Search

Solutions based on tree-ensemble models represent a considerable alternative to real-world prediction problems, but these models are considered black box, thus hindering their applicability in problems of sensitive contexts (such as: health and safety). Explainable Artificial Intelligence (XAI) aims to develop techniques that generate explanations of black box models, since these models are normally not self-explanatory. Methods such as Ciu, Dalex, Eli5, Lofo, Shap and Skater emerged with the proposal to explain black box models through global rankings of feature relevance, which based on different methodologies, generate global explanations that indicate how the model's inputs explain its predictions. This research aims to present an innovative XAI method, called eXirt , capable of carrying out the process of explaining tree-ensemble models, based on Item Response Theory (IRT). In this context, 41 datasets, 4 tree-ensemble algorithms (Light Gradient Boosting, CatBoost, Random Forest , and Gradient Boosting), and 7 XAI methods (including eXirt) were used to generate explanations. In the first set of analyses, the 164 ranks of global feature relevance generated by eXirt were compared with 984 ranks of the other XAI methods present in the literature, being verified that the new method generated different explanations from other existing methods. In a second analysis, exclusive local and global explanations generated by eXirt were presented that help in understanding the model trust, since in this explanation it is possible to observe particularities of the model regarding difficulty (if the model had difficulty predicting the test dataset), discrimination (if the model understands the test dataset as discriminative) and guesswork (if the model got the test dataset right by chance). Thus, it was verified that eXirt is able to generate global explanations of tree-ensemble models and also local and global explanations of models through IRT , showing how this consolidated theory can be used in machine learning in order to obtain explainable and reliable models. [ABSTRACT FROM AUTHOR]

Parkinson's Disease (PD) is a prevalent neurodegenerative condition, ranking second in incidence only to Alzheimer's disease. Presently, the diagnosis of PD using sensors and computer vision can recognize characteristics like speech, respiration, and gait in patients. However, these methods typically require the use of complex hardware systems or extensive data analysis when collecting or processing motion data from various joints of the patients. In this research, an efficient diagnostic system for PD is introduced, which utilizes common smartphones to capture videos of 42 healthy volunteers and 30 Parkinson's patients walking under different conditions. The system utilizes human pose estimation technology to extract three-dimensional joint coordinates, simplifying the data collection process. Afterward, joint coordinates are transformed into two-dimensional images using color mapping techniques, thus converting the diagnosis into an image recognition challenge. In contrast to video processing, this approach reduces 6.51 million parameters (22%) and reduces testing time by 62.92%. Finally, by combining ResNet50 and random forest algorithms, decision accuracy is improved using two-dimensional images. Experimental evidence substantiates the system's capability to efficiently identify Parkinsonian gait. The recognition rate for the leg joint group is 88.89%, and it would increase to 91.67% when factoring in the four-limb joint group. The system provides the potential for preliminary at-home diagnosis, holding significant clinical application potential, and diagnostic value for other movement-related disorders. • Combines mobile phone video and human assessment technology. • Mapping of joint coordinates to 2D images. • Integration of ResNet50 and Random Forest algorithms. • Joint group testing. [ABSTRACT FROM AUTHOR]

Detecting bearing faults helps ensure the healthy operation of machinery and prevents serious accidents. However, fault diagnosis method based on deep learning relies on the correlation between the extracted vibration signal features. It does not consider the causal relationship between fault, noise, and vibration waveform changes, resulting in lower bearing fault diagnosis accuracy under realistic working conditions. This study proposes a graph neural network (GNN) method based on the Granger causality test for bearing fault detection called GCT-GNN to address this issue. The proposed method first performs feature transformation on the original signals to extract time-domain and frequency-domain features, forming a feature matrix. Then, spectral analysis is conducted on both faulty signals with noise and healthy signals to calculate the lag order between them. Subsequently, the Granger causality test is employed to quantify the impact of faults and noise on signal changes, and the quantified results are used to calculate weights, constructing an adjacency matrix. The completed adjacency matrix and feature matrix are input into the GNN for feature mapping, last classifying the bearing fault data. In this paper, five models, Deep residual shrinkage Network (DRSN), GNN, Support Vector Machines(SVM), Random Forests (RF), and Convolutional Neural Network(CNN) were selected, and comparative experiments were carried out on two public data sets. The results show that, compared with other models, GCT-GNN has better anti-noise ability and is better than other models in various fault diagnoses. • The effect of noise on vibration signal is quantified by Granger causality test. • The lag order was calculated by using spectrum analysis method. • The method of calculating weight by causality is proposed to construct causality graph. • Construct a graph neural network using causal graphs. [ABSTRACT FROM AUTHOR]

Truck drivers are essential to modern supply chains. However, the companies that employ them are having trouble keeping them. At many trucking firms, annual turnover rates approach 100 percent or higher. This paper develops a method for predicting individual truck driver turnover events before they happen by applying supervised machine learning classifiers to a new source of operational truck driver data. Our paper uses Electronic Logging Device (ELD) data, which comprise newly federally-mandated, time-stamped work logs collected from approximately 1200 American long-haul truck drivers over 3 years. We train three supervised machine learning classifiers (logistic regression, random forests, and support vector machines) on this data and achieve 60 to 70 percent prediction accuracy and 50 to 60 percent recall across two 5-fold cross-validated experiments. We observe that the quantity and consistency of week-day driving assignments explain most of our models' predictive power. We offer these results as both a new technical tool, as well as novel managerial insights for improving global truck driver retention problems for the benefit of supply chains worldwide. [Display omitted] • Newly available operational data from long haul truck drivers is analyzed. • Three machine learning classifiers are trained on the operational data. • Predictive models achieve 60–70 percent accuracy, 50–60 percent recall. • Models show potential to enable interventions to retain truck drivers. [ABSTRACT FROM AUTHOR]

Telemonitoring of Parkinson's disease has important implications for early diagnosis and treatment of patients. Most of the existing feature selection methods for remote prediction of PD severity are based on correlation and rarely consider causality, thus compromising the robustness of the model. Therefore, a causal game-based feature selection (CGFS) model is proposed for remote PD symptom severity assessment. Firstly, to address the challenge of small data size, the similar patient transfer strategy is designed to find data from source domain patients with conditions similar to those of the target patient. Secondly, the undirected equivalent greedy search method is employed to construct the causal graph between features and PD severity scores, and the robustness of the model is improved by selecting causal features. Then, to enhance the prediction accuracy, this paper utilizes the cooperative game approach Shapley value to evaluate the contribution of neighborhood nodes of the target value, and selects the features with causality and high contribution to form the final feature subset. Finally, the subset is input into the random forest to further enhance robustness and performance of the model. Experiments on Parkinson's telemonitoring dataset and the tapping dataset with different biomarkers show that the robustness of the feature subset selected by the CGFS model, and the prediction performance is better than advanced models compared. Therefore, the validity and universality of the CGFS method is demonstrated in remote PD severity prediction. [ABSTRACT FROM AUTHOR]

Traditional machine learning algorithms face challenges in assessing flood susceptibility reliably due to their low robustness and the inherent 'black-box' nature. This paper utilizes five hyperparameter optimization algoirthms (HPO), namely grid search (GS), random search (RS), gauss process (GP), tree-structured parzen estimator (TPE) and simulated annealing (SA), to tune the traditional random forest's (RF) hyperparameters to improve the robustness of flood hazard mapping (FHM) models at Ningxiang City Hunan Province, China. Additionally, SHapley Additive exPlanations (SHAP) method were used to interpret the decision-mechanisms of these flood hazard models. This study considers 19 pluvial flood influencing factors and 2064 flood locations to create a geospatial database. The performance of each hybrid model was evaluated by area under the receiver operating characteristic (ROC) curve (AUC) and several validation methods. The results demonstrate that the developed hybrid models demonstrated good performance, with RF-TPE achieving the highest AUC (0.9660), followed by RF-GP (0.9648), RF-SA (0.9624), RF-GS (0.9612), RF-RS (0.9600), and RF (0.9539). The RF-TPE model exhibits superior robustness than other models, and the FHM constructed using it is more reliable. HPO is an effective approach to improve the predictive accuracy and robustness of FHM models. When considering limited computational resources, Bayesian optimization (TPE) should be prioritized for optimizing FHM models, followed by metaheuristic algorithms and model-free algorithms. Moreover, the study revealed that distance from river, peak rainfall intensity, continuous rainfall, antecedent effective rainfall, and terrain relief, are the most significant for pluvial FHM modeling in this region. [ABSTRACT FROM AUTHOR]

The US Census Bureau has collected two rounds of experimental data from the Commodity Flow Survey, providing shipment-level characteristics of nationwide commodity movements, published in 2012 (i.e., Public Use Microdata) and in 2017 (i.e., Public Use File). With this information, data-driven methods have become increasingly valuable for understanding detailed patterns in freight logistics. In this study, we used the 2017 Commodity Flow Survey Public Use File data set to explore building a high-performance freight mode choice model, considering three main improvements: (1) constructing local models for each separate commodity/industry category; (2) extracting useful geographical features, particularly the derived distance of each freight mode between origin/destination zones; and (3) applying additional ensemble learning methods such as stacking or voting to combine results from local and unified models for improved performance. The proposed method achieved over 92% accuracy without incorporating external information, an over 19% increase compared to directly fitting Random Forests models over 10,000 samples. Furthermore, SHAP (Shapely Additive Explanations) values were computed to explain the outputs and major patterns obtained from the proposed model. The model framework could enhance the performance and interpretability of existing freight mode choice models. [ABSTRACT FROM AUTHOR]

This study presents a comprehensive and systematic review of the machine learning (ML) techniques used to predict employee turnover in the past decade. A total of 52 relevant peer-reviewed studies published between 2012 and April 2023 were selected. The results indicate that over 20 ML techniques have been used to predict employee turnover in various institutions. In addition, this comprehensive review demonstrates that most machine learning approaches used to predict employee turnover were based on supervised learning, with 96% of the articles (50 out of 52) in this category, random forest technique. Furthermore, the review reveals that the most critical factors for predicting employee turnover have been salary and overtime. This study makes a valuable contribution to the field by offering a systematic analysis of the ML algorithms used for predicting employee turnover, in addition to providing an overview of the most significant works in this field produced in the past decade. This study offers important reference regarding the essential ML approaches used in employee turnover prediction and provides future directions for researchers and industries. [ABSTRACT FROM AUTHOR]

The proactive prediction and systematic classification of students' academic performance empower educational administrators with the invaluable capability to not only pinpoint potential challenges but also to craft targeted strategies and interventions that are tailored to the unique needs of students. This multifaceted approach enables educational institutions to proactively address issues within the education system, fostering a more equitable and effective learning environment for all, while simultaneously fostering a culture of continuous improvement and accountability in the pursuit of educational excellence. Hence, the current investigation aims to classify and predict the students' performance by examining and comparing the machine learning and artificial neural network assessments. Five methods of the Random Forest Classifier, the Decision Tree Classifier, the K Neighbors Classifier, the MLP Classifier, and the XG-Boost Classifier are used. These methods' performances are compared through the accuracy, precision, recall, and F1-score indicators. This comparison is applied to the base data and balanced data, which is carried out by the SVM-SMOTE technique. Finally, five metaheuristic algorithms are applied to the selected method to evaluate the performance indicators of the hybrid models. The results indicate that applying the SVM-SMOTE technique improves the methods' performance, in which the XG-Boost represented the best performance. As a result, the metaheuristic algorithms are applied to the XG-Boost, yielding to 9.33%, 8.44%, 9.33%, and 9.27% enhancement of the Accuracy, Precision, Recall and F1-Score values. Subsequently, the Enhanced Artificial Ecosystem-Based Optimization + XG-Boost hybrid method provides the accuracy and F1-score values of 0.9417 and 0.9413. These results underscore the potential of combining machine learning techniques with metaheuristic algorithms to enhance the accuracy and effectiveness of predicting and classifying students' performance, thus providing valuable insights for educational administrators to address issues and improve the education system. [ABSTRACT FROM AUTHOR]

As global warming continues to cause environmental concerns, finding alternative refrigerants is a current research focus in the refrigeration industry. R290 has desirable physical properties as a refrigerant, but its flammable and explosive nature restricts its usage. In this study, the authors designed a real refrigerant leakage experiment for an R290 room air conditioner based on the latest international standard, IEC 60335-2-40-2022. The experiment system was in a professional explosion-proof enthalpy difference chamber refrigeration. Leakage experiments were conducted under minimal, rated, and maximal operating refrigeration conditions, with a maximum leakage of 30% of the rated charge. To identify refrigerant leakage faults, the authors developed a feature parameter selection method based on the random forest algorithm variable importance measures coupled with the Pearson correlation coefficient. They then used a genetic algorithm to optimize the parameters in the random forest algorithm and obtained an optimal parameter set to establish a random forest fault identification model. With 35 fault identification rules extracted, the accuracy of the test set was 91.5%. To confirm the suitability of the fault identification rules for the entire operation of the R290 room air conditioner, 23 non-modeled experimental data sets were used, resulting in an accuracy rate of 91.3%. As a result, this study concludes that the fault identification rules proposed in this paper can be applied to the real operating conditions of the R290 room air conditioner. [ABSTRACT FROM AUTHOR]

The paper presents a machine learning based Smart Intrusion and Fault Identification (SIFI) method to identify the cyber-physical abnormalities in an inverter-based cyber-physical microgrid (CPM). The SIFI method utilizes an ensemble classifier (EC) to assimilate the decisions from three distinct classifiers (C4.5 decision tree, random forest (RF), and forest by penalizing attributes (FPA)). The proposed method employs a voting mechanism to classify and localize abnormal events for improved accuracy. The paper investigates the effects of cyberattacks (Denial-of-Service (DoS) and Malicious Data Injection (MDI) attacks) and physical abnormalities caused by line faults in microgrids. To train the classifiers, SIFI employs dataset with statistical attributes extracted from measurements. The renewable alternative power systems simulation (RAPSim) software tool is utilized to model the proposed system. The effectiveness of the presented model is assessed with respect to mean value error (MVE). The efficiency of the classifier is demonstrated by comparing its performance with other classifiers in terms of the MVE. For physical abnormalities identification and localization, the approach provides enhanced outputs with the MVE of 0.157 % and 0.162 % correspondingly. For identifying MDI anomalies, the model provides better results with a lower error rate. For DoS anomaly detection, the model provides better classification performance with 0.136 % error. The extensive empirical analysis proves that the anticipated ensemble classifier-based SIFI model yields the minimum MVE for identifying physical faults and cyberattacks. Furthermore, the effectiveness of the proposed method is identified by considering a case study in which the SIFI method outperforms well for the MDI and DoS cyber-attacks in the CPM. [ABSTRACT FROM AUTHOR]

The aim of ordinal classification is to predict the ordered labels of the output from a set of observed inputs. Interval-valued data refers to data in the form of intervals. For the first time, interval-valued data and interval-valued functional data are considered as inputs in an ordinal classification problem. Six ordinal classifiers for interval data and interval-valued functional data are proposed. Three of them are parametric, one of them is based on ordinal binary decompositions and the other two are based on ordered logistic regression. The other three methods are based on the use of distances between interval data and kernels on interval data. One of the methods uses the weighted k -nearest-neighbor technique for ordinal classification. Another method considers kernel principal component analysis plus an ordinal classifier. And the sixth method, which is the method that performs best, uses a kernel-induced ordinal random forest. They are compared with naïve approaches in an extensive experimental study with synthetic and original real data sets, about human global development, and weather data. The results show that considering ordering and interval-valued information improves the accuracy. The source code and data sets are available at https://github.com/aleixalcacer/OCFIVD. • Six ordinal classifiers for interval data-valued data are proposed. • Ordinal methods for interval-valued functional data are also proposed. • Considering ordering and interval-valued information improves the accuracy. • A kernel-induced ordinal random forest performs best. [ABSTRACT FROM AUTHOR]

Healthcare fraud detection algorithms are mostly based on applying machine learning methods to payer-claims data transactions to detect fraudulent activities. However, claim transactions are often analyzed in isolation, disregarding the interconnections that naturally exist in the claims generated from a single unit such as the patient, payer, or provider. Fraud detection frameworks have untapped potential to aid downstream processes such as investigations and audits through transparent and interpretable machine learning models. In this study, we show how fraud detection frameworks can benefit from use of a graphical network model called Bayesian Belief Network (BBN) which exploits the relational structure of attributes in a transaction and has better interpretability properties than many competing machine learning methods. BBN model performance is found to be comparable to commonly utilized baseline models such as logistic regression and random forest after validation with a real-life insurance claims dataset. In particular, the F1 score of fraud class for the best performing BBN was 0.15. The proposed BBN model is also considered to be interpretable to auditors and investigators based on common sense qualitative evaluation. [ABSTRACT FROM AUTHOR]

• Formulation of a perishable inventory model in multi-echelon distribution systems. • Integration of surrogate models significantly improves the performance of particle swarm optimization and differential evolution algorithms. • Development of a co-evolutionary strategy for efficient data point sampling of surrogate models in differential evolution algorithm. • The proposed algorithm outperformed 12 different metaheuristics. • Introduction of a new performance indicator to analyze the mechanism of surrogate-assisted evolutionary computation techniques. Simulation is widely used for analyzing supply chains with complex structures and stochastic nature. However, optimizing supply chain simulation models is usually computationally expensive. This study proposes a surrogate-assisted evolutionary optimization approach to optimize the inventory policies in multi-echelon distribution systems for perishable items under a limited number of evaluations. The random forest algorithm is used to build the surrogate model for a faster estimation of the performance of the inventory policies. A co-evolutionary differential evolution algorithm is proposed to simultaneously evolve the population through multiple searching strategies. The generated solutions are estimated by the low-cost surrogate model to select the promising solutions, which will be evaluated by the inventory model. Moreover, this study also integrates the surrogate model into two classic metaheuristic algorithms, particle swarm optimization and differential evolution. Also, a new performance indicator is proposed to examine the efficiency of the surrogate model in evolutionary computation. Two case studies are used to investigate the performance of the proposed algorithms. The experimental results show that both particle swarm optimization and differential evolution exhibit performance improvements exceeding 55% by using surrogate models under the limited number of function evaluations. Furthermore, the surrogate model reduces computational time for both algorithms by over 34% to achieve equivalent objective values. Finally, the proposed co-evolutionary differential evolution algorithm is compared with 12 algorithms, and the results show that the proposed algorithm consistently outperforms them. These findings confirm the usefulness of the surrogate model in evolutionary algorithms and the effectiveness of the proposed co-evolutionary strategy for solving perishable inventory problems. [ABSTRACT FROM AUTHOR]

HER-2/neu is a protein present on the surface of specific cancer cells and has been linked to the development and progression of certain cancer types. It is present in 15 to 20% of breast cancers and is clinically significant due to the availability of multiple anti-Her2 treatment options. Immunohistochemistry (IHC) is the most commonly used method to evaluate and quantify the expression of Her-2/neu. Although IHC is well-standardized in clinical practice, it is still subjected to inter-observer variability. Automating Her-2/neu scoring can improve accuracy, efficiency, consistency, and cost-effectiveness while reducing pathologists' workload. A deep learning-based automatic framework was utilized for the automatic detection of Her-2/neu score from whole slide images (WSI). The framework consists of three phases: identification of tumor patches, scoring of tumor patches, and Her-2/neu score prediction for whole slide images (WSI) based on the distribution of each score. This work used the dataset from the University of Warwick HER2 challenge contest. Two expert pathologists evaluated all 86 WSIs and assigned Her-2/neu scores to them. In addition, patches were generated from 50 WSIs and annotated individually by the pathologists. A total of 6641 extracted patches were generated out of which, 947 were labeled as 0, 327 as 1+, 1401 as 2+, 2950 as 3+, and 1016 were marked for discarding. Four pre-trained image classification models, namely DenseNet201, GoogleNet, MobileNet_v2, and a Vision Transformer based model, were fine-tuned, and tested on the generated patches. In order to predict the Her-2/neu score of the entire WSI, a random forest classifier was trained to predict the Her-2/neu score from the percentages of patches of each score present in the whole slide image. In patch classification performances, the vision transformer-based model outperformed the other models by achieving an accuracy of 92.6% on tumor patch classification and 91.15% on patch score classification. The random forest classifier achieved an accuracy of 88% on four score classification (0, 1+, 2 + and 3 +) and 96% on three score classification (0/1+, 2 + and 3 +). The proposed deep learning-based framework for the automatic detection and evaluation of Her-2/neu expression in breast cancer obtained encouraging results. This framework has the potential to be used as a prognostic tool, providing a cost-effective and time-efficient alternative for generating clinically relevant results. However, additional research is required to assess the applicability of this pipeline in different contexts. [ABSTRACT FROM AUTHOR]

Data-driven models have been recognized as powerful tools to support Structural Health Monitoring (SHM). This paper contributes to the literature by exploring two data-driven approaches to detect damage through changes in a set of variables that assess the condition of the structure, and accommodates the challenge that may arise due to the influence of environmental and operational variabilities. This influence is reflected in the response of the structure and can reduce the probability of detecting damage in a structure or increase the probability of signaling false positives. This paper conducts a comparative study between a machine learning detection approach (supported by linear regression, random forest, support vector machine, and neural networks) and a cointegration approach, with the aim of detecting damage as early as possible. This study also contributes to the literature by evaluating the merits of the damage detection methods using real data collected from a small-scale structure. The structure is analyzed in a reference state and a perturbed state in which damage is emulated. The results show that both approaches are able to detect damage within the first 24 h, without ever signaling false positives. The cointegration based approach can notably detect damage after 10 h and 15 minutes, while the machine learning approach takes 20 h 30 m to detect damage. • Comparison of Machine Learning and Cointegration in SHM. • Consideration of damages and false alarms. • Usage of a reduced scale structure to reproduce the behavior of a bridge column. • Cointegration is the fastest to detect damage. [ABSTRACT FROM AUTHOR]

Hierarchical binary classifiers are often used for multiclass problems when the number of classes is significantly more. The hierarchical structure suffers from a decrease in detection probability when there is an increase in the number of layers and classes. The order and the choice of classifier used in individual blocks of the hierarchical structure affect the overall probability of detection. In this paper, Ant Colony Optimization (ACO) is proposed to optimize the order of the classifier block at each level. It is also proposed to assign a suitable classifier model for the individual classifier block at each level to maximize the overall probability of detection. The proposed techniques also use an Adaptive boosting (Adaboost) classifier model (weighted average of the constructed classifier) as an alternative approach instead of fixing one classifier model at each classifier block and optimizing the classifier block's order using ACO. The description using the probability of false negative (miss → p M), and the probability of false positive (false alarm → p F) gives the characteristics of the individual classifier blocks in the hierarchical structure. Monte carlo simulation is performed by randomly assigning values for p M and p F. It was found that the overall detection rate is consistently increasing using the proposed method for various attempts made in the Monte carlo simulation. The performance of the optimized binary hierarchical structures obtained using the proposed technique for the Engine vibration data with 42 classes and 53 block classifiers are compared with the default hierarchical structure. It was observed that there has been a significant improvement in the overall detection rate from 87.235% using the default structure (with Random forest ensemble to classify all classes in the hierarchy) to 90.92% using the proposed technique (ACO with Adaboost). The proposed technique can be used for applications that demand hierarchical, multi-level classifications. • Optimal Hierarchical binary classifier structure (HBCS) using an ACO-based algorithm. • Closed-form expression for computing the overall detection rate of the arbitrary HBCS. • Monte-Carlo simulation-based performance analysis of the proposed technique. • Demonstration of the proposed method using engine vibration application. [ABSTRACT FROM AUTHOR]

Consumer preference prediction aims to predict consumers' future purchases based on their historical behavior-level data. Using machine learning algorithms, the prediction results provide evidence to conduct commercial activities and further improve consumer experiences. However, missing values and imbalanced class problems of consumer behavioral data always make machine learning algorithms ineffective. While several methods have been proposed to address missing data or imbalanced class problems, few works have considered the relationships among missing mechanisms, imputation algorithms, imbalanced class methods, and the effectiveness of classification algorithms that use impute data. In this study, we aim to propose an adaptive process for selecting the optimal combination of amputation, imputation, imbalance treatment, and classification based on classification performance. Our research extends the literature by showing significant interaction effects between 1) the amputation mechanism and imputation algorithms, 2) imputation and imbalance treatments, and 3) imbalance treatments and classification algorithms. Using three consumer behavioral datasets from the UCI Machine Learning Repository, we empirically show that, among different classification methods, the overall performance of Random Forest is better than that of Logit, SVM, or Decision Tree. Moreover, Logit, as the most widely used classification method, suffers most from imbalance issues in real-world datasets. Furthermore, Metacost is always the best imbalance treatment for different imputation techniques or missing value mechanisms. [ABSTRACT FROM AUTHOR]

• Propose an improved random forest based on the improvement of decision trees. • Improve the evaluation mechanism for the classification effect of decision trees. • Propose a method for quantifying the diversity between decision trees. • Multiple tests verify the superiority of the proposed improved random forest. Random forest is one of the most widely used machine learning algorithms. Decision trees used to construct the random forest may have low classification accuracies or high correlations, which affects the comprehensive performance of the random forest. Aiming at these problems, the authors proposed an improved random forest based on the classification accuracy and correlation measurement of decision trees in this paper. Its main idea includes two parts, one is retaining the classification and regression trees (CARTs) with better classification effects, the other is reducing the correlations between the CARTs. Specifically, in the classification effect evaluation part, each CART was applied to make predictions on three reserved data sets, then the average classification accuracies were achieved, respectively. Thus, all the CARTs were sorted in descending order according to their achieved average classification accuracies. In the correlation measurement part, the improved dot product method was proposed to calculate the cosine similarity, i.e., the correlation, between CARTs in the feature space. By using the achieved average classification accuracy as reference, the grid search method was used to find the inner product threshold. On this basis, the CARTs with low average classification accuracy among CART pairs whose inner product values are higher than the inner product threshold were marked as deletable. The achieved average classification accuracies and correlations of CARTs were comprehensively considered, those with high correlation and weak classification effect were deleted, and those with better quality were retained to construct the random forest. Multiple experiments show that, the proposed improved random forest achieved higher average classification accuracy than the five random forests used for comparison, and the lead was stable. The G-means and out-of-bag data (OBD) score obtained by the proposed improved random forest were also higher than the five random forests, and the lead was more obvious. In addition, the test results of three non-parametric tests show that, there were significant diversities between the proposed improved random forest and the other five random forests. This effectively proves the superiority and practicability of the proposed improved random forest. [ABSTRACT FROM AUTHOR]

Use of social media for communication, sharing or expressing views, broadcasting news, threatening and blackmailing has become an integral part of society. One such activity is understanding multi-cultural wedding images uploaded on social media. This paper presents a novel method based on the combination of U-Net, Convolutional Neural Network and Random Forest for classification of multicultural wedding images. In the case of wedding images, bride and bridegroom draw the attention of the viewers. This observation led to propose a U-Net for segmenting the region of bride and bridegroom in a novel way. Similarly, it is noted that the costumes of bride and bridegroom are vital information for differentiating different cultures. This cue motivated us to extract features using CNN for classification. Since the extracted features using CNN are capable of discriminating images of different classes, we propose a simple and effective Random-Forest for Multicultural Wedding Image Classification. The efficiency of the proposed model is demonstrated by testing it on our own dataset of six multi-cultural wedding classes and standard dataset of wedding and non-wedding images classes. Experimental results on both the datasets show that the proposed model outperforms the state-of-the-art models in terms of average classification rate. [ABSTRACT FROM AUTHOR]

Several giant water diversion projects go through large expansive soil areas in China. It is challenging to ensure the slope stability of the deep excavated expansive soil canal segments due to its undesirable geologic, and physical and mechanical properties. This study develops a novel deformation prediction method for deep excavated expansive soil canal slopes based on the eXtreme Gradient Boosting (XGBoost) algorithm, and provides a result interpretation using SHAP (SHapley Additive exPlanations). Firstly, the main factors influencing the slope deformation of deep excavated expansive soil canal segments are discussed. Subsequently, the integrated XGBoost-SHAP method is proposed for the prediction model construction of slope deformation and its result interpretation, the related the methodologies including the variational mode decomposition (VMD), XGBoost, and SHAP are also presented. The ten-fold cross-validation is employed to find the hyperparameter combination for the deformation prediction models, and the model performances are compared with common validation measurements. The effectiveness of the explanatory deformation prediction method is verified through one typical deep excavated expansive canal segment of the mid-route of the South-to-North Water Diversion Project in China. The case study shows that the VMD algorithm can well decompose the trend, periodic and fluctuating displacements. Compared with random forest and least squares-support vector machines, the XGBoost-algorithm based model achieves the better prediction performance. SHAP provides prediction result interpretations at both global and local levels. Time dependent effect and groundwater have a significant impact on the slope deformation of deep excavated expansive soil canal segments. The integrated SHAP-XGBoost method can provide a reference for predicting slope deformation of the similar projects. It can also quantify the contribution and mechanism of influencing factors on slope deformation at both global and local levels, which is conducive to implement reinforcement measures. [ABSTRACT FROM AUTHOR]

The accumulation of large amounts of historical data has led to the wide application of ensemble learning over the past few decades, but the balance between the individual accuracy of base classifiers (BCs) and the diversity among these BCs is rarely considered in the construction of ensemble models. Since such a balance is crucial to the success of ensemble models, this paper proposes a Genetic Algorithm-based sequential instance selection framework to address this research gap. The novelties of the proposed framework include: transforming the balance between the individual accuracy of BCs and the diversity among BCs into a general combinatorial optimization model and designing a Genetic Algorithm-based evolutionary instance selection method to solve this model. The proposed framework not only overcomes the inherent limitations of the Genetic Algorithm in some high-dimensional tasks but also provides an explicit and automatic way to balance the accuracy and diversity by searching appropriate training data subsets for different component BCs. Based on obtained training data subsets, the component BCs of the ensemble model are generated sequentially, and their predictions are further combined with the weighted majority voting rule. Using 30 real datasets collected from various practical applications, such as medicine, business, and industry, the effectiveness of the proposed framework in constructing powerful ensemble models is examined and compared with six benchmark ensemble learning methods. In addition, the capability of the proposed framework to improve convergence performances is also examined by the comparison with the traditional Genetic Algorithm. [ABSTRACT FROM AUTHOR]

Location indices are key in explaining variation in house prices. However, the definition of comprehensive indices capturing all locational features, along with their efficient and timely calculation, is usually one of the most complex dimensions of house price modeling. Existing difficulties result in partial location specifications, mostly due to three hurdles: (1) there is not a consensus on the best method to construct these indices, (2) what features (variables) to include: labor, demographic, commuting, etc., and (3) its creation requires granular and updated datasets. We introduce a methodology based on computer algorithms to create car and walk accessibility indices that address the previous concerns and capture location interactions among a wide range of variables. The selection of variables is based on an automated search of the best performing utility bearing gravitational accessibility indices for price prediction. Once these optimal indices are obtained, the method applies principal components analysis to secure their orthogonality. Using a unique dataset from a leading real estate portal in Europe, we illustrate and test for the city of Madrid their applicability in several house asking price models that are estimated using regression analysis and random forests (as a representative family of machine learning techniques). The experimental analysis reveals that using the optimal indices results in significant improvements in accuracy, for both regression-based models (13%) and random forests (21.6%), while achieving a substantial reduction in spatial autocorrelation (around 35%). The generated indices are clearly interpretable, which makes them a valuable tool for urban analyses (planning, transportation, sustainability, etc.). Finally, the methodology can be extended to other types of real state (commercial, industrial, etc.) and location (country, region, etc.). • Common accessibility variables in hedonic price models are too simply defined. • We introduce algorithms to calculate optimal gravity-based accessibility indices. • The new indices are used in hedonic regression and random forest predicting prices. • The new indices increase the accuracy and predicting power of the models. • The new indices reduce spatial correlation and dependence. [ABSTRACT FROM AUTHOR]

This study was aimed at comparing the differences in image textures of windfall apple samples dried using different non-thermal and thermal techniques without and with ultrasound pretreatment. Six classes of sliced windfall apple samples, such as freeze-dried (control), freeze-dried (ultrasound pretreatment), greenhouse-dried (control), greenhouse-dried (ultrasound pretreatment), open-sun-dried (control), and open-sun-dried (ultrasound pretreatment) were obtained. The models to distinguish apple samples were developed based on selected image textures using machine learning algorithms. In the first step of the analysis, all six classes were distinguished. Then, apple samples were classified in pairs without and with ultrasound pretreatment to determine the effect of pretreatment on apple structure. Additionally, models were built to distinguish three control samples from each other as well as separately to classify three samples with ultrasound pretreatment. In the case of the classification of all six apple samples, an overall accuracy reached 80.83% for a model built using a Logistic Model Tree algorithm from the group of Trees and there were no mixing cases between freeze-dried without and with ultrasound pretreatment and samples dried using other techniques. Samples freeze-dried without and with ultrasound pretreatment were distinguished from each other with an accuracy reaching 95% (Random Forest from Trees). Greenhouse-dried (control) and greenhouse-dried with ultrasound pretreatment apple samples were classified with an accuracy of up to 93.5% (Bayes Net from Bayes). Whereas an accuracy reached 95% (Bayes Net) for the classification of open-sun-dried (control) vs. open-sun-dried with ultrasound pretreatment apples. Apple samples dried using freeze drying, greenhouse drying and open-sun drying without pretreatment were distinguished with an overall accuracy of up to 83.67% (Random Forest), and samples dried by freeze drying, greenhouse drying and open-sun drying with ultrasound pretreatment – with an accuracy of up to 93.67% (Bayes Net). The developed models may be used in practice to distinguish apple samples dried using different techniques and samples dried with and without pretreatment. [ABSTRACT FROM AUTHOR]

In the context of global sustainable development, environmental, social, and governance (ESG) investment has become a frontier topic in the field of asset management. This paper focuses on ESG-valued multi-period portfolio selection problem which incorporates ESG factor into traditional portfolio optimization. Using ESG-valued returns instead of traditional returns, we propose an ESG-valued fused LASSO model to promote an optimal sparse multi-period portfolio strategy. The objective of the model is to minimize a combination of returns and risks with respect to ESG ratings based on classical Markowitz mean–variance framework. The sparsity of the portfolio at each period and the turnover across periods is achieved by the ℓ 1 regularization approach. To improve the out-of-sample performance of the ESG-valued model, we introduce two machine learning methods, namely random forest and support vector regression, to predict the ESG-valued returns. We develop a symmetric alternating direction method of multipliers to solve the regularized optimization model. Additionally, some other sparsity-driven penalty functions are discussed which results in a general framework of multi-period portfolio optimization. Finally, numerical experiments on several real datasets from both in-sample and out-of-sample demonstrate the positive effect of ESG in multi-period portfolio optimization. [ABSTRACT FROM AUTHOR]

The detection of epileptic seizures in EEG signals is a challenging task because it requires careful review of multi-channel EEG recordings over a lengthy time interval. In general, EEG-based seizure detection is strongly dependent on the ability to select descriptive features that are also stable in the sense that they are not sensitive to changes in the training data. This study proposes and investigates a patient-independent seizure detection model that uses stable EEG-based features obtained by comparing multiple feature selection methods. The schemes considered can be divided into five categories, often referred to as similarity, information theoretic, sparse learning, statistical, and graph centrality feature selection methods. The stability of the features these schemes produce was evaluated by random forest classification, using the intersection, frequency, correlation, and similarity measures. In experiments with Temple University Hospital's Seizure database of EEG records of 341 patients, unsupervised graph centrality feature selection was the most effective method, with a correct classification rate of 91.5%. • This study introduces a patient-independent method for detecting epileptic seizures. • Stability of features selection methods is considered. • Highest stability obtained via an unsupervised graph-based feature selection method. • The stable selected features were mostly extracted from the F8 and T3 channels. • The proposed method achieves 91.5% classification accuracy. [ABSTRACT FROM AUTHOR]

The accumulation of dust and other particles on solar panels, known as soiling, is a significant factor that affects their performance, leading to reduced efficiency if not addressed properly. In this study, we propose a new methodology to estimate soiling on solar photovoltaic panels. To address this issue, we utilised data from the University of Jaén and satellite information from NASA. We applied five different machine learning models, including Linear Regression, Random Forest, Decision Tree, Multilayer Perceptron and Long Short-Term memory neural networks to estimate the extent of soiling on the panels. The input data consisted of weather data, as well as operational data of the solar panels. Our results showed that the MLP model had the lowest average error of 0.0003, indicating its effectiveness in estimating the extent of soiling on the panels. This is significantly lower compared to previous proposals in the literature, which had an average error of 0.026. This study demonstrates the effectiveness of using machine learning methods to forecast soiling on photovoltaic panels accurately. The implications of our findings are essential for optimising energy production and improving the efficiency of solar power systems. [ABSTRACT FROM AUTHOR]

Dysgraphia is a neurological disorder that hinders the acquisition process of normal writing skills in children, resulting in poor writing abilities. Poor or underdeveloped writing skills in children can negatively impact their self-confidence and academic growth. This work proposes various machine learning methods, including transfer learning via fine-tuning, transfer learning via feature extraction, ensembles of deep convolutional neural network (CNN) models, and fusion of CNN features, to develop a preliminary dysgraphia diagnosis system based on handwritten images. In this work, an existing online dysgraphia dataset is converted into images, encompassing various writing tasks. Transfer learning is applied using a pre-trained DenseNet201 network to develop four distinct CNN models separately trained on word, pseudoword, difficult word, and sentence images. Soft voting and hard voting strategies are employed to ensemble these CNN models. The pre-trained DenseNet201 network is used for CNN feature extraction from each task-specific handwritten image data. The extracted CNN features are then fused in different combinations. Three machine learning algorithms support vector machine (SVM), AdaBoost, and Random forest are employed to assess the performance of the CNN features and fused CNN features. Among the task-specific models, the SVM trained on word data achieved the highest accuracy of 91.7%. In the case of ensemble learning, soft voting ensembles of task-specific CNNs achieved an accuracy of 90.4%. The feature fusion approach substantially improved the classification accuracy, with the SVM trained on fused features from the task specific-data achieving an accuracy of 97.3%. This accuracy surpasses the performance of state-of-the-art methods by 16%. • Diagnosis of dysgraphia in children from offline handwritten images. • Ensemble of deep CNNs trained on handwriting task-specific data. • CNN features of handwriting task-specific data are fused for improved diagnosis. • High dysgraphia diagnosis sensitivity (97%). [ABSTRACT FROM AUTHOR]

In this paper, we investigate the impact of pattern leakage during data preprocessing on the reliability of Machine Learning (ML) based intrusion detection systems (IDS). Data leakage, also known as pattern leakage, occurs during data preprocessing when information from the testing set is used in training, leading to overfitting and inflated accuracy scores. Our study uses three well-known intrusion detection datasets: NSL-KDD, UNSW-NB15, and KDDCUP99. We preprocess the data to create versions with and without pattern leakage and train and test six ML models: Decision Tree (DT), Gradient Boosting (GB), K-neighbours (KNN), Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR). Our results show that building IDS models with data leakage leads to higher accuracy but is unreliable. Additionally, we find that some algorithms are more sensitive to data leakage than others, as seen by the drop in model accuracy when built without leakage. To address this problem, we provide suggestions for mitigating data leakage in the training process and analyzing the sensitivity of different algorithms. Overall, our study emphasizes the importance of addressing data leakage in the training process to ensure the reliability of ML-based IDS models. [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]

This paper proposes an intelligent slope stability prediction method based on the improved pelican optimization algorithm (IPOA) and the optimization random forest (RF) algorithm to reduce disasters and accidents caused by slope instability. First, exploratory data analysis (EDA) is performed using correlation diagrams, heat maps under different states, box plots, histograms, and quantile–quantile (Q-Q) diagrams of variables, followed by establishing a high-quality data set for slope engineering cases and an index system for slope stability prediction. Second, 10 benchmark functions reveal that the IPOA algorithm outperforms other algorithms. Accordingly, this paper develops a slope stability prediction model based on the IPOARF algorithm. Afterward, a set of intelligent slope stability prediction systems is created using MATLAB tools and applied to Lala Copper Mine in Sichuan Province. Finally, this paper compares the accuracy of various models and subjects the proposed model to additional testing. The results reveal that the prediction model based on improved IPOA and RF algorithms is reliable and effective, with an accuracy of up to 90.4%, which can serve as a solid technical basis for slope instability disaster prediction in geotechnical engineering. [ABSTRACT FROM AUTHOR]

Accurate and reliable runoff prediction is of great significance to water resources management, disaster monitoring and rational development and utilization of water resources. In this paper, a metaheuristic evolutionary deep learning model based on Temporal Convolutional Network (TCN), Improved Aquila Optimizer (IAO) and Random Forest (RF) is proposed for rainfall-runoff simulation and multi-step runoff prediction. In this study, the influence of various input variables on the prediction accuracy is discussed. First of all, in order to avoid the dimensional disaster problems and reduce the calculation time, RF is used to calculate the correlation between the input variables and the prediction object, and the data with high correlation is selected as the final inputs. Then, the filtered data are sent to the TCN model, and the parameters of the TCN model are optimized using the IAO algorithm, and the final prediction results are obtained. In this study, the rainfall and runoff data of five stations in the middle reaches of Jinsha River, China were selected, and the runoff of Panzhihua station was simulated and predicted by establishing multiple models. By analyzing and comparing the predictive results of several models, it shows that the models and improvements proposed in this study are effective. [ABSTRACT FROM AUTHOR]

Accurate and reliable runoff prediction is of great significance to water resources management, disaster monitoring and rational development and utilization of water resources. In this paper, a metaheuristic evolutionary deep learning model based on Temporal Convolutional Network (TCN), Improved Aquila Optimizer (IAO) and Random Forest (RF) is proposed for rainfall-runoff simulation and multi-step runoff prediction. In this study, the influence of various input variables on the prediction accuracy is discussed. First of all, in order to avoid the dimensional disaster problems and reduce the calculation time, RF is used to calculate the correlation between the input variables and the prediction object, and the data with high correlation is selected as the final inputs. Then, the filtered data are sent to the TCN model, and the parameters of the TCN model are optimized using the IAO algorithm, and the final prediction results are obtained. In this study, the rainfall and runoff data of five stations in the middle reaches of Jinsha River, China were selected, and the runoff of Panzhihua station was simulated and predicted by establishing multiple models. By analyzing and comparing the predictive results of several models, it shows that the models and improvements proposed in this study are effective. [ABSTRACT FROM AUTHOR]

This paper proposes an intelligent slope stability prediction method based on the improved pelican optimization algorithm (IPOA) and the optimization random forest (RF) algorithm to reduce disasters and accidents caused by slope instability. First, exploratory data analysis (EDA) is performed using correlation diagrams, heat maps under different states, box plots, histograms, and quantile–quantile (Q-Q) diagrams of variables, followed by establishing a high-quality data set for slope engineering cases and an index system for slope stability prediction. Second, 10 benchmark functions reveal that the IPOA algorithm outperforms other algorithms. Accordingly, this paper develops a slope stability prediction model based on the IPOARF algorithm. Afterward, a set of intelligent slope stability prediction systems is created using MATLAB tools and applied to Lala Copper Mine in Sichuan Province. Finally, this paper compares the accuracy of various models and subjects the proposed model to additional testing. The results reveal that the prediction model based on improved IPOA and RF algorithms is reliable and effective, with an accuracy of up to 90.4%, which can serve as a solid technical basis for slope instability disaster prediction in geotechnical engineering. [ABSTRACT FROM AUTHOR]

• Use feature selection to extract the most discriminating features. • Improve accuracy and reduce bias using an ensemble technique. • A One-vs-Rest Consensus Learning (OvRCL) technique is proposed. Research in crash severity prediction is necessary to allow safety planners to take precautionary measures and enable first aiders to remain prepared for assisting the injured. Existing literature in the field of crash severity prediction is mostly focused on generating the attributes for predicting the severity. However, in reality, not all features are discriminating, and certain classes are challenging to detect even employing the entire feature set. Although to tackle these problems several techniques are developed in the Machine Learning (ML) literature. But their application to crash severity prediction and an optimal strategy for the best combination of features or classifiers for achieving high accuracy is a less studied area. To address these problems, this work first provides a comparison of widely used classifiers for predicting crash severity; and secondly, by combining class-wise majority voting with One-vs-Rest (OvR) approach, a novel classification framework named, OvR consensus learning (OvRCL) is proposed. The proposed method avail a feature selection technique, Mutual information (MI), to acquire the most relevant feature set regarding the output class (i.e. severity). Moreover, to differentiate each class in the multi-class data, OvRCL iteratively runs ML algorithms as binary classifiers in an ensemble framework to significantly ameliorate classification performance. In our experiments, we use four classifiers, namely, the k -Nearest Neighbors (k -NN), Support Vector Machines (SVM), Random Forest (RF), and Bagging classifier, to get the consensus. Analysis was done using a real crash dataset obtained from an open data source of Leeds city council. A four-year crash data (2015–2018) is used for training and the OvRCL is tested on the 2019 data. Moreover, to validate the performance of OvRCL, this study also utilizes two more datasets with high-class imbalance. In contrast to conventional ML algorithms, our experiments depict that the OvRCL is a potent method for forecasting crash severity levels on the data under test. [ABSTRACT FROM AUTHOR]

The primary challenge faced in extractive text summarization is related to the scoring of sentences, with the critical factor for scoring being the manner in which the sentence representation is conducted. This study aims to investigate this hypothesis and to perform a detailed analysis of the impact of sentence representation techniques that have been used both semantically and syntactically. The study initially evaluated the empirical impact of individual syntactic and semantic features on the accuracy of summarization. To examine syntactic usage, a comprehensive list of 40 syntactic features was developed, while semantic representation was accomplished using sentence embeddings. Subsequently, an improved feature set was proposed that jointly utilizes syntactic and semantic features. To assess the impact of this feature set on the resulting summaries, the proposed sentence representation was tested on three distinct summarization corpora consisting of lengthy scientific documents across diverse domains. The assessment of summary evaluation and classification performance evaluation metrics was conducted to evaluate the quality of the resulting summaries. The findings of the experiments indicated that the summaries generated by the proposed feature set performed better than not only those obtained using individual features but even summaries produced by state-of-the-art methods. • In literature, text summarization has been handled by syntactic or semantic approach. • A new enhanced sentence representation was proposed to score sentences. • The sentence scoring model was a random forest fed by this enhanced representation. • The proposed model remarkably outperformed the baseline and state-of-the-art models. [ABSTRACT FROM AUTHOR]

Flight delay prediction has become one of the most critical topics in intelligent aviation systems due to its essential role in flight scheduling, airline planning, and airport operation. The accurate prediction of flight delays is very challenging because numerical factors will affect flight delays. Moreover, owing to the connectivity of the aviation system, flight delays also present complex spatial–temporal correlations, including the spatial correlations between airports along with the temporal correlations among timestamps. To address these challenges, we proposed a CNN-LSTM deep learning framework to consider the spatial–temporal correlations together with the extrinsic features for flight delay prediction. The CNN-LSTM model consists of a Convolution neural network (CNN) architecture to learn the spatial correlations followed by a Long short-term memory (LSTM) architecture to capture the temporal correlations. The spatial–temporal correlations obtained from the CNN-LSTM framework are then fused with the extrinsic features (e.g., airline issues, distance, schedule fly time, etc.) as inputs of the random forest (RF) model for flight delay prediction. The U.S. domestic flights in 2019 are collected from the Bureau of Transport Statistics to confirm the outperformance of the proposed model. The results show that the accuracy of the CNN-LSTM model reaches 92.39%. For the on-time samples, approximately 91% are correctly identified; for the delayed samples, the classification accuracy reaches 84% which exhibits better performance compared with several benchmark models. The created prediction model of this study could provide useful information for airport regulators in understanding the potential delays in advance and developing effective airport management strategies for improving airport on-time performance. [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]

Preserving integrity in online assessments is a matter of concern, worldwide. There exist several ways for cheating in online assessments. Exploiting the available Internet for finding solutions is one of the popular ways of cheating. Several researchers have proposed solutions to handle Internet cheating. Some of them suggest the use of secure browsers, however, these are vulnerable to hacking and prone to technical errors. Many works propose e-proctoring, but it is a resource-intensive method and has significant privacy concerns. Other works propose preventive measures like paraphrasing of questions, but smart google search algorithms defeat the purpose. In this work, we use machine learning to create a model by analysing the assessment log files, for the detection of cheaters who indulge in Internet cheating. Additionally, to address the persistent problem of the identification of ground truth in academic dishonesty, we modify an online quiz tool to collect labelled data. We transform the raw dataset using feature engineering methods, to derive thirteen features from the student and question-related attributes of the assessment log files. Our objective is to obtain the best predictive model for the classification of honest and dishonest students. We create models using two feature selection algorithms (ANOVA and Mutual Information) and five machine learning classifiers (Logistic regression, Support vector machines, Naïve Bayes, K nearest neighbour and Random forest) and evaluate them. From among all the models, Random forest classifier with top features selected by the MI method obtains the best performance with an accuracy of about 85%. We discuss the features that are most influential for the automated detection of cheaters. We also give insights into the critical aspects of a cheat-proof assessment design. [ABSTRACT FROM AUTHOR]

Polycystic Ovary syndrome (PCOS) is the most common serious health problem in women worldwide. It affects the women's fertility and may lead to various health problems like type 2 diabetes, cardiovascular diseases, etc. Early diagnosis of PCOS is recommended for proper treatment. In this paper, we propose SmS: SMOTE-Stacked hybrid model, for the early diagnosis of PCOS. In the proposed model, we combine the ensemble technique 'stacking' with oversampling technique 'SMOTE' i.e., Synthetic minority oversampling technique. In the Stacking ensemble technique, training is performed at two levels: base level and meta level. We use six classifiers Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Naïve Bayes (NB) and AdaBoost (AdaB) as base leaners at the base level and consider each of these classifiers separately at the meta -level to implement SmS hybrid models: Stack-LR, Stack-SVM, Stack-DT, Stack-RF, Stack-NB and Stack-AdaB. Standardization, SMOTE and Backward Feature Elimination (BFE) are performed before training the classifiers. We evaluate the performance of the individual classifiers and hybrid models on PCOS dataset (Imbalanced) taken form the Kaggle Repository and observe that Stack-AdaB shows the most promising results. [ABSTRACT FROM AUTHOR]

In the context of reinforced concrete members subjected to shear, the steel behaviour, assumed as embedded in the concrete, has been modelled through different strategies. One of them, the so-called Refined Compression Field Theory (RCFT), is based on the concept of the concrete tension stiffening area, and shows a better fitting with the experimental results than other shear theories. However, for certain standard design conditions, the RCFT non-linear formulation does not throw a real physical solution for the steel yield strain, what hinders its numerical calculation. In fact, previous works have defined a solvability region for such strain. This non-linear equation is usually solved by iterative methods. However, the convergence of these methods depends on the root location. Moreover, their robustness may diminish in the boundary of the solvability region. This work validates Machine Learning (ML) approaches as alternative tools for the iterative prediction of the steel yield strain. To this aim, the calculation of the steel yield strain is performed in two stages, first locating the root to calculate using ML-based classification techniques, and second calculating such root with an accurate adjustment as a function of its location, using ML-based regression algorithms. As result, the efficiency of the proposed method no longer depends on the steel strain location. Finally, despite the high influence of the concrete bond parameter in the yield strain field, the ML prediction error is homogeneous over all the problem physical domain. [Display omitted] • Accurate calculation of steel yield strain in reinforced concrete is tackled. • Iterative numerical methods usually present convergence problems. • Machine Learning algorithms may improve the calculation of steel yield strain. • Decision Trees, Random Forest and Support Vector Machines are applied. [ABSTRACT FROM AUTHOR]

The gradual depletion of surface water in major lakes and their impact in the sustainable development of local water resources has been a great challenge. Monitoring surface water and detecting changes in the lake are the main objectives of this study. Landsat Thematic Mapper (TM), Landsat Enhanced Thematic Mapper (ETM+) and Landsat Operational Land Imager (OLI) of 2010, 2000 and 2018 Lake Urmia images acquired from US Geological Survey were used for detecting the changes. Surface water changes are usually identified by extracting the water features from individual time series multispectral images. In this study, a novel change detection framework has been proposed involving pixel level fusion and classification. The spatial frequency based undecimated wavelet transform fusion (UDWT – SF) effectively extracted the spectral information from MS image and spatial information from PAN image of the same scene captured at different time periods. The endmembers of the fused images were selected using pixel purity index endmember extraction algorithm and the abundance estimation by the application of fully constrained least square spectral unmixing algorithm. An efficient sub-pixel classification process is designed by employing the spectral signatures, Normalized Difference Water Index (NDWI) and the abundance estimation generated from the pansharpened image in a random forest classifier. Experimental results indicate that the proposed classifier attained 99.9945% user's accuracy for water area and 99.9675% producer's accuracy for changed area. Similarly, the producer's accuracy of water and changed area are 99.9866% and 99.9868% respectively. The kappa coefficient and the overall accuracy of the proposed sub-pixel random forest classification on multitemporal multispectral image is 0.97 and 99.89% respectively. The lake surface area is computed and it is found that an area of 1369 Sq km has been decreased from the year 2000 to 2010 and 310 Sq km decreased from 2010 to 2018 as per the assessment based on the proposed random forest classifier. Spectral unmixing based random forest classifier (RF-SP) on fused image yields better results in terms of accuracy compared to other classifiers and it is more appropriate for detecting the multitemporal surface water changes. [ABSTRACT FROM AUTHOR]

• EEG signals were used for the identification of asthma severity levels. • Three methodologies were designed to identify a subject's asthma severity levels. • Identifications were made based on subject's data, new subject's data, and mixed data. • In two of the methodologies, the ensemble and XGBoost outperform other classifiers. • All three proposed methodologies may be employed for personalised home monitoring. Asthma is a complex respiratory disorder in which structural changes in the conducting airway cause variable airflow limitation. The excessive narrowing of the airway lumen underlies the morbidity and the mortality that are attributable to the disease, which reduces quality of life in people of all ages. Monitoring of the disease progression to gather data continuously as an objective marker of morbidity and to support therapy is essential. Currently, patients with asthma are examined only a few times a year, which prevents continuous monitoring of the disease progression and any advance in precision medicine. Remote non-invasive monitoring tools will enable the collection of data with minimal effort from patients. To date, however, assessment and monitoring of asthma based on electroencephalogram (EEG) signals have not been studied. The objective of this research was to develop a general approach for identifying asthma severity levels based on EEG signals for personalised remote and non-invasive monitoring of patients with asthma. Simultaneous measurements of EEG and respiration motion signals were acquired from adults with suspected asthma, during the entire methacholine challenge test. The EEG segments were categorized into three classes, each representing a level of asthma severity based on participant's spirometry score. Three artificial intelligence (AI) methodologies were designed and examined: the first aimed to identify a subject's asthma severity levels based on their known data, the second based on mixed data comprising data from all subjects including the subject's personal data, and the third methodology based on the datasets of all other subjects, reflecting a situation of a new patient. To overcome multi-subject variations in the third methodology, the probabilities of being at each one of the possible asthma severity levels in previous breathing cycles, as inputs for predictions of asthma severity levels in the current breathing cycle, was used. The classification was done based on ordinal and non-ordinal classification algorithms. In the first and second methodologies, an ensemble approach was also applied to enhance predictive performance. Good performance measures were obtained to identify asthma severity level using both the first and second methodologies, especially by the ensemble model and XGBoost classifier. In the third methodology, the combination of random forest and ordinal random forest in different stages of the methodology, yielded the best improvement in performance measures. The results of all three AI methodologies demonstrate that they may be employed for personalised home asthma monitoring and management that do not depend on patients exerting an effort. [ABSTRACT FROM AUTHOR]

• RF, AGD, GB, and XGB models can accurately predict the TL of strands in PSC beams. • GB and XGB models outperform RF and AGD models. • Ensemble ML models significantly improve prediction accuracy. • f ci , f pi , d b , c , b and h are found to have the most significant impact on the TL. • A web application can rapidly predict the TL of strands with less effort and cost. This study aims to develop four ensemble machine learning (ML) models, including Random Forest (RF), Adaptive Gradient Boosting (AGB), Gradient Boosting (GB), and Extreme Gradient Boosting (XGB), for estimating the transfer length of strands in prestressed concrete (PSC) beams. The results of eleven well-known empirical equations and four single ML models, including Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), K-nearest Neighbors (KNN), and Decision Tree (DT), are used to evaluate the performance of the developed ensemble ML models. This study shows that the GB and XGB models agree well with experimental results and significantly outperform empirical equations and single ML models. The SHapley Additive exPlanations method based on the GB and XGB models determines the compressive strength of concrete at prestress release, initial prestress, strand diameter, concrete cover, beam section width, and beam section height have the most significant effect on the transfer length of strands in PSC beams. Eventually, a web application is built based on the best ML models for practical use. It can predict the transfer length of strands in PSC beams without costly and time-consuming tests. [ABSTRACT FROM AUTHOR]

Members of the armed forces rely on having an effective and efficient medical evacuation (MEDEVAC) process for evacuating casualties from the battlefield to medical treatment facilities during combat operations. Determining which MEDEVAC units to task to respond to incoming requests for service, known as the MEDEVAC dispatching problem, is a natural sequential decision making problem and is the focus of this paper. Unlike previous research in this area, this paper incorporates triage classification errors and the possibility of having blood transfusion kits on board select MEDEVAC units. A discounted, infinite-horizon continuous-time Markov decision process model is formulated to examine the MEDEVAC dispatching problem and to compare generated dispatching policies to the currently practiced United States military policy. We utilize an approximate dynamic programming (ADP) technique that leverages a random forest value function approximation within an approximate policy iteration algorithmic framework to develop high-quality policies for both small-scale and large-scale problem instances. A representative planning scenario involving joint combat operations in South Korea is developed and utilized to investigate the differences between the various policies. Results from the analysis indicate that applying ADP techniques can improve current practices by as much as 29% with regard to a life-saving performance metric. This research is of particular interest to the military medical community and can inform the procedures of future military MEDEVAC operations. [ABSTRACT FROM AUTHOR]

A detailed and extensive empirical study of dynamic selection (DS) and random under-sampling (RUS) for the class imbalance problem is conducted in this paper. Total 20 state of the art DS methods are compared on 54 datasets. The empirical results clearly answer the following six key research questions in this direction, (1) how performances of ensembles with and without RUS compare with respect to different DS and static ensemble (SE) methods (2) how whether RUS is used affects performances of ensembles with respect to different DS/SE methods (3) how performances of different DS/SE methods compare with respect to ensembles with and without RUS (4) whether DS methods perform better than SE methods no matter whether RUS is used and what types of ensembles are used (5) how numbers of base classifiers affect how performances of ensembles with and without RUS compare with respect to different DS/SE methods (6) how numbers of base classifiers affect how performances of different DS/SE methods compare with respect to ensembles with and without RUS. The answers to the six research questions based on the experimental results in this study and the experimental findings are the main contributions of this work. • An extensive empirical study of DES with RUS for class imbalance is conducted. • Total 20 state of the art dynamic selection methods are compared on 54 datasets. • Three types of ensembles, bagging, random forests and AdaBoosting, are compared. • The experimental results and findings are important for the DES research community. [ABSTRACT FROM AUTHOR]

Increasing the accuracy of prediction models in financial markets is an important but difficult task due to the natural complexities of financial time series, which are nonlinear and nonstationary. This challenge has made machine learning methods popular in recent years. However, the noise contained in financial series dramatically distorts the performance of such approaches. This paper proposes an adaptive denoising method called MIC-EMD that is data-driven and removes the noise contained in input features based on the nonlinear relationship between the target variable and the input features. To verify the advantages of MIC-EMD, a simulation experiment is conducted to compare the performance of several representative denoising methods with that of MIC-EMD. Finally, a comprehensive empirical analysis is performed for the trend predictions of six major indexes in Asian markets using three prevalent machine learning methods (SVM, random forest and LightBGM). After the input features are denoised by MIC-EMD, the results reveal the following: (i) its prediction performance outperforms that of the three learning models with input features denoised by state-of-the-art denoising methods such as ICA, WT, WF, EMD, aEMD, P-EMD and S-EMD, e.g., the prediction accuracies of the three machine learning models increase by 9.77%, 13.5% and 12.3%; and (ii) we obtain a prediction accuracy as high as 70%. • The new denoising method is data-driven. • It outperforms prevalent state-of-the-art denoising methods. • It increases prediction accuracies of machine learning methods by 10% or more. • We get accuracies up to 70% of trend predictions for major Asian stock indexes. [ABSTRACT FROM AUTHOR]

• A novel hybrid data reconstruct strategy is developed. • Comprehensive and rational subsystem determining and training strategies are applied. • A suitable interval forecasting framework was established successfully. • A scientific and pluralistic evaluation system is constructed. With the continuous increase in global photovoltaic installations, the importance of photovoltaic power generation to the power industry has gradually increased, which means that accurate forecasting and real-time management of photovoltaics have become indispensable. In recent years, although many photovoltaic prediction systems based on different mechanisms have been proposed, most of them are point prediction methods and do not fully consider the impact of various factors on photovoltaic power generation. Therefore, in order to will fill this gap, a novel hybrid interval prediction system that combines hybrid signal preprocessing technique, random forest algorithm, deep learning model, neural network model and swarm intelligence optimization strategy is designed in this paper. The system can make full use of the characteristics of independent variables, and effectively improve the stability and accuracy of photovoltaic prediction. According to the data obtained from Yulara Solar System, the prediction efficiency of the suggested system is verified. Specifically, when the interval width coefficient is 0.15, the prediction interval coverage probabilities obtained by the presented system are 73.373%, 92.899%, and 92.781%, respectively. Furthermore, this paper identifies the superior stability and application possibilities of the proposed interval forecasting system from multiple perspectives. [ABSTRACT FROM AUTHOR]

• Construct a general incomplete data processing model for the same structure incomplete data set. • Newly define the concept of abnormal degree. • Apply the Bisection method to the division of the missing rate intervals. • Newly proposed two judgment rules to ensure the feasibility of the processing method. • Apply to incomplete data processing in similar fields or with similar structures. Aiming at the problem of incomplete data such as outliers and missing values in machine learning, an incomplete data processing method based on the measurement of missing rate and abnormal degree was proposed in this paper. In this method, the outlier problem that are easy to be ignored was fully considered, and the Bisection method in the data structure field was applied to reduce the interval length and find the data distribution law. And then, an incomplete data processing model for static-structure or fixed-structure data set in a certain field or research direction was constructed. In the model establishment stage, the rules of the outlier processing part were first explored, including the boundary condition for outlier processing, and the relative applicable condition of using the direct-discarding method for outliers. Then, the rules of the missing value processing part were later explored, including the boundary condition for missing value processing, the relative applicable condition of using the direct-discarding method for missing values, the division of different missing rate intervals and the corresponding applicable numerical filling methods. In this research, the loose particle localization data set was taken as the test object, and the applicable incomplete data processing model was established. Specifically, the model can process outliers with the boundary condition of 23%, and the relative applicable condition of using the direct-discarding method for outliers is 2%. While the boundary condition that can process missing values is 67%, and the relative applicable condition of using the direct-discarding method for missing values is 3%. In the range of 3% to 67%, the range of missing rate that processing by statistical-filling methods is 3% to 7%, and the range of missing rate that processing by kNN prediction model is 7% to 67%. The applicable incomplete data processing model was verified on multiple loose particle localization data sets. Test results show that the prediction accuracy achieved by the classification learner based on the parameter-optimized Random Forest has significantly improved on the data set before and after processing, and the average improvement is 5.04%, which effectively proves the feasibility and practicability of the method proposed in this paper. Theoretically, it can be extended and applied to the processing of incomplete data sets in similar fields or with same structures, and has important reference significance for the research on incomplete data processing in machine learning. In addition, it has important practical value for improving the data completeness and data regularity of feature data sets in specific application fields. [ABSTRACT FROM AUTHOR]

Millimeter-wave communication systems design requires accurate path-loss (PL) prediction, which is critical to determining coverage area and system capacity. In this work, four machine learning (ML) algorithms are proposed for PL prediction in an indoor environment for 5G millimeter-wave frequencies, from 26.5 to 40 GHz. They are artificial neural network (ANN), support vector regression (SVR), random forest (RF), and gradient tree boosting (GTB). We compare their performances, including also the empirical PL models ABG and CIF incorporating the number of crossed walls, as we propose. For the ML models, we present a methodology to select the predictor coalition by examining the marginal performance and the interpretability gains. The methodology allows choosing the most appropriate predictor coalition for building machines while not overlooking the non-linear connections between predictors and the path-loss. The results show that the ML techniques improve the prediction accuracy of empirical models. Among the ML algorithms employed, considering the optimal subset of predictors, the GTB model presents the best prediction and generalization capacities. • Accurate machine learning regression for the path-loss of millimeter waves indoors. • Incorporating the number of crossed walls improves empirical path-loss models. • Methodology for predictor coalition selection considering interpretability. • The methodology examines the marginal performance and the interpretability gains. • The resulting models present feasible accurate path-loss prediction capacity. [ABSTRACT FROM AUTHOR]