Balanced Random Forest Research Articles

CHARACTERIZATION OF MORTALITY PREDICTION: AN ENSEMBLE LEARNING ANALYSIS USING THE MIMIC-III DATASET

Applications that employ medical data are directly impacted by the classification of imbalanced data. It is vital due to the nature of classification and solutions about medical data. The purpose of this article is to identify a machine learning model that may be successfully applied in the medical field to reduce the number of mortality and optimize the efficiency of hospital resources. For this reason, it is thought that the better the performance of the ML model, the more a different perspective will be gained on the problems in today's medicine. Therefore, in the study, Weighted Random Forest (WRF) and Balanced Random Forest (BRF) which are ensemble machine learning (ML) methods for imbalanced data were implemented to identify the performance of the algorithms for mortality determination from open-source MIMIC-III dataset by using vital signs, comorbidities, and laboratory variables with demographic characteristic information. To evaluate the performance of WRF and BRF, a Random Forest Classifier (RFC) was also implemented to investigate the power of developed models for imbalanced data. In addition, the features used in the ML methods were separated into three groups to explore the impact of the vital signs, comorbidities, and laboratory variables with demographic characteristics separately on mortality identification. In addition to previous applications on UCI datasets, the present study revealed that the BRF method for imbalanced medical data provides high performance in determining the majority and minority classes of the data by using vital signs and laboratory variables with demographic characteristics.

A Radiogenomic Machine Learning Model for Glioblastoma Post-Resection Overall Survival Group Prediction

Novel methods are needed to better predict outcomes and tailor personalized therapies for patients with glioblastoma (GBM). To improve post-resection survival prediction in GBM patients, we developed a machine learning model that utilized 1) radiomic computational biomarkers from pre-resection multi-parametric MRI (mp-MRI) images, and 2) clinical information including tumor molecular features and extent of surgical resection. A cohort of 406 GBM patients treated with surgical resection was studied. Each patient received a pre-resection mp-MRI that included T1, contrast-enhanced T1 (T1-ce), T2, and FLAIR sequences. Three tumor subregions, i.e., enhanced tumor, tumor core, and whole tumor, were segmented with radiologists' correction. Based on survival outcomes in the literature, the cohort was categorized into three survival groups: group A (<9 mos, n = 148), B (9-21 mos, n = 181), and C (21+ mos, n = 77). We first extracted 88 radiomic features from each tumor subregion on each MR volume, and Z-score normalization was adopted. Three other patient-specific factors, including age, resection status (subtotal versus gross total), and IDH1 status, were concatenated with the radiomic features as a synthesized patient-specific feature vector. We then designed a two-step machine learning model: using the patient-specific feature vectors, the model 1) identified patients in group A with the shortest predicted survival using a balanced random forest (BRF) classifier, and 2) used a 2nd BRF classifier to segregate the remaining patients into groups B and C. For model training, a 7:3 training/test sample ratio was adopted, and 100 model versions were acquired through random validation sample assignments to study model robustness. Sensitivity, specificity, and accuracy results of each group were calculated, and an overall ROC was generated to represent the model's overall performance. The model demonstrated acceptable prediction performance with an ROC AUC value at 0.68. Individually, the model achieved good prediction accuracies for short and long-term survival prediction (Groups A and C), though we observed relatively limited accuracy in medium survival prediction (Group B). Model sensitivity in group A was promising, but was limited in the remaining two groups. Feature weight analysis showed that radiomic features from the enhanced tumor subregion were the leading variables in the BRF classifier. The developed radiogenomic machine learning model predicts GBM post-resection overall survival outcomes. Future work is necessary to further improve model sensitivity for patients with medium and long-term predicted survival.

Two-stage Hierarchical Framework for Solar Flare Prediction

Solar flares, often accompanied by coronal mass ejections and other solar phenomena, are one of the most important sources affecting space weather. It is important to investigate the forecast approach of solar flares to mitigate their destructive effect on the Earth. Statistical analysis, associated with data from 2010 to 2017 in Space-weather HMI Active Region Patches (SHARPs) collected by the Solar Dynamics Observatory's Helioseismic and Magnetic Imager, reveals that there is a distribution divergence between the two types of active regions (ARs) of solar flares. A two-stage hierarchical prediction framework is formulated to better utilize this intrinsic distribution information. Specially, we pick up the ARs where at least one solar flare event occurs within the next 48 hr as flaring ARs through balanced random forest and naive Bayesian methods and then predict the events from flaring ARs by a cascade module of learning models. The empirical evaluation of SHARPs data from 2016 to 2019 verifies the promising performance of our framework, e.g., 0.727 for the true skill statistic.

Classification of T lymphocyte motility behaviors using a machine learning approach.

T lymphocytes migrate into organs and interact with local cells to perform their functions. How human T lymphocytes communicate with organ-specific cells and participate in pathobiological processes remains unresolved. Brain infiltration of T lymphocytes is associated with multiple neurological disorders. Thus, to characterize the behavior of human T lymphocytes reaching the human brain, we performed time-lapse microscopy on human CD8+ T lymphocytes co-cultured with either primary human astrocytes or neurons. Using traditional manual and visual assessment of microscopy data, we identified distinct CD8+ T lymphocyte motility behaviors. However, such characterization is time and labor-intensive. In this work, we trained and validated a machine-learning model for the automated classification of behaviors of CD8+ T lymphocytes interacting with astrocytes and neurons. A balanced random forest was trained for the binary classification of established classes of cell behaviors (synapse vs. kinapse) as well as visually identified behaviors (scanning, dancing, and poking). Feature selection was performed during 3-fold cross-validation using the minimum redundancy maximum relevance algorithm. Results show promising performances when tested on a held-out dataset of CD8+ T lymphocytes interacting with astrocytes with a new experimenter and a held-out independent dataset of CD8+ T lymphocytes interacting with neurons. When tested on the independent CD8+ T cell-neuron dataset, the final model achieved a binary classification accuracy of 0.82 and a 3-class accuracy of 0.79. This novel automated classification approach could significantly reduce the time required to label cell motility behaviors while facilitating the identification of interactions of T lymphocytes with multiple cell types.

Unbalanced credit fraud modeling based on bagging and bayesian optimization

Credit fraud modeling is a crucial area of research that is highly relevant to the credit loan industry. Effective risk management is a key factor in providing quality credit services and directly impacts the profitability and bad debt ratio of leading organizations in this sector. However, when the distribution of credit fraud data is highly unbalanced, it can lead to noise errors caused by information distortion, periodic statistical errors, and model biases during training. This can cause unfair results for the minority class (target class) and increase the risk of overfitting. While traditional data balancing methods can reduce bias in models towards the majority class in relatively unbalanced data, they may not be effective in highly unbalanced scenarios. To address this challenge, this paper proposes using Bagging algorithms such as Random Forest and Bagging to model highly unbalanced credit fraud data. Bayesian optimization is utilized to find hyperparameters and determine the accuracy of the minority class as an optimization function for the model, which is tested with real European credit card fraud data. The results of the proposed packing algorithms are compared with traditional data balancing methods such as Balanced Bagging and Balanced Random Forest. The study found that traditional data balancing methods may not be compatible with excessively unbalanced data, whereas Bagging algorithms show promise as a solution for modeling such data. The proposed method for finding hyperparameters effectively deals with highly unbalanced data. It achieved precision, recall, and F1-score for the minority category of 0.94, 0.81, and 0.87, respectively. The study emphasizes the importance of addressing the challenges associated with unbalanced credit fraud data to improve the accuracy and fairness of credit fraud models

Sentiment Analysis on Movie Review from Rotten Tomatoes Using Modified Balanced Random Forest Method and Word2Vec

The film industry is one of the impacts of the rapid development of technology. This causes the film industry to increase every year. In addition, technological developments also affect the public to make it easier to access various movies from various websites. With many choices of movies, people need to know the quality of various movies by knowing the reviews of these movies from other people. However, the large number of audience reviews of a movie makes it difficult for people to categorize good movies and bad movies. The solution to the problem is to perform sentiment analysis on movie reviews. In this research, the classification method used is Modified Balanced Random Forest. This method was chosen because it can overcome imbalanced data and can increase accuracy and reduce time complexity. In this research, Word2Vec is also used as feature extraction. This feature extraction was chosen because previous research explained that Word2Vec has the advantage of being able to show the contextual similarity of two words in the resulting vector. The best model produced from this research is a model built without using stemming in the preprocessing stage, using 300 dimensions in Word2Vec, and using the Modified Balanced Random Forest classification method which produces an f1-score of 84.15%.

Efficient mapping of RNA-binding residues in RNA-binding proteins using local sequence features of binding site residues in protein-RNA complexes.

Protein-RNA interactions play vital roles in plethora of biological processes such as regulation of gene expression, protein synthesis, mRNA processing and biogenesis. Identification of RNA-binding residues (RBRs) in proteins is essential to understand RNA-mediated protein functioning, to perform site-directed mutagenesis and to develop novel targeted drug therapies. Moreover, the extensive gap between sequence and structural data restricts the identification of binding sites in unsolved structures. However, efficient use of computational methods demanding only sequence to identify binding residues can bridge this huge sequence-structure gap. In this study, we have extensively studied protein-RNA interface in known RNA-binding proteins (RBPs). We find that the interface is highly enriched in basic and polar residues with Gly being the most common interface neighbor. We investigated several amino acid features and developed a method to predict putative RBRs from amino acid sequence. We have implemented balanced random forest (BRF) classifier with local residue features of protein sequences for prediction. With 5-fold cross-validations, the sequence pattern derived dipeptide composition based BRF model (DCP-BRF) resulted in an accuracy of 87.9%, specificity of 88.8%, sensitivity of 82.2%, Mathew's correlation coefficient of 0.60 and AUC of 0.93, performing better than few existing methods. We further validated our prediction model on known human RBPs through RBR prediction and could map ~54% of them. Further, knowledge of binding site preferences obtained from computational predictions combined with experimental validations of potential RNA binding sites can enhance our understanding of protein-RNA interactions. This may serve to accelerate investigations on functional roles of many novel RBPs.

Development and benchmarking of machine learning models to classify patients suitable for outpatient lower extremity joint arthroplasty

Study objectivePerforming hip or knee arthroplasty as an outpatient surgery has been shown to be operationally and financially beneficial for selected patients. By applying machine learning models to predict patients suitable for outpatient arthroplasty, health care systems can better utilize resources efficiently. The goal of this study was to develop predictive models for identifying patients likely to be discharged same-day following hip or knee arthroplasty. DesignModel performance was assessed with 10-fold stratified cross-validation, evaluated over baseline determined by the proportion of eligible outpatient arthroplasty over sample size. The models used for classification were logistic regression, support vector classifier, balanced random forest, balanced bagging XGBoost classifier, and balanced bagging LightGBM classifier. SettingThe patient records were sampled from arthroplasty procedures at a single institution from October 2013 to November 2021. PatientsThe electronic intake records of 7322 knee and hip arthroplasty patients were sampled for the dataset. After data processing, 5523 records were kept for model training and validation. InterventionsNone. MeasurementsThe primary measures for the models were the F1-score, area under the receiver operating characteristic curve (ROCAUC), and area under the precision-recall curve. To measure feature importance, the SHapley Additive exPlanations value (SHAP) were reported from the model with the highest F1-score. ResultsThe best performing classifier (balanced random forest classifier) achieved an F1-score of 0.347: an improvement of 0.174 over baseline and 0.031 over logistic regression. The ROCAUC for this model was 0.734. Using SHAP, the top determinant features of the model included patient sex, surgical approach, surgery type, and body mass index. ConclusionsMachine learning models may utilize electronic health records to screen arthroplasty procedures for outpatient eligibility. Tree-based models demonstrated superior performance in this study.

Improving Wildfire Danger Assessment Using Time Series Features of Weather and Fuel in the Great Xing’an Mountain Region, China

Wildfires directly threaten the safety of life and property. Predicting wildfires with a model driven by wildfire danger factors can significantly reduce losses. Weather conditions continuously influence the drying rate of fuel as well as the occurrence probability and danger degree of wildfires. Previous studies have paid little attention to the continuous effects of weather and fuel on wildfires. This study improved the accuracy and effect of wildfire danger assessment using the time series features of weather and fuel. First, the time series features of weather and fuel factors within the 16 days before the fire were analyzed. Then, four feature groups were selected—feature group without time series values, feature group with time series values, feature group with Tsfresh transformation of time series values, and feature group with gradient and cumulative transformation of time series values—and three models were trained, respectively: random forest, balanced random forest, and extreme gradient boosting. The results showed that the f1-score of all feature groups with time series values (0.93) increased by 0.15, on average, compared with those without time series values (0.78) for the three models. The feature group with gradient and cumulative features had a more stable prediction accuracy and a more accurate wildfire danger map. The results suggest that using the appropriate time series features of weather and fuel can help improve the precision and effect of the wildfire danger assessment model.

Exploration of machine learning methods for maritime risk predictions

ABSTRACT Maritime applications such as targeting ships for inspections, improved domain awareness, and dynamic risk exposure assessments for strategic planning all benefit from ship-specific incident probabilities. Using a unique and comprehensive global data set, of 1.2 million observations over the period from 2014 to 2020, this study explores the effectiveness and suitability of 144 model variants from the field of machine learning for eight incident endpoints of interest and evaluating over 580 covariates. Furthermore, the importance of covariates is examined and visualized. The results differ for each endpoint of interest but confirm that random forest methods can improve prediction capabilities. Based on out-of-sample evaluations for the year 2020, targeting the top 10% most risky vessels would improve predictions by a factor of 2.7 to 4.9 compared to random selection and based on the top decile lift. Balanced random forests and random forests with balanced training variants outperform regular random forests, for which the selected variants also depend on aggregation types. The most important covariate groups for predicting incident probabilities relate to beneficial ownership, the safety management company, and the size and age of the vessel, while the relevance of these factors remains similar across the different endpoints of interest.

Faster and more accurate pathogenic combination predictions with VarCoPP2.0

BackgroundThe prediction of potentially pathogenic variant combinations in patients remains a key task in the field of medical genetics for the understanding and detection of oligogenic/multilocus diseases. Models tailored towards such cases can help shorten the gap of missing diagnoses and can aid researchers in dealing with the high complexity of the derived data. The predictor VarCoPP (Variant Combinations Pathogenicity Predictor) that was published in 2019 and identified potentially pathogenic variant combinations in gene pairs (bilocus variant combinations), was the first important step in this direction. Despite its usefulness and applicability, several issues still remained that hindered a better performance, such as its False Positive (FP) rate, the quality of its training set and its complex architecture.ResultsWe present VarCoPP2.0: the successor of VarCoPP that is a simplified, faster and more accurate predictive model identifying potentially pathogenic bilocus variant combinations. Results from cross-validation and on independent data sets reveal that VarCoPP2.0 has improved in terms of both sensitivity (95% in cross-validation and 98% during testing) and specificity (5% FP rate). At the same time, its running time shows a significant 150-fold decrease due to the selection of a simpler Balanced Random Forest model. Its positive training set now consists of variant combinations that are more confidently linked with evidence of pathogenicity, based on the confidence scores present in OLIDA, the Oligogenic Diseases Database (https://olida.ibsquare.be). The improvement of its performance is also attributed to a more careful selection of up-to-date features identified via an original wrapper method. We show that the combination of different variant and gene pair features together is important for predictions, highlighting the usefulness of integrating biological information at different levels.ConclusionsThrough its improved performance and faster execution time, VarCoPP2.0 enables a more accurate analysis of larger data sets linked to oligogenic diseases. Users can access the ORVAL platform (https://orval.ibsquare.be) to apply VarCoPP2.0 on their data.

Screening unknown novel psychoactive substances using GC–MS based machine learning

In recent years, there is a large increase in structural diversity of novel psychoactive substances (NPS), exacerbating drug abuse issues as these variants evade classical detection methods such as spectral library matching. Gas chromatography mass spectrometry (GC–MS) is commonly used to identify these NPS. To tackle this issue, machine learning models are developed to address the analytical challenge of identifying unknown NPS, using only GC–MS data. 891 GC–MS spectra are used to train and evaluate multiple supervised machine learning classifiers, namely artificial neural network (ANN), convolutional neural network (CNN) and balanced random forest (BRF). 7 classes, comprising 6 NPS classes (cathinone, cannabinoids, phenethylamine, piperazine, tryptamines and fentanyl) and other unrelated compounds can be effectively classified with a macro-F1 score of 0.9, averaged across all cross-validation folds. These results indicate that machine learning models are a promising complement as an effective NPS detection tool.

Using Machine Learning and Deep Learning Algorithms to Predict Postoperative Outcomes Following Anterior Cervical Discectomy and Fusion.

A retrospective cohort study from a multisite academic medical center. To construct, evaluate, and interpret a series of machine learning models to predict outcomes related to inpatient health care resource utilization for patients undergoing anterior cervical discectomy and fusion (ACDF). Reducing postoperative health care utilization is an important goal for improving the delivery of surgical care and serves as a metric for quality assessment. Recent data has shown marked hospital resource utilization after ACDF surgery, including readmissions, and ED visits. The burden of postoperative health care use presents a potential application of machine learning techniques, which may be capable of accurately identifying at-risk patients using patient-specific predictors. Patients 18-88 years old who underwent ACDF from 2011 to 2021 at a multisite academic center and had preoperative lab values within 3 months of surgery were included. Outcomes analyzed included 90-day readmissions, postoperative length of stay, and nonhome discharge. Four machine learning models-Extreme Gradient Boosted Trees, Balanced Random Forest, Elastic-Net Penalized Logistic Regression, and a Neural Network-were trained and evaluated through the Area Under the Curve estimates. Feature importance scores were computed for the highest-performing model per outcome through model-specific metrics. A total of 1026 cases were included in the analysis cohort. All machine learning models were predictive for outcomes of interest, with the Random Forest algorithm consistently demonstrating the strongest average area under the curve performance, with a peak performance of 0.84 for nonhome discharge. Important features varied per outcome, though age, body mass index, American Society of Anesthesiologists classification >2, and medical comorbidities were highly weighted in the studied outcomes. Machine learning models were successfully applied and predictive of postoperative health utilization after ACDF. Deployment of these tools can assist clinicians in determining high-risk patients. III.

Prediction of Postoperative Pulmonary Edema Risk Using Machine Learning.

Postoperative pulmonary edema (PPE) is a well-known postoperative complication. We hypothesized that a machine learning model could predict PPE risk using pre- and intraoperative data, thereby improving postoperative management. This retrospective study analyzed the medical records of patients aged > 18 years who underwent surgery between January 2011 and November 2021 at five South Korean hospitals. Data from four hospitals (n = 221,908) were used as the training dataset, whereas data from the remaining hospital (n = 34,991) were used as the test dataset. The machine learning algorithms used were extreme gradient boosting, light-gradient boosting machine, multilayer perceptron, logistic regression, and balanced random forest (BRF). The prediction abilities of the machine learning models were assessed using the area under the receiver operating characteristic curve, feature importance, and average precisions of precision-recall curve, precision, recall, f1 score, and accuracy. PPE occurred in 3584 (1.6%) and 1896 (5.4%) patients in the training and test sets, respectively. The BRF model exhibited the best performance (area under the receiver operating characteristic curve: 0.91, 95% confidence interval: 0.84-0.98). However, its precision and f1 score metrics were not good. The five major features included arterial line monitoring, American Society of Anesthesiologists physical status, urine output, age, and Foley catheter status. Machine learning models (e.g., BRF) could predict PPE risk and improve clinical decision-making, thereby enhancing postoperative management.

Predicting no-show appointments in a pediatric hospital in Chile using machine learning

The Chilean public health system serves 74% of the country’s population, and 19% of medical appointments are missed on average because of no-shows. The national goal is 15%, which coincides with the average no-show rate reported in the private healthcare system. Our case study, Doctor Luis Calvo Mackenna Hospital, is a public high-complexity pediatric hospital and teaching center in Santiago, Chile. Historically, it has had high no-show rates, up to 29% in certain medical specialties. Using machine learning algorithms to predict no-shows of pediatric patients in terms of demographic, social, and historical variables. To propose and evaluate metrics to assess these models, accounting for the cost-effective impact of possible intervention strategies to reduce no-shows. We analyze the relationship between a no-show and demographic, social, and historical variables, between 2015 and 2018, through the following traditional machine learning algorithms: Random Forest, Logistic Regression, Support Vector Machines, AdaBoost and algorithms to alleviate the problem of class imbalance, such as RUS Boost, Balanced Random Forest, Balanced Bagging and Easy Ensemble. These class imbalances arise from the relatively low number of no-shows to the total number of appointments. Instead of the default thresholds used by each method, we computed alternative ones via the minimization of a weighted average of type I and II errors based on cost-effectiveness criteria. 20.4% of the 395,963 appointments considered presented no-shows, with ophthalmology showing the highest rate among specialties at 29.1%. Patients in the most deprived socioeconomic group according to their insurance type and commune of residence and those in their second infancy had the highest no-show rate. The history of non-attendance is strongly related to future no-shows. An 8-week experimental design measured a decrease in no-shows of 10.3 percentage points when using our reminder strategy compared to a control group. Among the variables analyzed, those related to patients’ historical behavior, the reservation delay from the creation of the appointment, and variables that can be associated with the most disadvantaged socioeconomic group, are the most relevant to predict a no-show. Moreover, the introduction of new cost-effective metrics significantly impacts the validity of our prediction models. Using a prototype to call patients with the highest risk of no-shows resulted in a noticeable decrease in the overall no-show rate.

Exploring the Potential of Machine Learning Techniques for Predicting Travel Insurance Claims: A Comparative Analysis of Four Models

Travel insurance is a crucial component for any traveller as it offers protection against financial losses resulting from unforeseen events during a trip, such as trip cancellations, medical emergencies, lost luggage, and related issues. This study aims to investigate the potential of machine learning (ME) techniques for predicting the probability of travel insurance claims. In order to tackle the issue of managing extensive and intricate datasets, advanced statistical techniques were employed, including keyword extraction, feature extraction, and Chi-squared tests. Our evaluation of four popular ML models, namely balanced random forest (BRF), support vector machines (SVM), logistic regression (LR), and balanced bagging (BB), highlight that the BRF model outperforms the other models in predicting travel insurance claims. Our study emphasises the advantages of utilising machine learning algorithms in processing large datasets, producing predictions on future insurance claims, and adapting to changing circumstances, thus serving as a valuable tool for practitioners in the travel insurance industry.

Evaluation of four machine learning models for signal detection.

Logistic regression-based signal detection algorithms have benefits over disproportionality analysis due to their ability to handle potential confounders and masking factors. Feature exploration and developing alternative machine learning algorithms can further strengthen signal detection. Our objective was to compare the signal detection performance of logistic regression, gradient-boosted trees, random forest and support vector machine models utilizing Food and Drug Administration adverse event reporting system data. Cross-sectional study. The quarterly data extract files from 1 October 2017 through 31 December 2020 were downloaded. Due to an imbalanced outcome, two training sets were used: one stratified on the outcome variable and another using Synthetic Minority Oversampling Technique (SMOTE). A crude model and a model with tuned hyperparameters were developed for each algorithm. Model performance was compared against a reference set using accuracy, precision, F1 score, recall, the receiver operating characteristic area under the curve (ROCAUC), and the precision-recall curve area under the curve (PRCAUC). Models trained on the balanced training set had higher accuracy, F1 score and recall compared to models trained on the SMOTE training set. When using the balanced training set, logistic regression, gradient-boosted trees, random forest and support vector machine models obtained similar performance evaluation metrics. The gradient-boosted trees hyperparameter tuned model had the highest ROCAUC (0.646) and the random forest crude model had the highest PRCAUC (0.839) when using the balanced training set. All models trained on the balanced training set performed similarly. Logistic regression models had higher accuracy, precision and recall. Logistic regression, random forest and gradient-boosted trees hyperparameter tuned models had a PRCAUC ⩾ 0.8. All models had an ROCAUC ⩾ 0.5. Including both disproportionality analysis results and additional case report information in models resulted in higher performance evaluation metrics than disproportionality analysis alone.

Deep Attention-based Supernovae Classification of Multiband Light Curves

In astronomical surveys, such as the Zwicky Transient Facility, supernovae (SNe) are relatively uncommon objects compared to other classes of variable events. Along with this scarcity, the processing of multiband light curves is a challenging task due to the highly irregular cadence, long time gaps, missing values, few observations, etc. These issues are particularly detrimental to the analysis of transient events: SN-like light curves. We offer three main contributions: (1) Based on temporal modulation and attention mechanisms, we propose a deep attention model (TimeModAttn) to classify multiband light curves of different SN types, avoiding photometric or hand-crafted feature computations, missing-value assumptions, and explicit imputation/interpolation methods. (2) We propose a model for the synthetic generation of SN multiband light curves based on the Supernova Parametric Model, allowing us to increase the number of samples and the diversity of cadence. Thus, the TimeModAttn model is first pretrained using synthetic light curves. Then, a fine-tuning process is performed. The TimeModAttn model outperformed other deep learning models, based on recurrent neural networks, in two scenarios: late-classification and early-classification. Also, the TimeModAttn model outperformed a Balanced Random Forest (BRF) classifier (trained with real data), increasing the balanced-F 1score from ≈.525 to ≈.596. When training the BRF with synthetic data, this model achieved a similar performance to the TimeModAttn model proposed while still maintaining extra advantages. (3) We conducted interpretability experiments. High attention scores were obtained for observations earlier than and close to the SN brightness peaks. This also correlated with an early highly variability of the learned temporal modulation.

Personalized screening and risk profiles for Mild Cognitive Impairment via a Machine Learning Framework: Implications for general practice

ObjectivesDiagnosis of Mild Cognitive Impairment (MCI) requires lengthy diagnostic procedures, typically available at tertiary Health Care Centers (HCC). This prospective study evaluated a flexible Machine Learning (ML) framework toward identifying persons with MCI or dementia based on information that can be readily available in a primary HC setting. MethodsDemographic and clinical data, informant ratings of recent behavioral changes, self-reported anxiety and depression symptoms, subjective cognitive complaints, and Mini Mental State Examination (MMSE) scores were pooled from two aging cohorts from the island of Crete, Greece (N = 763 aged 60–93 years) comprising persons diagnosed with MCI (n = 277) or dementia (n = 153), and cognitively non-impaired persons (CNI, n = 333). A Balanced Random Forest Classifier was used for classification and variable importance-based feature selection in nested cross-validation schemes (CNI vs MCI, CNI vs Dementia, MCI vs Dementia). Global-level model-agnostic analyses identified predictors displaying nonlinear behavior. Local level agnostic analyses pinpointed key predictor variables for a given classification result after statistically controlling for all other predictors in the model. ResultsClassification of MCI vs CNI was achieved with improved sensitivity (74 %) and comparable specificity (73 %) compared to MMSE alone (37.2 % and 94.3 %, respectively). Additional high-ranking features included age, education, behavioral changes, multicomorbidity and polypharmacy. Higher classification accuracy was achieved for MCI vs Dementia (sensitivity/specificity = 87 %) and CNI vs Dementia (sensitivity/specificity = 94 %) using the same set of variables. Model agnostic analyses revealed notable individual variability in the contribution of specific variables toward a given classification result. ConclusionsImproved capacity to identify elderly with MCI can be achieved by combining demographic and medical information readily available at the PHC setting with MMSE scores, and informant ratings of behavioral changes. Explainability at the patient level may help clinicians identify specific predictor variables and patient scores to a given prediction outcome toward personalized risk assessment.

Machine Learning Model Development and Validation for Predicting Outcome in Stage 4 Solid Cancer Patients with Septic Shock Visiting the Emergency Department: A Multi-Center, Prospective Cohort Study.

A reliable prognostic score for minimizing futile treatments in advanced cancer patients with septic shock is rare. A machine learning (ML) model to classify the risk of advanced cancer patients with septic shock is proposed and compared with the existing scoring systems. A multi-center, retrospective, observational study of the septic shock registry in patients with stage 4 cancer was divided into a training set and a test set in a 7:3 ratio. The primary outcome was 28-day mortality. The best ML model was determined using a stratified 10-fold cross-validation in the training set. A total of 897 patients were included, and the 28-day mortality was 26.4%. The best ML model in the training set was balanced random forest (BRF), with an area under the curve (AUC) of 0.821 to predict 28-day mortality. The AUC of the BRF to predict the 28-day mortality in the test set was 0.859. The AUC of the BRF was significantly higher than those of the Sequential Organ Failure Assessment score and the Acute Physiology and Chronic Health Evaluation II score (both p &lt; 0.001). The ML model outperformed the existing scores for predicting 28-day mortality in stage 4 cancer patients with septic shock. However, further studies are needed to improve the prediction algorithm and to validate it in various countries. This model might support clinicians in real-time to adopt appropriate levels of care.