate information rather than with amenities, which is in contrast to estimating population counts. Another contribution of our work in population estimation is the use of previously unexploited POI and real estate datasets for it, such as property transactions, year of construction, and flat types (number of rooms). Advancing the domain of population estimation, this study reveals the prospects of a small set of detailed and strong predictors that might have the potential of estimating other demographic characteristics such as income. [ABSTRACT FROM AUTHOR]

Although Africa is the most vulnerable region to climate change, little research has focused on how climate change is perceived by Africans. Using random forest methodology, we analyze survey and climate data from second-order political boundaries to explore what predicts climate change beliefs in Africa. We include five different dimensions of climate change beliefs: climate change awareness, belief in anthropogenic climate change, risk perception, the need to stop climate change, and self-efficacy. Based on these criteria we identify five key results: (1) climate change in Africa is largely perceived through its negative impacts on agriculture; (2) actual changes in local climate conditions are related to climate change beliefs; (3) authoritarian and intolerant ideologies are associated to less climate change awareness, and a diminished risk perception and belief that it must be stopped; (4) women are less likely to be aware of climate change, and (5) not speaking French, English or Portuguese is linked to a hindered understanding of climate beliefs. Our combined results can help policy makers better understand the need to jointly consider the multilevel complexities of individual beliefs and hydroclimatic data for the development of more accurate adaptation and mitigation strategies to combat the impacts of climate change in Africa. [ABSTRACT FROM AUTHOR]

The present research reports on the use of data mining techniques for differentiating between translated and non-translated original Chinese based on monolingual comparable corpora. We operationalized seven entropy-based metrics including character, wordform unigram, wordform bigram and wordform trigram, POS (Part-of-speech) unigram, POS bigram and POS trigram entropy from two balanced Chinese comparable corpora (translated vs non-translated) for data mining and analysis. We then applied four data mining techniques including Support Vector Machines (SVMs), Linear discriminant analysis (LDA), Random Forest (RF) and Multilayer Perceptron (MLP) to distinguish translated Chinese from original Chinese based on these seven features. Our results show that SVMs is the most robust and effective classifier, yielding an AUC of 90.5% and an accuracy rate of 84.3%. Our results have affirmed the hypothesis that translational language is categorically different from original language. Our research demonstrates that combining information-theoretic indicator of Shannon's entropy together with machine learning techniques can provide a novel approach for studying translation as a unique communicative activity. This study has yielded new insights for corpus-based studies on the translationese phenomenon in the field of translation studies. [ABSTRACT FROM AUTHOR]

Accurate determination of the axial load capacity of the pile is of utmost importance when designing the pile foundation. However, the methods of determining the axial load capacity of the pile in the field are often costly and time-consuming. Therefore, the purpose of this study is to develop a hybrid machine-learning to predict the axial load capacity of the pile. In particular, two powerful optimization algorithms named Herd Optimization (PSO) and Genetic Algorithm (GA) were used to evolve the Random Forest (RF) model architecture. For the research, the data set including 472 results of pile load tests in Ha Nam province—Vietnam was used to build and test the machine-learning models. The data set was divided into training and testing parts with ratio of 80% and 20%, respectively. Various performance indicators, namely absolute mean error (MAE), mean square root error (RMSE), and coefficient of determination (R2) are used to evaluate the performance of RF models. The results showed that, between the two optimization algorithms, GA gave superior performance compared to PSO in finding the best RF model architecture. In addition, the RF-GA model is also compared with the default RF model, the results show that the RF-GA model gives the best performance, with the balance on training and testing set, meaning avoiding the phenomenon of overfitting. The results of the study suggest a potential direction in the development of machine learning models in engineering in general and geotechnical engineering in particular. [ABSTRACT FROM AUTHOR]

Autoimmune hepatitis (AIH) is a poorly understood, chronic disease, for which corticosteroids are still the mainstay of therapy and most patients undergo liver biopsy to obtain a diagnosis. We aimed to determine if there was a transcriptomic signature of AIH in the peripheral blood and investigate underlying biologic pathways revealed by gene expression analysis. Whole blood RNA from 75 AIH patients and 25 healthy volunteers was extracted and sequenced. Differential gene expression analysis revealed 249 genes that were significantly differentially expressed in AIH patients compared to controls. Using a random forest algorithm, we determined that less than 10 genes were sufficient to differentiate the two groups in our cohort. Interferon signaling was more active in AIH samples compared to controls, regardless of treatment status. Pegivirus sequences were detected in five AIH samples and 1 healthy sample. The gene expression data and clinical metadata were used to determine 12 genes that were significantly associated with advanced fibrosis in AIH. AIH patients with a partial response to therapy demonstrated decreased evidence of a CD8+ T cell gene expression signal. These findings represent progress in understanding a disease in need of better tests, therapies, and biomarkers. [ABSTRACT FROM AUTHOR]