Search

Post-pandemic student performance prediction has gained considerable attention within the academic community. Educational institutions worldwide have adopted blended learning approaches for delivering education and assessments. In this study, we harnessed data from the 'Offee' assessment platform based in Mumbai, India, spanning from September 2020 to October 2022. This dataset allows us to analyze the real impact of the Covid-19 pandemic on examinations. To tackle the challenges posed by high-dimensional data, we employed feature selection techniques to retain only the most significant features for predicting student performance. Specifically, we assessed the effectiveness of two decision tree-based methods, J48 classifier and Random Forest, in conjunction with the ranker search method using wrapper subset evaluation. The resulting feature set was then utilized to train and test a Support Vector Machine model for predicting student performance. Our findings reveal that both J48 and Random Forest feature selection methods yield impressive precision accuracy rates for student performance prediction, achieving 93.14% and 86.24%, respectively. In contrast, using the dataset without feature selection resulted in lower precision accuracy, standing at 79.9%, and consumed more time during model training and development. In summary, this study underscores the significance of feature selection in enhancing the accuracy of student performance prediction compared to utilizing the dataset without such feature curation. [ABSTRACT FROM AUTHOR]

Metabolic syndrome increases cardiovascular risk, necessitating accurate prediction models. However, prior studies demonstrate key gaps including small datasets limiting generalizability, insufficient model comparisons constraining algorithm selection, and lack of model interpretation impeding biomarker identification. This study addresses these limitations through a robust methodology. Using a dataset of 2400 patients, larger than previous studies, we evaluated nine machine learning classifiers including Logistic Regression, KNN, SVC, Decision Tree Classifier, Random Forest Classifier, Gradient Boosting Classifier, AdaBoost Classifier, XGBoost Classifier, and LightGBM Classifier, after optimized preprocessing like hyperparameter tuning to tackle overfitting. The XGBoost model outperformed other algorithms with 95% training accuracy and 88.97% testing accuracy owing to advance nonlinear data handling. It achieved high precision, recall, and a 0.913 F1 score. Feature importance analysis revealed waist circumference as the most predictive biomarker. Our extensive model comparisons over sizable data significantly advance metabolic syndrome prediction capabilities compared to prior work. The optimized and interpretable XGBoost model enables early risk identification to guide timely interventions mitigating associated complications. Additional validation across diverse populations can further confirm generalizability. This study successfully demonstrates a multifaceted machine learning approach to address key gaps in literature related to model evaluation, selection, and interpretation. [ABSTRACT FROM AUTHOR]

An enormous number of deaths occur every year as a result of heart disease, making it a major concern in world health. Improving patient outcomes and lowering death rates, early detection and correct diagnosis of cardiac disease play a key role. When the heart's arteries become blocked, oxygen-poor blood cannot reach the heart properly, resulting in coronary heart disease. Early detection of cardiac disease is viable because it reduces medical costs and potentially saves the patient's life. Recently presented methods have improved heart failure detection accuracy on testing data without sacrificing accuracy on training data, yet most of these algorithms are suffering from the issue of overfitting. Models that were created end up fitting the test data too well. In this study, we create a novel diagnostic system to address this issue, and the resulting system demonstrates high intelligence and excellent performance on both training and testing data. Machine learning (ML) algorithms have demonstrated promising potential in assisting healthcare professionals with timely and accurate diagnosis. In this paper, is based on supervised machine learning methods are decision tree (DT), random forest (RF), Support vector Machine (SVM), Principal Component Analysis(PCA). We compare their accuracy with each other by using bar plot. [ABSTRACT FROM AUTHOR]

Spare-parts provisioning strategy is crucial for airlines or Maintenance Repair Overhauls (MRO). The strategy is divided into multiple classes that refer to several criteria, then uses Material Requirements Planning (MRP) to run those strategies. However, classifying spare parts into a strategy is onerous and requires innovation. Therefore, we employ three algorithms: Decision Tree, Random Forest, and Support Vector Machine (SVM), and use spare parts and aircraft utilization data to classify the strategy and optimize spare parts provisioning of MRO over time. The results showed Random Forest performs best by accuracy, sensitivity, and specificity with a score of more than 97%. [ABSTRACT FROM AUTHOR]

Estimating an individual's height from a two-dimensional (2D) image has emerged as a focal point of investigation within Computer Vision. Notably, the research landscape has witnessed endeavors dedicated to human height estimation from frontal-face and full-body images. Nevertheless, there exists an imperative need for further exploration within this domain, primarily aimed at enhancing the precision of such estimations. Within this context, our study seeks to delve into the nuanced task of height from a solitary full-body image. Leveraging cutting-edge AI methodologies, we have harnessed the capabilities of Yolov7 for human pose estimation, alongside the DeepFace pre-trained model for age, gender, and race estimation, to serve as pivotal feature extraction mechanisms. Notably, our model's efficacy is intrinsically linked to the performance benchmarks set by these two foundational models. Our empirical observations underscore the commendable performance of Yolov7; however, it is incumbent to acknowledge that the DeepFace model has, regrettably, not demonstrated commensurate levels of accuracy in our experimentation. In the present study, we have employed a set of regression and tree-based predictive models, including Linear Regression, AdaBoost Regressor, Decision Tree, and Random Forest. Notably, the AdaBoost Regressor model has exhibited superior performance compared to the other models under consideration. Specifically, it has demonstrated a remarkable mean absolute error of 6.2 cms and a mean squared error of 7.9 cms, thereby establishing its efficacy in minimizing prediction errors in this context. [ABSTRACT FROM AUTHOR]

• It is possible to obtain good classification results of cardiovascular autonomic neuropathy using supervised machine learning. • The best classifier is the random forest classifier with a F1-score of 60% on the reduced dataset. • In the ensembled data no standard clinical biochemical analytes were present in the 10 best classifiers. Using supervised machine learning to classify the severity of cardiovascular autonomic neuropathy (CAN). The aims were 1) to investigate which features contribute to characterising CAN 2) to generate an ensembled set of features that best describes the variation in CAN classification. Eighty-two features from demographic, beat-to-beat, biochemical, and inflammation were obtained from 204 people with diabetes and used in three machine-learning-classifiers, these are: support vector machine, decision tree, and random forest. All data were ensembled using a weighted mean of the features from each classifier. The 10 most important features derived from the domains: Beat-to-beat, inflammation markers, disease-duration, and age. Beat-to-beat measures associate with CAN as diagnosis is mainly based on cardiac reflex responses, disease-duration and age are also related to CAN development throughout disease progression. The inflammation markers may reflect the underlying disease process, and therefore, new treatment modalities targeting systemic low-grade inflammation should potentially be tested to prevent the development of CAN. Cardiac reflex responses should be monitored closely to diagnose and classify severity levels of CAN accurately. Standard clinical biochemical analytes, such as glycaemic level, lipidic level, or kidney function were not included in the ten most important features. Beat-to-beat measures accounted for approximately 60% of the features in the ensembled data. [ABSTRACT FROM AUTHOR]

This study is targeted to use AI (artificial intelligence) to predict the English proficiency test scores (CET-4) of students in private universities. The study will train and test multiple machine learning models using a dataset of students' demographic information, English language background, and test scores. These models will use different algorithms such as linear regression, decision trees, and neural networks to identify patterns and relationships in the data and make accurate predictions. The study aims to develop a reliable and effective predictive tool for educators to assess students' language proficiency levels and provide targeted teaching strategies. Comprehensive evaluation of students' overall quality will play a crucial role in the prediction model. Factors such as educational resources, teaching methods, student learning habits, and learning atmosphere in each college will affect the accuracy of the prediction. The full-coverage classification prediction of student performance using modern scientific and technological strategies holds great research significance for private universities' educational tasks and targeted teaching plans. [ABSTRACT FROM AUTHOR]

Gesture recognition is able to transform how we interact with technology, including during presentations. We aim to classify slide-control gestures and create a model that allows users to add their personalized gestures to to control presentation com- mands. To avoid wasting time on capturing and training the model, we aim to construct a model with a short training time and high accuracy. We assessed the efficiency of various multi-class classification methods for gesture recognition in presentation settings. We compared K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Decision Tree, Linear Discriminant Analysis (LDA), and Random Forest. The results demonstrated that KNN offers the shortest training time while maintaining high accuracy. The findings of this study provide a basis for future research on gesture recognition technology and its imple- mentation to improve presentation experience. [ABSTRACT FROM AUTHOR]

The properties of silver nanoparticles (AgNPs) are affected by various parameters, making optimisation of their synthesis a laborious task. This optimisation is facilitated in this work by concurrent use of a T-junction microfluidic system and machine learning approach. The AgNPs are synthesized by reducing silver nitrate with tannic acid in the presence of trisodium citrate, which has a dual role in the reaction as reducing and stabilizing agent. The study uses a decision tree-guided design of experiment method for the size of AgNPs. The developed approach uses kinetic nucleation and growth constants derived from an independent set of experiments to account for chemistry of synthesis, the Reynolds number and the ratio of Dean number to Reynolds number to reveal effect of hydrodynamics and mixing within device and storage temperature to account for particle stability after collection. The obtained model was used to define a parameter space for additional experiments carried out to improve the model further. The numerical results illustrate that well-designed experiments can contribute more effectively to the development of different machine learning models (decision tree, random forest and XGBoost) as opposed to randomly added experiments. [Display omitted] • Decision tree guided synthesis is an effective strategy to design experiments. • Three models were used to test the performance of each area of parametric space. • The proposed method allowed the use of kinetics to predict the size of AgNPs. [ABSTRACT FROM AUTHOR]

IoT-based applications have witnessed a rapid surge in deployment in various domains. IoT infrastructure is the nervous system responsible for the effective functioning of Smart Cities. Nevertheless, the full-fledged deployment of IoT-based applications exposes this infrastructure to a high risk of cyberattacks. Since IoT devices establish communication with cloud services via inbuilt sensors, the probability of sabotaging the communication channel by malicious attacks is always high. This paper aims to explore an anomaly detection method that makes use of techniques like Support Vector Machine (SVM), Artificial Neural Networks (ANN), k- Nearest Neighbor (KNN), Linear Regression (LR), Decision Trees (DT), and Random Forest (RF) to neutralize threats and boost the cybersecurity of a smart city. The paper goes on to examine the role of ensemble techniques like bagging and boosting to provide an additional security layer to the detection architecture. This is where the paper departs from the erstwhile approaches that have revolved around single classifiers to boost the detection system and have not considered the integration of cross-validation and feature selection. The experimental results conducted on the datasets UNSW-BC15 and CICIDS2017 and several measures like Accuracy, Precision, Recall, and F1 Score establish that the proposed approach outperforms various state-of-the-art methods used in detecting rare attacks. [ABSTRACT FROM AUTHOR]

This paper presents tree-based machine learning techniques like decision trees, and random forest algorithms to categorize various permanent faults such as a line to ground, line to line, line-line to ground, and line-line-line to ground fault simulated in an underground cable. The simulation has been performed with the variation in fault types, fault location, fault resistance, and fault inception angle and the sending end current signals are presented as input to the classifier for classifying the fault. A relative comparison between the decision tree technique and random forest technique has been performed and the random forest algorithm turns out to be the best for classifying the permanent faults in an underground cable. Both the algorithms are deliberated by precision, F1 score, and recall. The best result is presented by Random Forest Algorithm with an accuracy of 98.8%. [ABSTRACT FROM AUTHOR]

Students face problems that might hinder their academic pursuit toward success, problems ranging from trivial matters such as class condition, feeling of the student to severe matters such as family breakdown, economic reasons, and many more. This is a major problem because students shape the future of a nation – which will affect many things in the future. Teachers are looking for an effective way to find what might generally be the best solution for solving certain problems, as each student may face different problems, solving one at a time is not possible with the number of students each year. In this paper, we will try to find factors that might hinder or improve student performance using Pearson correlation between each feature toward the student G3 result. Based on the result, past failures will negatively impact student grades with -0.360415 correlation, and then Mother's Education will positively impact student grades with 0.217147. After finding out which factor affects student grade, we try to predict student grade using ML models to prove whether that factor actually affects student grade. Our MLP 12-Neuron model performs the best with RMSE value of 4.32, followed by Random Forest with RMSE value of 4.52, and finally Decision Tree with RMSE value of 5.69. [ABSTRACT FROM AUTHOR]

A comprehensive multi-scale computational strategy was developed in this study based on mass transfer and machine learning for simulation of drug concentration distribution in a biomaterial matrix. The controlled release was modeled and validated via the hybrid model. Mass transfer equations along with kinetics models were solved numerically and the results were then used for machine learning models. We investigated the performance of three regression models, namely Decision Tree (DT), Random Forest (RF), and Extra Tree (ET) in predicting medicine concentration (C) based on r and z data. Hyper-parameter optimization is conducted using Glowworm Swarm Optimization (GSO). Results revealed high predictive accuracy across all models, with ET demonstrating superior performance, achieving a coefficient of determination value (R2) of 0.99854, an RMSE of 1.1446E-05, and a maximum error of 6.49087E-05. DT and RF also exhibit notable performance, with coefficients of determination equal to 0.99571 and 0.99655, respectively. These results highlight the effectiveness of ensemble tree-based methods in accurately predicting chemical concentrations, with Extra Tree (ET) Regression emerging as the most promising model for this specific dataset. • Developing a comprehensive computational strategy for drug release modeling. • Simulation of the distribution of drug concentrations in a biomaterial matrix. • Investigation of the performance of three regression models for drug release. • Use of Glowworm Swarm Optimization (GSO) for Hyper-parameter optimization. [ABSTRACT FROM AUTHOR]

Random Forest is an ensemble of decision trees based on the bagging and random subspace concepts. As suggested by Breiman, the strength of unstable learners and the diversity among them are the ensemble models' core strength. In this paper, we propose two approaches known as oblique and rotation double random forests. In the first approach, we propose rotation based double random forest. In rotation based double random forests, transformation or rotation of the feature space is generated at each node. At each node different random feature subspace is chosen for evaluation, hence the transformation at each node is different. Different transformations result in better diversity among the base learners and hence, better generalization performance. With the double random forest as base learner, the data at each node is transformed via two different transformations namely, principal component analysis and linear discriminant analysis. In the second approach, we propose oblique double random forest. Decision trees in random forest and double random forest are univariate, and this results in the generation of axis parallel split which fails to capture the geometric structure of the data. Also, the standard random forest may not grow sufficiently large decision trees resulting in suboptimal performance. To capture the geometric properties and to grow the decision trees of sufficient depth, we propose oblique double random forest. The oblique double random forest models are multivariate decision trees. At each non-leaf node, multisurface proximal support vector machine generates the optimal plane for better generalization performance. Also, different regularization techniques (Tikhonov regularization, axis-parallel split regularization, Null space regularization) are employed for tackling the small sample size problems in the decision trees of oblique double random forest. The proposed ensembles of decision trees produce trees with bigger size compared to the standard ensembles of decision trees as bagging is used at each non-leaf node which results in improved performance. The evaluation of the baseline models and the proposed oblique and rotation double random forest models is performed on benchmark 121 UCI datasets and real-world fisheries datasets. Both statistical analysis and the experimental results demonstrate the efficacy of the proposed oblique and rotation double random forest models compared to the baseline models on the benchmark datasets. [ABSTRACT FROM AUTHOR]

Diabetes has been an offensive condition in recent years, and it can lead to major health problems. If diabetes is not addressed, it can lead to a variety of health problems, including heart disease, stroke, blindness, and kidney failure. Diabetes condition must be addressed promptly to avoid a significant health risk. Machine learning algorithms can assist the doctor in identifying and diagnosing diabetes and other diseases. Different types of classifiers have been used to diagnose diabetes. To improve the performance of integrated flexible individual classifiers and lower the possibility of misclassifying a single instance, an ensemble approach named "Stacking Classifier" was developed. Several classifiers, such as Naïve Bayes, KNN, Linear regression, and decision tree (DT) were used but all these models have low accuracy. However, additional study is needed to detect diabetic condition due to a lack of major work and low accuracy. Therefore this study proposed an ensemble technique termed "Stacking Classifier" was designed to increase the performance of integrated flexible individual classifiers and reduce the probability of misclassifying a single instance. This study uses a variety of classifiers, including Naïve Bayes, KNN, Linear Discriminant Analysis, and Decision Tree, with Random Forest functioning as a Meta classifier. In terms of F-measure, Recall, Accuracy, and Precision, the proposed stacking classifier achieves a higher accuracy of 97.35 % when compared to current models such as Nave Naïve Bayes, KNN, Decision Tree, and Linear Discriminant Analysis, which are 74.60 %, 78.57 %, and 77.35 %, respectively. [ABSTRACT FROM AUTHOR]

Decision tree is a widely used non-parametric technique in machine learning, data mining and pattern recognition. It is simple to understand and interpret, however it faces challenges such as handling higher dimensional and class imbalanced datasets, over-fitting and instability. To overcome some of these issues, vertical partitioning approaches like serial partitioning, theme based partitioning are used in the literature. A vertical partitioning approach divides the feature set into subsets of features (blocks) and makes use of these subsets for subsequent tasks. In this work, we use the ideas of music rhythm tree to propose a novel vertical partitioning technique. It orders the features based on the average correlation strength of the features before partitioning the feature set. The proposed method is proved to be superior by showing an average of 13.8 % , 6 % , 9.8 % , 19.7 % , 9.4 % , and 29.4 % higher classification accuracy over C4.5, Random Forest, Bagging, Adaboost, an ensemble technique and a vertical partitioning technique respectively. Our empirical results on 15 datasets demonstrate that the proposed vertical partitioning method is more stable and better in handling class-imbalanced data. Finally, some popular statistical tests are conducted to validate the statistical significance of the results of the proposed method. [ABSTRACT FROM AUTHOR]

In this work, the gas phase photocatalytic CO 2 reduction was analyzed via machine learning, and the results were compared with those obtained in liquid phase process. Total 549 data points (268 for gas and 281 for liquid phase) were extracted from 80 published papers for this purpose. The general trends in the literature were analyzed using simple descriptive statistics first; then, the random forest (RF) regression was used for the band gap prediction while the decision tree (DT) classification was utilized to deduce heuristics for higher CO 2 reduction rates. It was found that H 2 , CO, and CH 4 are the main products in gas phase while CH 3 OH production is more dominant in liquid phase. Random forest prediction was quite successful in predicting the band gap with the root mean square error of 0.15 for testing. Decision tree models for the total gas production rates were also successful; for example, the accuracy rates for training and testing were 80% and 79% respectively for the gas phase processes. The high precision for the high gas production class allowed to deduce some rules indicating the semiconductor and co-catalysts options for high CO 2 photoreduction rates while the reaction temperatures was also found to be influential in the liquid phase. [Display omitted] • Photocatalytic CO 2 reduction was studied using machine learning. • Descriptive statistics, decision tree (DT) and random forest (RF) were used. • The bandgap prediction with RF was quite successful. • Heuristic rules were deduced for high total gas production rate using DT. • Semiconductor and co-catalyst were found to be deterministic. [ABSTRACT FROM AUTHOR]

A respiratory syndrome COVID-19 pandemic has become a serious public health issue nowadays. The COVID-19 virus has been affecting tens of millions people worldwide. Some of them have recovered and have been released. Others have been isolated and few others have been unfortunately deceased. In this paper, we apply and compare different machine learning approaches such as decision tree models, random forest, and multinomial logistic regression to predict isolation, release, and decease states for COVID-19 patients in South Korea. The prediction can help health providers and decision makers to distinguish the states of infected patients based on their features in early intervention to take an action either by releasing or isolating the patient after the infection. The proposed approaches are evaluated using Data Science for COVID-19 (DS4C) dataset. An analysis of DS4C dataset is also provided. Experimental results and evaluation show that multinomial logistic regression outperforms other approaches with 95% in a state prediction accuracy and a weighted average F1-score of 95%. [ABSTRACT FROM AUTHOR]

Decision forests, such as random forest (RF) are widely used for tabular data, mainly due to their predictive performance and ease of usage. However, given that the forest's trees may produce contradictory predictions for a certain sample, the usage of decision forests in applications that involve decision-making necessitates further reliability assessment of the predictions for generating a trustworthy combined prediction. Model distillation using a born-again tree is a common approach for converting a decision forest into a single decision tree (DT) while preserving the forest's predictive performance and supporting decision-makers. In this paper, we introduce PnT (Path in Tree), a novel approach that learns a path-based encoding from a decision forest. PnT applies an iterative algorithm, where in each iteration, a batch of trees is trained and then used to identify informative paths. These paths are then encoded and utilized for PnT-DT, an approach for producing a contradiction-free born-again DT. We also show that PnT can be leveraged for PnT-RF, a born-again forest approach, capable of improving the predictive performance of a plain decision forest. We evaluate PnT-DT and PnT-RF on 40 classification datasets and demonstrate that both PnT-DT and PnT-RF significantly outperform existing state-of-the-art (SOTA) born-again DT and decision forest methods in terms of predictive performance. • PnT, a novel approach that learns a path-based encoding from a decision forest. • PnT-DT, a contradiction-free born-again tree, converting forest knowledge into a DT. • PnT-RF enhances RF's predictive performance via path-based encoding. [ABSTRACT FROM AUTHOR]

• Among decision tree-based MLs, RF Performance compared better than CART and XGB. • DyLEMa can reconstruct ET based on seasonal segregated atmospheric and land data. • DyLEMa ET PBIAS was reduced in temporal validation compared to the USGS ET data. • DyLEMa ET estimates are robust across various cloud contamination rates. • DyLEMa ET estimates are robust across Landsat sensor failure. This study presents the "Dynamic Land Cover Evapotranspiration Model Algorithm: DyLEMa" for continuous spatiotemporal evapotranspiration (ET) estimates across diverse land uses. DyLEMa employs a coupled Random Forest model with a novel dynamic recalibrating strategy to improve pre-optimized seasonal hyperparameters following satellite acquisition alongside land cover classes. An analysis of feature importance indicated the significant variability in ET processes across different land cover classes and seasons. Hence, DyLEMa was applied to 20 years of daily 30x30 m pixel resolution Landsat-derived ET data in Illinois to address spatial and temporal discontinuities due to cloud contamination and sensor failures. DyLEMa performance was evaluated on Eddy Covariance measurements to find out that DyLEMa predictions reduced the average PBIAS error from + 31 % to −7% compared to existing US Geological Survey ET products. Spatially, DyLEMa underscores the value of a land cover-aware approach in ET estimation under varied cloud cover rates and their ability to preserve landscape features. However, the performance of DyLEMa was affected by the quality of land cover classification, suggesting the need for a refined region-specific land cover classification. DyLEMa's flexibility and performance suggest its applicability to other regions and satellite datasets, offering a promising reduction in uncertainty of ET estimates with impacts on environmental and water resources assessments on regional scales. [ABSTRACT FROM AUTHOR]

In this paper, researchers propose a classification method for any institution's campus placement possibility using Placement Data Full Class for campus recruitment dataset. Researchers attempt to study the supervised learning classification algorithms such Logistic Regression, Support Vector Classifier (SVC), K-Nearest Neighbors (KNN), Gaussian Naive Bayes, Decision Tree, Random Forest, Gradient Boosting, and Linear Discriminant Analysis (LDA). Hyperparameter optimization also used to optimize the supervised algorithms for better results. Experimental results have found that by using hyperparameter tuning in Linear Discriminant Analysis (LDA), it can increase the accuracy performance results, and also given a better result compared to other algorithms. [ABSTRACT FROM AUTHOR]

Technological advances are accelerating the dissemination of information. Today, millions of devices and their users are connected to the Internet, allowing businesses to interact with consumers regardless of geography. People all over the world send and receive emails every day. Email is an effective, simple, fast, and cheap way to communicate. It can be divided into two types of emails: spam and ham. More than half of the letters received by the user – spam. To use Email efficiently without the threat of losing personal information, you should develop a spam filtering system. The aim of this work is to reduce the amount of spam using a classifier to detect it. The most accurate spam classification can be achieved using machine learning methods. A natural language processing approach was chosen to analyze the text of an email in order to detect spam. For comparison, the following machine learning algorithms were selected: Naive Bayes, K-Nearest Neighbors, SVM, Logistic regression, Decision tree, Random forest. Training took place on a ready-made dataset. Logistic regression and NB give the highest level of accuracy – up to 99%. The results can be used to create a more intelligent spam detection classifier by combining algorithms or filtering methods. [ABSTRACT FROM AUTHOR]

• The performance of an evaporative condenser was studied via data-driven models. • A large experimental dataset (339 tests) compared to previous studies was employed. • The robustness of SVR, DT, and RF models compared to MLPANN was investigated. • Feature importance analysis was performed by the random forest model. • The effect of dataset size on the predictive performance of the models was studied. Evaporative condensers are regarded as highly-efficient and eco-friendly heat exchangers in refrigeration systems. Data-driven methods can play a key role in performance prediction of evaporative condensers, conducted without the complexity of theoretical analysis. In this study, four machine learning models including multi-layer perceptron artificial neural network (ANNMLP), support vector regression (SVR), decision tree (DT), and random forest (RF) models have been employed to predict heat transfer rate and overall heat transfer coefficient of a small-scale evaporative condenser functioning under a wide range of working conditions. A set of experimental tests were conducted, where inlet air dry/wet-bulb temperatures, spraying water and condenser saturation temperatures, refrigerant, and air flow rates were considered as main influencing parameters. The results show that the ANNMLP followed by SVR, and RF models possess the best generalization capability. Further, the dataset size analysis indicates that SVR is the best model to predict heat transfer rate for small dataset sizes. Additionally, feature importance analysis by the RF model reveals that refrigerant flow rate is the most influencing parameter. [ABSTRACT FROM AUTHOR]

[Display omitted] • Introduction of machine learning for modeling Sacran/CNF solution viscosity using six inputs for the first time. • High coefficient of determination (R2) of 0.99 achieved, demonstrating the effectiveness of the models. • Emphasis on key model features: shear rate or speed rate and the ratio of CNF content in solutions. • Notable importance given to shear rate, accounting for 60 % of model significance, and 31 % from CNF content ratio. • Machine learning algorithms showcased their value in analyzing viscosity data. This study explores the predictive capabilities of machine learning algorithms for Sacran/CNF solution viscosity. Leveraging data on initial CNF amount, shear rate, shear stress, speed, torque, and temperature, decision tree (DT) and random forest (RF) models were employed. Both models demonstrated strong alignment with experimental results, achieving a coefficient of determination (R2) of 0.99. The main features of those models established that 60 % of the importance is attributed to speed or shear rate, while 31 % is attributed to the ratio of CNF content in the solutions. This research underscores the potential of machine learning in effectively analyzing viscosity datasets. [ABSTRACT FROM AUTHOR]

The transmission dynamics of COVID-19 depend largely on the structure of the underlying contact network and individuals susceptibility. The latter factor is a primary concern of Covid-19 since recent studies reported conflicting conclusions regarding the characteristics of susceptible individuals. A susceptibility classification model of individuals is of great importance to governments and decision makers, as it facilitates imposing personalized protective strategies, which can save lives and minimize social and economic consequences. Machine learning models have been successfully used to predict individual susceptibility to different diseases such as cancer and visceral fat associated diseases. Existing contributions that predict susceptibility of COVID-19 are limited and use a large number of features including clinical data and medical history, which may be hard to obtain or unavailable. In this work, we investigate the use of machine learning models, namely multilayer perceptron (MLP), support vector machine (SVM), decision tree (DT), and random forest (RF) to predict susceptibility of individuals based on demographic data, including age, gender, nationality, and location. We use a dataset of test records of all individuals who were admitted to take nasopharyngeal swabs for COVID-19 in Saudi Arabia between March 2, 2020 until April 25, 2020. Our experiments show promising results with prediction accuracies of 85.6%, 85.3%, 77.2%, and 74.2% using DT, RF, MLP, and SVM, respectively. [ABSTRACT FROM AUTHOR]

Financial decision supporting is a very important and complex problem. The aim of this paper is to develop the Supervised Learning method based on the random forest algorithm for decision support on stock exchange. Contemporarily, machine learning methods, including decision trees, are often used. Many research works and practical implementation projects focus on supporting decisions on stock markets. However, most of them are related to developed markets in the USA or Western Europe, or Asian stock markets. There is a lack of research related, for example, to the Warsaw Stock Exchange (WSE). The findings concern determining which of the most popular technical analysis indicators have the greatest predictive power for a successful transaction with the feature importance method. [ABSTRACT FROM AUTHOR]

• Dual Deep CNN is used for extracting valuable features from input data. • Adapted to perform feature extraction in predicting disease. • XGBoost algorithm comprising of certain innate advantages like high convergence and modest computations. A disease is said to be chronic when a disease occurs in an individual and their health conditions due to the disease last one year/more. Predicting chronic diseases have become crucial to save individual's life and enhance their well-being. Though traditional clinical procedures are considered to perform this, it seems to be a time consuming process. Concurrently, with the progress of data mining algorithms, researchers have attempted to use different algorithms for such prediction. Nevertheless, they have been ineffective in feature extraction that negatively affected the prediction rate. To combat issues with regard to low accuracy rate, the present research intends to perform prediction of four common chronic diseases (breast cancer, heart disease, diabetes and kidney disease) affecting people worldwide. To accomplish this, the research proposes dual Deep CNN (Deep Convolutional Neural Network) for feature extraction. In this case, optimal, maximum and minimum hidden layers are used for extracting relevant features. Further, ME-RF (Modified Extreme-Random Forest) is used for classification. In this process, the research considers XGBoost algorithm comprising of certain innate advantages like high convergence and modest computations. However, when the predictability of this model is poor, it works in an ideal manner with numerous leaves in DT (Decision Tree). Simultaneously, RF comprise of several trees with equal weighted leaves by which, maximum precision and accuracy could be attained flexibly with the prevailing data. Considering this, the trees are built with RF and the research introduces this process as ME-RF. Classification performance is evaluated individually on four different considered datasets under the implementation of minimum and maximum Deep CNN network and also with the use of combined dual deep CNN networks. The overall analytical outcomes confirms the effectiveness of the proposed system. [ABSTRACT FROM AUTHOR]

Calcium (Ca2+) and magnesium (Mg2+) are the major scaling ions of reverse osmosis concentrate in zero-liquid discharge systems, causing performance decline. In this study, we predicted the removal of Ca2+ and Mg2+ from simulated reverse osmosis concentrate by functional polyketones (FPKs). Four amines, including 1,2-diaminopropane (DAP), 1-(2-aminoethyl) piperazine (AEP), 1-(3-aminopropyl) imidazole (API), and butyl amine (BA) used to synthesize FPKs. The effects of various factors such as the amount of adsorbent, feed water concentration, and pH were investigated for process optimization. In this study, ensemble learner artificial intelligence models, decision tree (DT), extreme gradient boost (XGB), and random forest (RF) were used to predict Ca2+ and Mg2+ removal by the FPKs. Datasets were collected experimentally using FPKs to remove Ca2+ and Mg2+ from the simulated reverse osmosis concentrate. The predictions were made by XGB, DT, and RF models for the first chosen amine for Ca2+ and then for Mg2+, subsequently, this process was repeated with each amine. The developed DT, RF, and XGB models demonstrated higher coefficients of determination for predicting Mg2+ removal by AEP and DAP (R2 = 0.841–0.935) than by API and BA (R2 = 0.774–0.801) except in the RF and XGB model results (R2 = 0.801–0.846). Overall, the XGB model displayed good results for both Ca2+ and Mg2+ removal but slight changes were observed in the AEP and BA predictions by DT and RF. Therefore, artificial intelligence models may be a viable alternative for further insight in predicting Ca2+ and Mg2+ removal by FPKs from simulated reverse osmosis concentrate. [Display omitted] • Observed ≥71% removal of Ca2+ and ≥66% of Mg2+ by the FPKs. • Studied predictive modeling of Ca2+ and Mg2+ removal from ROC by FPKs. • Tested the feasibility of ensemble AI techniques (DT, RF, and XGB). • Conducted a comparative analysis between DT, RF and XGB and a recent study. • Proposed XGB model predicted reasonably better results for Ca2+ and Mg2+ removal from ROC by FPKs. [ABSTRACT FROM AUTHOR]

In machine learning, some models can make uncertain and imprecise predictions, they are called evidential models. These models may also be able to handle imperfect labeling and take into account labels that are richer than the commonly used hard labels, containing uncertainty and imprecision. This paper proposes an Evidential Decision Tree, and an Evidential Random Forest. These two models use a distance and a degree of inclusion to allow the model to group observations whose response elements are included in each other into a single node. Experimental results showed better performance for the presented methods compared to other evidential models and to recent Cautious Random Forests when the data is noisy. The models also offer a better robustness to the overfitting effect when using datasets that are effectively uncertainly and imprecisely labeled by the contributors. The proposed models are also able to predict rich labels, an information that can be used in other approaches, such as active learning. • Source uncertainty and imprecision can be modeled in richer labels. • Richer labels improve classification performance. • Evidential decision tree and evidential random forest are robust to overfitting. [ABSTRACT FROM AUTHOR]

Optimization of SO 2 emission and the process cost in catalytic hydrodesulfurization (HDS) is of great importance for the petroleum industry. Given that the process of HDS is complicated, machine learning-based models are suitable for the purpose of process optimization in which the cost and separation efficiency can be optimized efficiently. In this investigation, we are working with a data collection on the HDS process to model via machine learning models. Pressure, temperature, initial sulfur content, and catalyst dose constitute the inputs for the models. Outputs include sulfur concentration (ppm), emission of gas (%), and HDS process cost ($). To model the process, for the first time, four tree-based ensemble methods are developed including Gradient Boosting, Extreme gradient boosting, Random Forest, and Extra Trees to optimize the HDS process. The models tuned on the available dataset and then the best ones selected for each output For sulfur concentration the extra tree model is the most accurate and for other outputs extreme gradient boosting has the best performance. For the models, the R2 scores for outputs are 0.983, 0.982, and 0.995, respectively. [ABSTRACT FROM AUTHOR]

The Notification Oriented Paradigm (NOP) emerges as an alternative to develop and execute applications. The NOP brings a new inference concept based on precise notifying collaborative minimal entities. This inference implicitly allows achieving decoupled solutions, thereby enabling parallelism at a granularity level as fine-grained as possible in the envisaged computational platform. Previous research has proposed a digital circuit solution based on the NOP model, which is called NOP to Digital Hardware (DH), as a sort of High-Level Synthesis (HLS) prototype tool. The results with NOP-DH were encouraging indeed. However, the previous NOP-DH works lack benchmarks that exploit well-known algorithms against known HLS tools, such as the Vivado HLS tool, which is one of the suitable commercial HLS solutions. This work proposes evaluating the NOP-DH applied to develop the well-known Random Forest algorithm. The Random Forest is a popular Machine Learning algorithm used in several classification and regression applications. Due to the high number of logic-causal evaluations in the Random Forest algorithm and the possibility of running them in parallel, it is suitable for envisaged benchmark purpose. Experiments were performed to compare NOP-DH, and two Vivado HLS approaches (an ad hoc code and a hls4ml tool-based code) in terms of performance, amount of logic elements, maximum frequency, and the number of predictions per second. Those experiments demonstrated that NOP-DH circuits achieve better results concerning the number of logical elements and prediction rates, with some scalability limitations as a drawback. On average, the NOP-DH uses 52.5% fewer resources, and the number of predictions per second is 4.7 times higher than Vivado HLS. Finally, our codes are made publicly available at https://nop.dainf.ct.utfpr.edu.br/nop-public/nop-dh-random-forest-algorithm. [ABSTRACT FROM AUTHOR]

The fluctuation of stock price is a major concern for stock investors. In order to provide investors with decision-making suggestions for buying or selling stocks, we use big data computing platform-Spark to perform binary classification prediction of stocks' rise and fall according to the price and related information in stock trading day, and compare four classification algorithms: naïve bayes, random forest, decision tree, and logistic regression from four metrics: run time, classification accuracy, AUC(Area under the receiver operating characteristic curve) and PR(Area under the Precision-Recall curve). Experimental results show that: (1) the run time is negatively correlated with the scale of the cluster. When the number of node in the cluster is increased by 3, the run time is reduced to about two-fifths; 2. Naïve Bayes has the lowest accuracy 0.7436, and the accuracy of random forests, decision trees, and logistic regression are 0.8297, 0.8318, 0.8085 respectively; (3) both the random forest and decision tree algorithm have a higher AUC and PR value, the values of the random forest are 0.6891,0.7497 and the value of the decision tree are 0.7165,0.7485. The evaluation results show that the random forest and decision tree algorithms are more suitable for stock classification prediction, because they pay more attention to the features with large information gain values and can reduce the error of the classification. [ABSTRACT FROM AUTHOR]

• Longitudinal prognosis models are not available for acute liver failure patients. • We develop prediction models using novel BiMM tree and BiMM forest methodology. • BiMM forest has slightly higher accuracy, but BiMM tree is simpler to use at the bedside. Assessing prognosis for acetaminophen-induced acute liver failure (APAP-ALF) patients during the first week of hospitalization often presents significant challenges. Current models such as the King's College Criteria (KCC) and the Acute Liver Failure Study Group (ALFSG) Prognostic Index are developed to predict outcome using only a single time point on hospital admission. Models using longitudinal data are not currently available for APAP-ALF patients. We aim to develop and compare performance of prediction models for outcomes during the first week of hospitalization for APAP-ALF patients. Models are developed for the ALFSG registry data to predict longitudinal outcomes for 1042 APAP-ALF patients enrolled 01/1998–02/2016. The primary outcome is defined as daily low versus high coma grade. Accuracy in prediction of outcome (AC), sensitivity (SN), specificity (SP) and area under the receiver operating curve (AUC) are compared between the following models: classification and regression tree, random forest, frequentist generalized linear mixed model (GLMM), Bayesian GLMM, BiMM tree, and BiMM forest using original and imputed datasets. BiMM tree offers predictive (test set) 63% AC, 72% SP and 53% SN for the original dataset, whereas BiMM forest offers predictive (test set) 69% AC, 63% SP and 74% SN for the imputed dataset. BiMM tree has the highest AUC for the original testing dataset (0.697), whereas BiMM forest and standard random forest have the highest AUC for the imputed testing dataset (0.749). The three most important predictors of daily outcome for the BiMM tree are pressor use, bilirubin and creatinine. The BiMM forest model identifies lactate, ammonia and ALT as the three most important predictors of outcome. BiMM tree offers a prognostic tool for APAP-ALF patients, which has good accuracy and simple interpretation of predictors which are consistent with clinical observations. BiMM tree and forest models are developed using the first week of in-patient data and are appropriate for predicting outcome over time. While the BiMM forest has slightly higher predictive AC, the BiMM tree model is simpler to use at the bedside. [ABSTRACT FROM AUTHOR]

Background: Some patients have features that indicate possible difficulty with direct laryngoscopy for tracheal intubation. Prediction of the likely outcome and selection of patients for an enhanced management algorithm would reduce the possible harm from failed intubation attempts.Methods: Adult elective patients were assessed for seven features associated with difficult direct laryngoscopy, ranked in difficulty from 0 to 3. For a patient with at least one Class 3 feature, or two or more features of class 1 or higher, the enhanced management used a channelled videolaryngoscope Airtraq™ instead of a Macintosh laryngoscope. A long flexible angulated stylet and a flexible fibrescope would be used as the second and third steps. For patients with lesser difficulty scores, a Macintosh laryngoscope was used. Outcomes of enhanced management were analysed. Logistic regression and Random Forest algorithm, using the ranks of the predictive features, were used to predict difficulty during enhanced management.Results: We prospectively studied 16 695 patients. We selected 1501 (9%) for enhanced management, and tracheal intubation was successful in all of them. Of these, 73% were intubated in less than 30 s, and only 4.5% required more than 4 min for intubation. Progression to the second and third steps of enhanced management was predicted by restriction of mouth opening and reduced cervical spine mobility.Conclusions: An enhanced management algorithm allowed successful tracheal intubation of all patients with anticipated difficult laryngoscopy. The need to combine the use of a stylet and a fibrescope with the Airtraq™ could be predicted with a high degree of certainty. [ABSTRACT FROM AUTHOR]

Label ranking aims to learn a mapping from instances to rankings over a finite number of predefined labels. Random forest is a powerful and one of the most successful general-purpose machine learning algorithms of modern times. In this paper, we present a powerful random forest label ranking method which uses random decision trees to retrieve nearest neighbors. We have developed a novel two-step rank aggregation strategy to effectively aggregate neighboring rankings discovered by the random forest into a final predicted ranking. Compared with existing methods, the new random forest method has many advantages including its intrinsically scalable tree data structure, highly parallel-able computational architecture and much superior performance. We present extensive experimental results to demonstrate that our new method achieves the highly competitive performance compared with state-of-the-art methods for datasets with complete ranking and datasets with only partial ranking information. [ABSTRACT FROM AUTHOR]

Highlights • The evaluated models are able to predict the potential occurrence of an invasive species. • The Hasse diagram technique proves to be useful to compare machine learning methods. • A specific combination of machine learning methods is superior to a single model. • The random combination of models can be inferior to the output of a single model. Abstract We tested four machine learning methods for their performance in the classification of mosquito species occurrence related to weather variables: support vector machine, random forest, logistic regression and decision tree. The objective was to find a method which showed the most accurate model for the prediction of the potential geographical distribution of Aedes japonicus japonicus , an invasive mosquito species in Germany. The evaluation of the model trainings was conducted using derivations of a confusion matrix. Furthermore, we introduced two quality indices, 'selectivity' and 'exactness', for the evaluation of the spatial simulation, visualised through the Hasse diagram technique. From the evaluation results we can conclude that a specific combination of two to three models performs better in predicting the potential distribution of the mosquito species than a single model or the random combination of models. [ABSTRACT FROM AUTHOR]

Highlights • UTI is a common disease and its diagnosis is not easy with routine examination procedures. • It is aimed at developing a model that diagnoses cystitis and nonspecific urethritis diseases with similar symptoms from urinary tract infections. • Comparison of 3 different artificial intelligence methods was presented. • This study was able to diagnose UTI with 98.3% accuracy using just two symptoms and one laboratory finding. Abstract Background and Objective Urinary tract infection (UTI) is a common disease affecting the vast majority of people. UTI involves a simple infection caused by urinary tract inflammation as well as a complicated infection that may be caused by an inflammation of other urinary tract organs. Since all of these infections have similar symptoms, it is difficult to identify the cause of primary infection. Therefore, it is not easy to diagnose a UTI with routine examination procedures. Invasive methods that require surgery could be necessary. This study aims to develop an artificial intelligence model to support the diagnosis of UTI with complex symptoms. Methods Firstly, routine examination data and definitive diagnosis results for 59 UTI patients gathered and composed as a UTI dataset. Three classification models namely; decision tree (DT), support vector machine (SVM), random forest (RF) and artificial neural network (ANN), which are widely used in medical diagnosis systems, were created to model the definitive diagnosis results using the composed UTI dataset. Accuracy, specificity and sensitivity statistical measurements were used to determine the performance of created models. Results DT, SVM, RF and ANN models have 93.22%, 96.61%, 96.61%, 98.30% accuracy, 95.55%, 97.77%, 95.55%, 97.77% sensitivity and 85.71%, 92.85%, 100%, 100% specificy results, respectively. Conclusions ANN has the highest accuracy result of 98.3% for UTI diagnosis within the proposed models. Although several symptoms, laboratory findings, and ultrasound results are needed for clinical UTI diagnosis, this ANN model only needs pollacuria, suprapubic pain symptoms and erythrocyte to get the same diagnosis with such accuracy. This proposed model is a successful medical decision support system for UTI with complex symptoms. Usage of this artificial intelligence method has its advantages of lower diagnosis cost, lower diagnosis time and there is no need for invasive methods. [ABSTRACT FROM AUTHOR]

Random Forest is an ensemble classification approach. It aims to design a discrete finite group of decision trees constructed based on bootstrap samples and random attribute selection. Random Forests have strong generalization capacities due to the variance in the training and attribute couple subsets used for constructing different decision trees in the forest. However, to construct a robust and effective random forest, two main issues need to be taken into account namely: (1) increasing the accuracy and diversity of decision trees; (2) decreasing the number of decision trees. In this paper, a genetic algorithm-based approach to tackle the aforementioned challenges related to random forest construction is proposed. Three objectives are taken into consideration. First, strengthening the classification accuracy of individual decision trees as well as that of the forest. Second, making use of diversity measures among the decision trees to improve the generalization of the constructed model. Third, minimizing the number of trees in the forest and finding an optimal subset of the random forest. An experimental evaluation on several datasets from the UCI Machine Learning Repository is conducted. The obtained results show that the proposed approach outperforms state-of-the-art classical as well as evolutionary random forest construction methods. Finally, the proposed approach is used to build a reliable random forest model for detecting Botnet traffic in Internet of Things environment. • We propose a genetic algorithm to construct an improved random forest. • We design a theoretic solution allowing a smart convergence to the best solutions. • We propose a Powerset system to optimize the encoding of the forests in chromosomes. • Our objective function optimizes the accuracy and diversity of the trees of models. • Application to the detection of Botnet traffic in Internet of Things networks. [ABSTRACT FROM AUTHOR]

The area-to-mass ratio (AMR) of a resident space object (RSO) is an important parameter for improved space situation awareness capability due to its effect on the non-conservative forces including the atmosphere drag force and the solar radiation pressure force. However, information about AMR is often not provided in most space catalogs. The present paper investigates recovering the AMR information from the consistency error, which refers to the difference between the orbit predicted from an earlier estimate and the orbit estimated at the current epoch. A data mining technique, particularly the random forest (RF) method, is used to discover the relationship between the consistency error and the AMR. Using a simulation-based space catalog environment as the testbed, this paper demonstrates that the classification RF model can determine the RSO's category AMR and the regression RF model can generate continuous AMR values, both with good accuracies. Furthermore, the paper reveals that by recording additional information besides the consistency error, the RF model can estimate the AMR with even higher accuracy. [ABSTRACT FROM AUTHOR]

In this paper, we propose a new decision forest algorithm that builds a set of highly accurate decision trees by exploiting the strength of all non-class attributes available in a data set, unlike some existing algorithms that use a subset of the non-class attributes. At the same time to promote strong diversity, the proposed algorithm imposes penalties (disadvantageous weights) to those attributes that participated in the latest tree in order to generate the subsequent trees. Besides, some other weight-related concerns are taken into account so that the trees generated by the proposed algorithm remain individually accurate and retain strong diversity. In order to show the worthiness of the proposed algorithm, we carry out experiments on 20 well known data sets that are publicly available from the UCI Machine Learning Repository. The experimental results indicate that the proposed algorithm is effective in generating highly accurate and more balanced decision forests compared to other prominent decision forest algorithms. Accordingly, the proposed algorithm is expected to be very effective in the domain of expert and intelligent systems. [ABSTRACT FROM AUTHOR]

For the purpose of carbon capture, utilization, and storage, carbon dioxide (CO 2) injection in coal formations can enhance methane recovery and mitigate climate change. However, measuring CO 2 adsorption isotherms using experimental or mathematical models can be time-consuming, expensive, and inaccurate. Thus, this study presents a machine-learning framework that predicts CO 2 adsorption in coal formations based on various coal properties and testing conditions. Machine-learning (ML) framework was applied using a dataset of 1,064 points collected for different coal samples at different operating conditions to predict the CO 2 adsorption in coal surface. The ML techniques include decision tree regression (DT), random forests (RF), gradient boost regression (GBR), K-nearest neighbor (KNN), artificial neural network (ANN), function network (FN), and adaptive neuro-fuzzy inference system (ANFIS). The applied framework determines CO 2 adsorption as a function of coal's physical and chemical properties (moisture, ash, volatile matter, and fixed carbon content), the vitrinite reflectance of the coal samples, and testing conditions (pressure and temperature). Classical statical tools such as R2, root mean square error (RMSE), and average absolute percentage error (AAPE) were used to evaluate the model's performance analysis. The results demonstrated the ability to determine CO 2 adsorption for varying coal types and at different temperature and pressure conditions. The statistical measures suggested that RF, GBR, and KNN are very reliable ML models, with RF being the best. At low operating pressure (P < 4 MPa), CO 2 adsorption is impacted by any pressure changes, while it is stabilized at high-pressure values and becomes more dependent on the rock properties at high operating pressure (P > 4 MPa). The introduced ML framework offers a technique to evaluate the capability of different algorithms and accurately estimate CO 2 adsorption without the requirement of additional experimental measurements or complicated mathematical techniques. [ABSTRACT FROM AUTHOR]

The Decision Tree algorithm is one of the first machine learning algorithms developed. It is used both as a standalone model and as an ensemble of many cooperating trees like Random Forest, AdaBoost, Gradient Boosted Trees, or XGBoost. In this work, a new version of the Decision Tree was developed for classifying real-world signals using Gaussian distribution functions and a fuzzy decision process. The research was carried out on Power Quality Classification Dataset hosted on the platform kaggle.com for accessing the algorithm's classification. The proposed algorithm modification produces a sparse tree of multidimensional Gaussian kernels performing fuzzy, proximity-based division of solution space instead of a typical Decision Tree performing definite space restrictions. The machine learning model can cluster samples based on the common pattern and evaluate the input's similarity in a fuzzified fashion. The studies were conducted for the prediction of 6 classes of current. The proposed modification achieved the best accuracy and F1 score compared to the default Decision Tree and other machine learning algorithms. [ABSTRACT FROM AUTHOR]

Annual cumulative soil nitrous oxide (N 2 O) emissions calculated from discrete chamber-based flux measurements have unknown uncertainty. We used outputs from simulations obtained with an agroecosystem model to design sampling strategies that yield accurate cumulative N 2 O flux estimates with a known uncertainty level. Daily s oil N 2 O fluxes were simulated for Ames, IA (corn-soybean rotation), College Station, TX (corn-vetch rotation), Fort Collins, CO (irrigated corn), and Pullman, WA (winter wheat), representing diverse agro-ecoregions of the United States. Fertilization source, rate, and timing were site-specific. These simulated fluxes surrogated daily measurements in the analysis. We “sampled” the fluxes using a fixed interval (1–32 days) or a rule-based (decision tree-based) sampling method. Two types of decision trees were built: a high-input tree (HI) that included soil inorganic nitrogen (SIN) as a predictor variable, and a low-input tree (LI) that excluded SIN. Other predictor variables were identified with Random Forest. The decision trees were inverted to be used as rules for sampling a representative number of members from each terminal node. The uncertainty of the annual N 2 O flux estimation increased along with the fixed interval length. A 4- and 8-day fixed sampling interval was required at College Station and Ames, respectively, to yield ±20% accuracy in the flux estimate; a 12-day interval rendered the same accuracy at Fort Collins and Pullman. Both the HI and the LI rule-based methods provided the same accuracy as that of fixed interval method with up to a 60% reduction in sampling events, particularly at locations with greater temporal flux variability. For instance, at Ames, the HI rule-based and the fixed interval methods required 16 and 91 sampling events, respectively, to achieve the same absolute bias of 0.2 kg N ha −1 yr −1 in estimating cumulative N 2 O flux. These results suggest that using simulation models along with decision trees can reduce the cost and improve the accuracy of the estimations of cumulative N 2 O fluxes using the discrete chamber-based method. [ABSTRACT FROM AUTHOR]

Accurate estimates of air demand are critical when addressing the cavitation phenomenon in bottom outlet conduits of dams. In the present study, this accuracy was evaluated for air demand estimation models using gene expression programming (GEP), classification tree methods [including boosted regression trees (BRT) and random forest (RF) algorithms], and empirical relationships. Using k -fold cross validation and drawing on data from 6 dams, two different air-demand-related parameters were estimated (aerator air discharge demand and aeration coefficient) and the estimation accuracy was assessed using the Nash-Sutcliffe (NS) efficiency and other statistics. Outperforming the other models, the GEP model performed well in estimating aerator air demand (NS =0.674) and the aeration coefficient (NS =0.489). Moreover, on average, the GEP model showed 45% and 12% improvements in aeration coefficient estimation accuracy over empirical relationships and decision tree methods, respectively. Finally, in the function finding phase, the derived mathematical models were presented as non-linear equations to estimate the air demand of aerator and aeration coefficients using a non-linear relation derived from the GEP model. [ABSTRACT FROM AUTHOR]

Tree-based ensemble algorithms (T E A s) have made significant advances in recent years due to their simple algorithmic design. However, when the proportion of the 'most informative' features is low, the performance of conventional T E A s degrades significantly. The primary rationale for performance degradation is that traditional algorithmic design appears to be biased toward the least informative features, and the sub-space selection procedure contains uninformative features. This paper proposes a logically randomized forest (L R F) algorithm by incorporating two different enhancements into existing T E A s. The first enhancement is made to address the issue of biasness by performing feature-level engineering. The second enhancement is the approach by which individual feature sub-spaces are selected. To derive the first enhancement, we use the graph-theoretic principle of minimal vertex cover to construct a relevant assemblage of features. Following that, the permutation-based feature importance technique is employed to calculate the 'informativeness' of the relevant features in order to infuse logical randomness into the individual trees in the forest. For the second enhancement, the stratified sampling method is used to ensure that the most informative features are present in all newly created feature sub-spaces. Consequently, individual trees are generated using the Roulette wheel-based selection (R W S) algorithm. The proposed algorithm has been evaluated on two real-world genomic data sets, ten hybrid-synthetic classification data sets, and twenty multidisciplinary benchmark data sets with varying characteristics. The experimental findings demonstrate that the L R F outperforms the existing benchmark and cutting-edge T E A s. • Proposed a novel tree-based ensemble algorithm. • Appropriate for applications with sparsely populating most informative features. • Feature engineering by graph-theoretic principal and permutation-based algorithm. • Forest generation by stratified sampling and Roulette wheel selection algorithm. • Proposed algorithm outperformed the benchmark and state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Herein, we analyze the effects of critical materials, electrode preparation methods, and operational descriptors on the discharge capacity and cycle life of the Na-ion batteries. An extensive dataset is constructed from literature and analyzed using machine learning. It is found that alloy-based anodes have the highest average discharge capacity, while the carbon group exhibits higher cycle life performance; it is also deduced that the discharge capacity is improved when alloy-based anodes are coupled with metal oxide cathodes. Random forest models are reasonably good for providing rough predictions for discharge capacity and demonstrating the relative significance of descriptors; the root mean square error for training and testing are 75 mAh/g and 157 mAh/g, respectively, for the anode, and 13 mAh/g and 48 mAh/g, respectively, for the cathode studies. Anode and cathode types are the most influential descriptors for the model, as expected, while the synthesis conditions and crystal structure are also effective. Decision tree classification of cycle life (cycle number at which 80% of peak discharge capacity is retained) is also quite successful in leading heuristic rules for electrode preparation. Material synthesis conditions are highly influential for high cycle life for the anode, while solvent selection seems to be also significant for cathode studies. [Display omitted] • Performance of Na-ion batteries was studied using machine learning. • Dataset contained 1227 data points obtained from 335 published articles. • Random forest was used to develop predictive models for discharge capacity. • Decision tree was used to develop heuristics for cycle life. • Significant factors affecting high performance were also identified. [ABSTRACT FROM AUTHOR]