Search

We developed an interpretable machine learning algorithm that prospectively predicts the risk of thrombocytopenia in older critically ill patients during their stay in the intensive care unit (ICU), ultimately aiding clinical decision-making and improving patient care. Data from 2286 geriatric patients who underwent surgery and were admitted to the ICU of Dongyang People's Hospital between 2012 and 2021 were retrospectively analyzed. Integrated algorithms were developed, and four machine-learning algorithms were used. Selected characteristics included common demographic data, biochemical indicators, and vital signs. Eight key variables were selected using the Least Absolute Shrinkage and Selection Operator and Random Forest Algorithm. Thrombocytopenia occurred in 18.2% of postoperative geriatric patients, with a higher mortality rate. The C5.0 model showed the best performance, with an area under the receiver operating characteristic curve close to 0.85, along with unparalleled accuracy, precision, specificity, recall, and balanced accuracy scores of 0.88, 0.98, 0.89, 0.98, and 0.85, respectively. The support vector machine model excelled at predictively assessing thrombocytopenia severity, demonstrating an accuracy rate of 0.80 in the MIMIC database. Thus, our machine learning-based models have considerable potential in effectively predicting the risk and severity of postoperative thrombocytopenia in geriatric ICU patients for better clinical decision-making and patient care. [ABSTRACT FROM AUTHOR]

Within the Hadoop ecosystem, MapReduce stands as a cornerstone for managing, processing, and mining large-scale datasets. Yet, the absence of efficient solutions for precise estimation of job execution times poses a persistent challenge, impacting task allocation and distribution within Hadoop clusters. In this study, we present a comprehensive machine learning approach for predicting the execution time of MapReduce jobs, encompassing data collection, preprocessing, feature engineering, and model evaluation. Leveraging a rich dataset derived from comprehensive Hadoop MapReduce job traces, we explore the intricate relationship between cluster parameters and job performance. Through a comparative analysis of machine learning models, including linear regression, decision tree, random forest, and gradient-boosted regression trees, we identify the random forest model as the most effective, demonstrating superior predictive accuracy and robustness. Our findings underscore the critical role of features such as data size and resource allocation in determining job performance. With this work, we aim to enhance resource management efficiency and enable more effective utilisation of cloud-based Hadoop clusters for large-scale data processing tasks. [ABSTRACT FROM AUTHOR]

Despite its proven success in various fields such as engineering, business, and healthcare, human–machine collaboration in education remains relatively unexplored. This study aims to highlight the advantages of human–machine collaboration for improving the efficiency and accuracy of decision-making processes in educational settings. High school dropout prediction serves as a case study for examining human–machine collaboration's efficacy. Unlike previous research prioritizing high accuracy with immutable predictors, this study seeks to bridge gaps by identifying actionable factors for dropout prediction through a framework of human–machine collaboration. Utilizing a large dataset from the High School Longitudinal Study of 2009 (HSLS:09), two machine learning models were developed to predict 9th-grade students' high school dropout history. Results indicated that the Random Forest algorithm outperformed the deep learning algorithm. Model explainability revealed the significance of actionable variables such as students' GPA in the 9th grade, sense of school belonging, self-efficacy in mathematics and science, and immutable variables like socioeconomic status in predicting high school dropout history. The study concludes with discussions on the practical implications of human–machine partnerships for enhancing student success. [ABSTRACT FROM AUTHOR]

The sleeping stage classification plays an important role in the medical science because it helps the diagnosis of the mental health diseases. The conventional approach for performing the sleeping stage classification is based on the electroencephalograms (EEGs). However, it is worth noting that the EEGs reflect the brain activities. Nevertheless, the brain activities are very complicated even though the subjects are sleeping. Hence, performing the sleeping stage classification via the EEGs may yield the low classification accuracy. On the other hand, the electrooculograms (EOGs) are the voltages between the front eyes and the back eyes which are related to the eye ball movement. As it can directly reflect the various sleeping stages, it can achieve a higher classification accuracy. Therefore, this paper employs the two channel EOGs for performing the sleeping stage classification. The major contribution of this paper is to 1) employ the singular spectrum analysis (SSA) to exploit the latent intrinsic high dimensional dynamics of the one dimensional EOGs for performing the sleeping stage classification, 2) employ the approximate entropy as the features for performing the sleeping stage classification, and 3) assign the same features of different SSA components of different channels of the epochs of the EOGs into the same group and perform the principal component analysis (PCA) on each group of the feature vectors so that the properties of each type of the features are preserved. The results show that our proposed method yields the five sleeping stage classification accuracy at 93.73% and the sensitivity of the stage one non-rapid eye movement (S1) at 78.44%, which achieves the significant improvements compared to the existing methods. Therefore, our proposed method could be used to reduce the workload of the medical officers. [ABSTRACT FROM AUTHOR]

For the detection of food adulteration, sensitive and reproducible analytical methods are required. Liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS) is a highly sensitive method that can be used to obtain analytical fingerprints consisting of a variety of different components. Since the comparability of measurements carried out with different devices and at different times is not given, specific adulterants are usually detected in targeted analyses instead of analyzing the entire fingerprint. However, this comprehensive analysis is desirable in order to stay ahead in the race against food fraudsters, who are constantly adapting their adulterations to the latest state of the art in analytics. We have developed and optimized an approach that enables the separate processing of untargeted LC‑HRMS data obtained from different devices and at different times. We demonstrate this by the successful determination of the geographical origin of honey samples using a random forest model. We then show that this approach can be applied to develop a continuously learning classification model and our final model, based on data from 835 samples, achieves a classification accuracy of 94% for 126 test samples from 6 different countries. [ABSTRACT FROM AUTHOR]

Impervious surface data are increasingly important for research and planning. Despite the availability of global and local urban land cover maps, regional data are lacking in Africa. We generated annual 30 m impervious cover data from 2001–2020 for Ghana, Togo, Benin, and Nigeria using the Landsat archive. We used random forest to predict impervious cover using 11 spectral indices and applied pixel-level temporal segmentation with the LandTrendr algorithm. Processing with LandTrendr improved the accuracy of the random forest predictions, with higher predicted-observed r2 (0.81), and lower mean error (−0.03), mean absolute error (5.73%), and root mean squared error (9.93%). We classified pixels >20% impervious as developed and < = 20% impervious as undeveloped. This classification had 93% overall accuracy and similar producer's (79%) and user's (80%) accuracies for developed area. Our maps had higher accuracy and captured more developed areas than comparable global datasets. This is the first regionally calibrated 30 m resolution impervious dataset in West Africa, which can support research on drivers and impacts of urban expansion and planning for future growth. [ABSTRACT FROM AUTHOR]

This study addresses the critical issue of anemia detection using machine learning (ML) techniques. Although a widespread blood disorder with significant health implications, anemia often remains undetected. This necessitates timely and efficient diagnostic methods, as traditional approaches that rely on manual assessment are time-consuming and subjective. The present study explored the application of ML – particularly classification models, such as logistic regression, decision trees, random forest, support vector machines, Naïve Bayes, and k-nearest neighbors – in conjunction with innovative models incorporating attention modules and spatial attention to detect anemia. The proposed models demonstrated promising results, achieving high accuracy, precision, recall, and F1 scores for both textual and image datasets. In addition, an integrated approach that combines textual and image data was found to outperform the individual modalities. Specifically, the proposed AlexNet Multiple Spatial Attention model achieved an exceptional accuracy of 99.58%, emphasizing its potential to revolutionize automated anemia detection. The results of ablation studies confirm the significance of key components – including the blue-green-red, multiple, and spatial attentions – in enhancing model performance. Overall, this study presents a comprehensive and innovative framework for noninvasive anemia detection, contributing valuable insights to the field. [ABSTRACT FROM AUTHOR]

Introduction: The Indian Himalayas' susceptibility to landslides, particularly as a location where climate change effects may be event catalysts, necessitates the development of dependable landslide susceptibility maps (LSM). Method: This study diverges from traditional binary classification models, framing LSM as a positive-unlabeled learning problem. This approach acknowledges that regions without recorded landslides are not necessarily at low risk but could simply have not experienced landslides yet. The study utilizes novel positive-unlabeled learning-enhanced algorithms—Random Forest, K-Nearest Neighbor, and Decision Tree—to create LSM for Chamoli district, India. Eleven causative factors for landslides are identified, including elevation, aspect, slope, geology, geomorphology, distance to lineament, lithology, NDVI, distance to river, distance to road and residential land use. To address spatial correlation biases, instead of randomly splitting the dataset, the study adopts spatial splitting to get the training and testing datasets. Conclusion: The study reveals that positive-unlabeled learning substantially improves the Area Under Curve and recall, leading to a more conservative LSM compared to binary classification methods. Analysis shows that the southern region of Chamoli exhibits high recall but lower accuracy, suggesting a latent high landslide susceptibility despite a lack of historical landslides in this region. The study also quantifies the impact of human activity on landslide risk, indicating an elevated threat to life and the local economy, especially in Chamoli's southwestern areas. [ABSTRACT FROM AUTHOR]

Malaria is a deadly disease that can take a person's life if not predicted or cured correctly. Numerous factors like temperature, humidity, precipitation, etc., impact India's increasing cases of malaria diseases. This research presents an advanced machine learning regression technique recently developed to anticipate the prevalence of malaria in Tripura using a real-world data. The proposed structure uses nine different regression methods, such as multilayer perceptron (MLP), random forest, support vector regressor, gradient boosting regressor, Bayesian ridge, kernel ridge, extreme gradient boost regressor (XGB), light gradient boosting machine (LGBM regressor and linear regression, to predict malaria using the most affecting factors of malaria diseases, namely temperature, humidity, precipitation, month, and years as input. Furthermore, to opt out the best suited technology for malaria cases prediction, the range of value method (ROVM)–multi-criteria decision methods (MCDM) technique has been applied, considering various statistical measurements as criteria. Ultimately, a comparison of various MCDM techniques reveals that MLP, XGB, and RF emerge as the top three choices. MLP regression, with root-mean-square error (RMSE) value of 0.03357, yields the lowest RMSE value, and the coefficient of determination (R2) is 0.97616, yielding the maximum among other regressions. To effectively battle the illness in Tripura, it could be useful for continuing intervention tactics by governmental, profit and nonprofit organizations. [ABSTRACT FROM AUTHOR]

Machine learning has been widely used to build intrusion detection models in detecting unknown attack traffic. How to train a model properly in order to attain the desired intrusion detection is an important topic. In contrast to offline learning, online learning proves more practical as it can update models simultaneously in the detecting process to comply with real network traffic. Active learning is an effective way to realize online learning. Among existing active learning mechanisms proposed to perform intrusion detection, most fail to meet the real online environment or to run persistently. This paper presents a new active online learning mechanism to secure better intrusion detection performance. The new mechanism advances related works in bringing the lifelong learning practice to fit in the online environment. It uses the efficient random forest (RF) as the detection model to train samples and adds a new tree to train a new batch of data when updating the model at each online stage, to pursue lifelong learning. By training a new batch of data only, it can keep the previously trained weights from being updated so as to preserve the past knowledge. Our mechanism is experimentally proved to yield better overall results than existing mechanisms: It produces superior training efficiency and detection performance—with the least training time, best training data quality and much reduced training data quantity. [ABSTRACT FROM AUTHOR]

In river research, forecasting flow velocity accurately in vegetated channels is a significant challenge. The forecasting performance of various independent and hybrid machine learning (ML) models are thus quantified for the first time in this work. Utilizing flow velocity measurements in both natural and laboratory flume experiments, we assess the efficacy of four distinct standalone machine learning techniques—Kstar, M5P, reduced error pruning tree (REPT) and random forest (RF) models. In addition, we also test for eight types of hybrid ML algorithms trained with an Additive Regression (AR) and Bagging (BA) (AR-Kstar, AR-M5P, AR-REPT, AR-RF, BA-Kstar, BA-M5P, BA-REPT and BA-RF). Findings from a comparison of their predictive capabilities, along with a sensitivity analysis of the influencing factors, indicated: (1) Vegetation height emerged as the most sensitive parameter for determining the flow velocity; (2) all ML models displayed outperforming empirical equations; (3) nearly all ML algorithms worked optimal when the model was built using all of the input parameters. Overall, the findings showed that hybrid ML algorithms outperform regular ML algorithms and empirical equations at forecasting flow velocity. AR-M5P (R2 = 0.954, R = 0.977, NSE = 0.954, MAE = 0.042, MSE = 0.003, and PBias = 1.466) turned out to be the optimal model for forecasting of flow velocity in vegetated-rivers. [ABSTRACT FROM AUTHOR]

Supply chain planning during disasters can be challenging due to uncertainty in demand and travel time, leading to limited stocks and delivery delays. While previous studies have focused on network planning for disaster relief supply chains under uncertainty, they have not fully integrated all network components while considering various potential factors. This integration is crucial for successful humanitarian relief operations. To address this issue, we propose a comprehensive model using a two-stage mixed-integer stochastic linear programming. The model incorporates facility location, pre-positioning, direct allocation, and multi-depot vehicle routing under demand and travel time uncertainties while examining multi-echelon, multi-commodity, response deadlines, and deprivation costs. We also create an improved random forest algorithm to enhance the accuracy of demand and travel time forecasts. To obtain accurate information for effective decision-making, we develop a data-driven, exact algorithm by combining an improved random forest algorithm and Benders decomposition. Computational experiments show that our proposed algorithm outperforms the L-shaped method in finding a better solution with less running time. We provide a real case to validate our model and algorithms. Our model and solution scheme can help improve efficiency and timeliness while minimizing deficiencies in disaster relief efforts. [ABSTRACT FROM AUTHOR]

The past few years have seen an emergence of interest in examining the significance of machine learning (ML) in the medical field. Diseases, health emergencies, and medical disorders may now be identified with greater accuracy because of technological advancements and advances in ML. It is essential especially to diagnose individuals with chronic diseases (CD) as early as possible. Our study has focused on analyzing ML's applicability to predict CD, including cardiovascular disease, diabetes, cancer, liver, and neurological disorders. This study offered a high-level summary of the previous research on ML-based approaches for predicting CD and some instances of their applications. To wrap things up, we compared the results obtained by various studies and the methodologies as well as tools employed by the researchers. The factors or parameters that are responsible for improving the accuracy of the predicting model for different previous works are also identified. For identifying significant features, most of the authors employed a variety of strategies, where least absolute shrinkage and selection (LASSO), minimal-redundancy-maximum-relevance (mRMR), and RELIEF are extensively used methods. It is seen that a wide range of ML approaches, including support vector machine (SVM), random forest (RF), decision tree (DT), naïve Bayes (NB), etc., have been widely used. Also, several deep learning techniques and hybrid models are employed to create CD prediction models, resulting in efficient and reliable clinical decision-making models. For the benefit of the whole healthcare system, we have also offered our suggestions for enhancing the prediction results of CD. [ABSTRACT FROM AUTHOR]

Predicting hotel ratings from reviews involves natural language processing techniques to extract sentiment and features from text data, then applying machine learning (ML) models like regression or classification to estimate the corresponding rating based on these features. This study proposes an ensemble learning approach for predicting hotel ratings from user reviews. By integrating multiple ML algorithms trained on various textual features, including linguistic, semantic, and sentiment-based attributes, our model achieves superior predictive accuracy. The primary aim of this study is to predict user experience ratings with high accuracy. To achieve this, this research utilizes an Ensemble learning approach known as majority voting to make these predictions effectively. In this investigation, initially, the dataset undergoes cleaning and is subsequently subjected to a series of pre-processing steps using Natural Language Processing (NLP) techniques. The research includes a comparative analysis of various classifiers along with different embedding methods. Seven different types of classifiers are used alongside three embedding techniques i.e., Term Frequency-Inverse Document Frequency (TF-IDF), Bag of Words (BoW), and Word2Vec. The classifiers include, Logistic Regression without Cross Validation (LR), LR with Cross Validation (LRCV), Decision Tree Classifier (DTC), Stochastic Gradient Descent Classifier (SGDC), Random Forest Classifier (RFC), Support Vector Classifier (SVC) and K-Nearest Neighbour (KNN). Our proposed methodology demonstrates higher accuracy and robust performance. Accuracy and frequency are utilized as performance metrics for assessing and validating both classifiers and embedding techniques. As per the simulation results, TF-IDF in combination with LRCV achieves an accuracy rate of 61%. [ABSTRACT FROM AUTHOR]

Mathematical programming techniques are widely used in the determination of optimal functional configuration of a wireless sensor network (WSN). But these techniques have usually high computational complexity and are often considered as Non Polynomial complete problem. Therefore, machine learning techniques can be utilized for the prediction of the WSN parameters with high accuracy and lesser computational complexity than the mathematical programming techniques. This paper focuses on developing the prediction model for determination of the node status to be included in the set cover based on the coverage probability and trust values of the nodes. The set covers are defined as the subset of nodes which are scheduled to monitor the region of interest with the desired coverage level. Several machine learning techniques have been used to determine the node activation status based on which the set covers are obtained. The results show that the random forest based prediction model yields the highest accuracy for the considered network setting. [ABSTRACT FROM AUTHOR]

Purpose: 123I-Ioflupane SPECT is an effective tool for the diagnosis and progression assessment of Parkinson's disease (PD). Radiomics and deep learning (DL) can be used to track and analyze the underlying image texture and features to predict the Hoehn-Yahr stages (HYS) of PD. In this study, we aim to predict HYS at year 0 and year 4 after the first diagnosis with combined imaging, radiomics and DL-based features using 123I-Ioflupane SPECT images at year 0. Methods: In this study, 161 subjects from the Parkinson's Progressive Marker Initiative database underwent baseline 3T MRI and 123I-Ioflupane SPECT, with HYS assessment at years 0 and 4 after first diagnosis. Conventional imaging features (IF) and radiomic features (RaF) for striatum uptakes were extracted from SPECT images using MRI- and SPECT-based (SPECT-V and SPECT-T) segmentations respectively. A 2D DenseNet was used to predict HYS of PD, and simultaneously generate deep features (DF). The random forest algorithm was applied to develop models based on DF, RaF, IF and combined features to predict HYS (stage 0, 1 and 2) at year 0 and (stage 0, 1 and ≥ 2) at year 4, respectively. Model predictive accuracy and receiver operating characteristic (ROC) analysis were assessed for various prediction models. Results: For the diagnostic accuracy at year 0, DL (0.696) outperformed most models, except DF + IF in SPECT-V (0.704), significantly superior based on paired t-test. For year 4, accuracy of DF + RaF model in MRI-based method is the highest (0.835), significantly better than DF + IF, IF + RaF, RaF and IF models. And DL (0.820) surpassed models in both SPECT-based methods. The area under the ROC curve (AUC) highlighted DF + RaF model (0.854) in MRI-based method at year 0 and DF + RaF model (0.869) in SPECT-T method at year 4, outperforming DL models, respectively. And then, there was no significant differences between SPECT-based and MRI-based segmentation methods except for the imaging feature models. Conclusion: The combination of radiomic and deep features enhances the prediction accuracy of PD HYS compared to only radiomics or DL. This suggests the potential for further advancements in predictive model performance for PD HYS at year 0 and year 4 after first diagnosis using 123I-Ioflupane SPECT images at year 0, thereby facilitating early diagnosis and treatment for PD patients. No significant difference was observed in radiomics results obtained between MRI- and SPECT-based striatum segmentations for radiomic and deep features. [ABSTRACT FROM AUTHOR]

In this study, optical photothermal infrared (O-PTIR) spectroscopy combined with machine learning algorithms were used to evaluate 46 tissue cores of surgically resected cervical lymph nodes, some of which harboured oral squamous cell carcinoma nodal metastasis. The ratios obtained between O-PTIR chemical images at 1252 cm−1 and 1285 cm−1 were able to reveal morphological details from tissue samples that are comparable to the information achieved by a pathologist's interpretation of optical microscopy of haematoxylin and eosin (H&E) stained samples. Additionally, when used as input data for a hybrid convolutional neural network (CNN) and random forest (RF) analyses, these yielded sensitivities, specificities and precision of 98.6 ± 0.3%, 92 ± 4% and 94 ± 5%, respectively, and an area under receiver operator characteristic (AUC) of 94 ± 2%. Our findings show the potential of O-PTIR technology as a tool to study cancer on tissue samples. [ABSTRACT FROM AUTHOR]

Purpose: To investigate and implement semiautomated screening for meta-analyses (MA) in urology under consideration of class imbalance. Methods: Machine learning algorithms were trained on data from three MA with detailed information of the screening process. Different methods to account for class imbalance (Sampling (up- and downsampling, weighting and cost-sensitive learning), thresholding) were implemented in different machine learning (ML) algorithms (Random Forest, Logistic Regression with Elastic Net Regularization, Support Vector Machines). Models were optimized for sensitivity. Besides metrics such as specificity, receiver operating curves, total missed studies, and work saved over sampling were calculated. Results: During training, models trained after downsampling achieved the best results consistently among all algorithms. Computing time ranged between 251 and 5834 s. However, when evaluated on the final test data set, the weighting approach performed best. In addition, thresholding helped to improve results as compared to the standard of 0.5. However, due to heterogeneity of results no clear recommendation can be made for a universal sample size. Misses of relevant studies were 0 for the optimized models except for one review. Conclusion: It will be necessary to design a holistic methodology that implements the presented methods in a practical manner, but also takes into account other algorithms and the most sophisticated methods for text preprocessing. In addition, the different methods of a cost-sensitive learning approach can be the subject of further investigations. [ABSTRACT FROM AUTHOR]

Historical architecture has different special styles attributed to each era, dynasty, or region. These styles are common features such as geometric properties, ratios, scales, colors, and artistic techniques. Historical geometric ornaments have an enormous capability for classification based on their geometric characteristics. Smart pattern recognition allows researchers to classify huge databases of heritage for useful internet searches. So, our main goal in this paper is to implement the detection of categories in geometric patterns for classification and documentation in which by the photography of ornaments in every monument, the type of patterns and the number of every type of pattern would be estimated quickly. Furthermore, due to occurring deterioration in these patterns, our method also contributes to recognizing the deteriorated patterns. When we encounter numerous pieces of deteriorated patterns, manual recognition in order to reassemble and reconstruct is usually impossible or time-consuming. With the aid of artificial intelligence, in this paper, our aim is to seek to solve the automatic recognition of historical geometric patterns, even patterns having deterioration as an occlusion via image processing and machine learning methods. A challenging issue that researchers would tackle in detecting historical geometric pattern's types is the variety in geometric textures, especially when they have occlusion such as deterioration. This issue leads to limited success in classifying via extracting only one feature. The other issue is that the extracted feature must be invariant to the transformation, such as scale, rotation, and noise variation. To cope with the challenges mentioned above and accurately classify, we plan to use the fusion method based on extracting global and local features. So, the features extracted from images in this research are based on local and global. In other words, the proposed fusion strategy lies both in feature and decision level, but the core is the proposed three combination methods in fusion decision methods. In this method, the dataset is composed of four main Persian geometric pattern types: Tond dah, Kond tablghenas, Hashtva 4 lengeh, and HashtvatablKond. So, the model will be trained by extracting global and local features of the images separately. Random forest, as the prevalent machine learning algorithm, is proposed for training data and predicting the class of input images. Finally, the probability of prediction for random forest classifiers is fused by the Decision Templates (DT) combiner, Naïve Bayes (NB) combiner, and Dempster–Shafer combination methods. In comparison with the individual classifier accuracy results of 80% and 85% for global and local features respectively, our proposed approach achieves an improved accuracy of 90%, 88%, and 90% in three fusion decision methods including DT combiner, NB combiner, and Dempster–Shafer combination methods. [ABSTRACT FROM AUTHOR]

Image recognition and classification in the domain of cultural heritage is a complex task that usually requires good image quality and a large dataset. Small dataset is an apparent problem when data availability is limited, often resulting in poor classification performance. The aim of this study is to analyze and compare the performance of four machine learning algorithms, namely random forest, multilayer perceptron classifier, Naïve Bayes, and decision tree, that are used to classify features extracted using eleven pre-trained deep learning architectures from a small set of images representing cultural heritage. The findings imply that random forest and multilayer perceptron classifiers are the most appropriate for the task of image classification of small cultural heritage datasets, as they obtained the highest performance compared to the other two algorithms, particularly when DenseNet121, EfficientNetB0 and NASNetMobile architectures are used for feature extraction. In addition, for the classification of features extracted using NASNetMobile all the four utilized machine learning algorithms obtained high accuracy ranging from 88.89 to 95.56%. [ABSTRACT FROM AUTHOR]

ALS (Amyotrophic Lateral Sclerosis) is a fatal neurodegenerative disease of the human motor system. It is a group of progressive diseases that affects the nerve cells in the brain and spinal cord that control the muscle movement of the body hence, detection and classification of ALS at the right time is considered to be one of the vital aspects that can save the life of humans. Therefore, in various studies, different AI techniques are used for the detection of ALS, however, these methods are considered to be ineffectual in terms of identifying the disease due to the employment of ineffective algorithms. Hence, the proposed model utilizes Modified Principal Component Analysis (MPCA) and Modified Random Forest (MRF) for performing dimensionality reduction of all the potential features considered for effective classification of the ALS presence and absence of ALS causing mutation in the corresponding gene. The MPCA is adapted for capturing all the Low-Importance Data transformation. Furthermore, The MPCA is objected to performing three various approaches: Covariance Matrix Correlation, Eigen Vector- Eigenvalue decomposition, and selecting the desired principal components. This is done in aspects of implying the LI (Lower-Importance) Data Transformation. By choosing these potential components without any loss of features ensures better viability of selecting the attributes for ALS-causing gene classification. This is followed by the classification of the proposed model by using Modified RF by updating the clump detector technique. The clump detector is proceeded by clustering approach using K-means, and the data reduced by their dimension are grouped accordingly. These clustered data are analyzed either for ALS causing or devoid of causing ALS. Finally, the model's performance is assessed using different evaluation metrics like accuracy, recall, F1 score, and precision, and the proposed model is further compared with the existing models to assess the efficacy of the proposed model. [ABSTRACT FROM AUTHOR]

In the context of contemporary urbanization, the significance of diversifying urban public facilities has attracted significant attention. This study examines the relationship between road network and the diversity of urban public facilities in Changsha City. These factors are measured through the Shannon-Weiner Index and Space Syntax method. To provide greater specificity, the study employs Random Forest and Geographically Weighted Regression models to analyze the relationship between road networks and the diversity of urban public facilities. The results identify a nonlinear relationship between these variables. In addition, high accessibility exhibits a stronger association with diversity than accessibility alone, and this correlation is reflected in varying degrees of inconsistency across different study areas. Finally, the cartographic depiction of diversity clusters is overlaid on the road network, demonstrating a significant relationship between the configuration of the road network and diversity patterns. In conclusion, this research emphasizes the robust correlation between the roadway network and the diversity of urban public facilities. It prompts local governments to focus beyond mere equality, with a greater commitment to enhancing the quality of life for residents through diversity. [ABSTRACT FROM AUTHOR]

In healthcare, observing the features and areas of malaria in microscopic images is crucial for the diagnosis and treatment of plasmodium malaria parasites for automated detection. The classification of malaria parasites can be challenging due to factors such as limited training data, class imbalance, and variability in parasite stages. Moreover, the parasites often belong to similar classes, and the quality of malaria blood cell smears is often poor. To address this issue, a Deep Convolutional Neural Network (Deep-CNN) model is combined with Random Forest (RF) for the detection of plasmodium malaria parasites. The Deep-CNN-RF approach leverages domain-specific expertise for improved learning. Moreover, a novel modification is applied to the pooling layer by incorporating the Global Average-Pooling layer (GAP) without an additional flattening layer, enhancing parasite area visualization. Additionally, Canny edge detection is employed for precise parasite boundary edge detection. The experimental performance of the Deep-CNN-RF model is evaluated using a malaria cell dataset from the National Library of Medicine (NLM), Kaggle, and the National Institutes of Health (NIH). The proposed scheme effectiveness is assessed through performance metrics, including accuracy, precision, recall, Mean Square Error (MSE), and F1-Score. The results demonstrate that the proposed technique outperforms state-of-the-art classification models on malaria parasite datasets. [ABSTRACT FROM AUTHOR]

Groundwater in northwestern parts of Bangladesh, mainly in the Chapainawabganj District, has been contaminated by arsenic. This research documents the geographical distribution of arsenic concentrations utilizing machine learning techniques. The study aims to enhance the accuracy of model predictions by precisely identifying occurrences of groundwater arsenic, enabling effective mitigation actions and yielding more beneficial results. The reductive dissolution of arsenic-rich iron oxides/hydroxides is identified as the primary mechanism responsible for the release of arsenic from sediment into groundwater. The study reveals that in the research region, alongside elevated arsenic concentrations, significant levels of sodium (Na), iron (Fe), manganese (Mn), and calcium (Ca) were present. Statistical analysis was employed for feature selection, identifying pH, electrical conductivity (EC), sulfate (SO4), nitrate (NO3), Fe, Mn, Na, K, Ca, Mg, bicarbonate (HCO3), phosphate (PO4), and As as features closely associated with arsenic mobilization. Subsequently, various machine learning models, including Naïve Bayes, Random Forest, Support Vector Machine, Decision Tree, and logistic regression, were employed. The models utilized normalized arsenic concentrations categorized as high concentration (HC) or low concentration (LC), along with physiochemical properties as features, to predict arsenic occurrences. Among all machine learning models, the logistic regression and support vector machine models demonstrated high performance based on accuracy and confusion matrix analysis. In this study, a spatial distribution prediction map was generated to identify arsenic-prone areas. The prediction map also displays that Baroghoria Union and Rajarampur region under Chapainawabganj municipality are high-risk areas and Maharajpur Union and Baliadanga Union are comparatively low-risk areas of the research area. This map will facilitate researchers and legislators in implementing mitigation strategies. Logistic regression (LR) and support vector machine (SVM) models will be utilized to monitor arsenic concentration values continuously. [ABSTRACT FROM AUTHOR]

Existing research on non-suicidal self-injury (NSSI) among adolescents has primarily concentrated on general risk factors, leaving a significant gap in understanding the specific NSSI characteristics that predict diverse psychopathological outcomes. This study aims to address this gap by using Random Forests to discern the significant predictors of different clinical outcomes. The study tracked 348 adolescents (64.7% girls; mean age = 13.31, SD = 0.91) over 6 months. Initially, 46 characteristics of NSSI were evaluated for their potential to predict the repetition of NSSI, as well as depression, anxiety, and suicidal risks at a follow-up (T2). The findings revealed distinct predictors for each psychopathology. Specifically, psychological pain was identified as a significant predictor for depression, anxiety, and suicidal risks, while the perceived effectiveness of NSSI was crucial in forecasting its repetition. These findings imply that it is feasible to identify high-risk individuals by assessing key NSSI characteristics, and also highlight the importance of considering diverse NSSI characteristics when working with self-injurers. [ABSTRACT FROM AUTHOR]

Social interactions are essential for well-being. Therefore, researchers increasingly attempt to capture an individual's social context to predict well-being, including mood. Different tools are used to measure various aspects of the social context. Digital phenotyping is a commonly used technology to assess a person's social behavior objectively. The experience sampling method (ESM) can capture the subjective perception of specific interactions. Lastly, egocentric networks are often used to measure specific relationship characteristics. These different methods capture different aspects of the social context over different time scales that are related to well-being, and combining them may be necessary to improve the prediction of well-being. Yet, they have rarely been combined in previous research. To address this gap, our study investigates the predictive accuracy of mood based on the social context. We collected intensive within-person data from multiple passive and self-report sources over a 28–day period in a student sample (Participants: N = 11, ESM measures: N = 1313). We trained individualized random forest machine learning models, using different predictors included in each model summarized over different time scales. Our findings revealed that even when combining social interactions data using different methods, predictive accuracy of mood remained low. The average coefficient of determination over all participants was 0.06 for positive and negative affect and ranged from − 0.08 to 0.3, indicating a large amount of variance across people. Furthermore, the optimal set of predictors varied across participants; however, predicting mood using all predictors generally yielded the best predictions. While combining different predictors improved predictive accuracy of mood for most participants, our study highlights the need for further work using larger and more diverse samples to enhance the clinical utility of these predictive modeling approaches. [ABSTRACT FROM AUTHOR]

In the present study, for the first time the machine learning (ML) based refractive index (n) approach is established depends on the density (ρ) parameter of glasses for a dataset of 2000 oxide glasses to predict refractive index of B2O3-CaO-Li2O-Sb2O3 glasses. Density of the investigated glasses varied from 2.56 to 2.97 gm/cm3. The corresponding refractive index was changed from 2.540 to 2.405. The refractive index prediction based on density parameter derived from the density of glasses and constant 'K'. For all M-L techniques including gradient descent (GD), artificial neural network (ANN), and random forest regression (RFR), the density factor is used as an independent variable and the experimental refractive index as a dependent variable. The data set of 10,000 oxide glass samples was employed to forecast density using a variety of machine learning approaches. In comparison to other models, the Random forest regression (RFR) model fitted the glass data with the highest R2 value of 0.949 for refractive index prediction and 0.925 for density prediction. For both the prediction of density and refractive index, the R2 is controlled to 0.932 and 0.9223, respectively. The highest R2 values for refractive index and density prediction were gained when the tanh activation function was used in an artificial neural network (ANN) with varied activation functions. [ABSTRACT FROM AUTHOR]

The present study introduces a novel approach utilizing machine learning techniques to predict the crucial mechanical properties of engineered cementitious composites (ECCs), spanning from typical to exceptionally high strength levels. These properties, including compressive strength, flexural strength, tensile strength, and tensile strain capacity, can not only be predicted but also precisely estimated. The investigation encompassed a meticulous compilation and examination of 1532 datasets sourced from pertinent research. Four machine learning algorithms, linear regression (LR), K nearest neighbors (KNN), random forest (RF), and extreme gradient boosting (XGB), were used to establish the prediction model of ECC mechanical properties and determine the optimal model. The optimal model was utilized to employ SHapley Additive exPlanations (SHAP) for scrutinizing feature importance and conducting an in-depth parametric analysis. Subsequently, a comprehensive control strategy was devised for ECC mechanical properties. This strategy can provide actionable guidance for ECC design, equipping engineers and professionals in civil engineering and material science to make informed decisions throughout their design endeavors. The results show that the RF model demonstrated the highest prediction accuracy for compressive strength and flexural strength, with R2 values of 0.92 and 0.91 on the test set. The XGB model outperformed in predicting tensile strength and tensile strain capacity, with R2 values of 0.87 and 0.80 on the test set, respectively. The prediction of tensile strain capacity was the least accurate. Meanwhile, the MAE of the tensile strain capacity was a mere 0.84%, smaller than the variability (1.77%) of the test results in previous research. Compressive strength and tensile strength demonstrated high sensitivity to variations in both water-cement ratio (W) and water reducer (WR). In contrast, flexural strength exhibited high sensitivity solely to changes in W. Conversely, the sensitivity of tensile strain capacity to input features was moderate and consistent. The mechanical attributes of ECC emerged from the combined effects of multiple positive and negative features. Notably, WR exerted the most significant influence on compressive strength among all features, whereas polyethylene (PE) fiber emerged as the primary driver affecting flexural strength, tensile strength, and tensile strain capacity. [ABSTRACT FROM AUTHOR]

Metastatic gastric cancer (GC) presents significant clinical challenges due to its poor prognosis and limited treatment options. To address this, we conducted a targeted protein biomarker discovery study to identify markers predictive of metastasis in advanced GC (AGC). Serum samples from 176 AGC patients (T stage 3 or higher) were analyzed using the Olink Proteomics Target panels. Patients were retrospectively categorized into nonmetastatic, metastatic, and recurrence groups, and differential protein expression was assessed. Machine learning and gene set enrichment analysis (GSEA) methods were applied to discover biomarkers and predict prognosis. Four proteins (MUC16, CAIX, 5'-NT, and CD8A) were significantly elevated in metastatic GC patients compared to the control group. Additionally, GSEA indicated that the response to interleukin-4 and hypoxia-related pathways were enriched in metastatic patients. Random forest classification and decision-tree modeling showed that MUC16 could be a predictive marker for metastasis in GC patients. Additionally, ELISA validation confirmed elevated MUC16 levels in metastatic patients. Notably, high MUC16 levels were independently associated with metastatic progression in T3 or higher GC. These findings suggest the potential of MUC16 as a clinically relevant biomarker for identifying GC patients at high risk of metastasis. [ABSTRACT FROM AUTHOR]

Rockburst is a mine dynamic disaster caused by the rapid release of elastic strain energy of surrounding rock. As the depth of engineering project operations increases, accurate classification of rockburst intensity cannot be achieved based on conventional criteria due to high uncertainty and unpredictability of rockburst. In this regard, an AOA-Voting-Soft ensemble machine learning was proposed in this study by combining seven individual classifiers, i.e., eXtreme gradient boosting, support vector machines, multilayer perceptron, k-nearest neighbor, random forest, naive Bayesian, and gradient boosting decision Tree. In addition, outliers were eliminated by means of density-based spatial clustering of applications with noise, and CURE-MeanradiusSMOTE was adopted to obtain a balanced data structure. Furthermore, the optimal combination of classifiers in Voting was determined by the game theory and the exhaustive search method. Weights of individual learners in Voting were determined through the arithmetic optimization algorithm and fivefold cross-validation. The results show that the prediction accuracy of the ensemble algorithm proposed in this study is 4.4% higher than that of the individual classifier with optimal performance. The importance analysis indicates that the elastic energy index is the most important variable that affects rockburst intensity grades. Moreover, this rockburst ensemble method can be applied further to solve other classification problems in underground engineering projects. Highlights: This study improves the data preprocessing method, outliers were eliminated by means of density-based spatial clustering of applications with noise, and CURE-MeanradiusSMOTE was proposed to obtain a balanced data structure. This study presents a hybrid ensemble model for Rockburst intensity grade prediction, combining a new metaheuristic method with the Voting-Soft model. This study combines game theory and method of exhaustion to determine the best classifier combination in voting. The weights of individual learners in Voting were determined through arithmetic optimization algorithm and fivefold cross-validation. Sensitivity study was conducted on input variables with RBD-FAST, and the results suggest that W et is the most important input variable. [ABSTRACT FROM AUTHOR]

Background: Anastomotic leakage (AL), a severe complication following colorectal surgery, arises from defects at the anastomosis site. This study evaluates the feasibility of predicting AL using machine learning (ML) algorithms based on preoperative data. Methods: We retrospectively analyzed data including 21 predictors from patients undergoing colorectal surgery with bowel anastomosis at four Swiss hospitals. Several ML algorithms were applied for binary classification into AL or non-AL groups, utilizing a five-fold cross-validation strategy with a 90% training and 10% validation split. Additionally, a holdout test set from an external hospital was employed to assess the models' robustness in external validation. Results: Among 1244 patients, 112 (9.0%) suffered from AL. The Random Forest model showed an AUC-ROC of 0.78 (SD: ± 0.01) on the internal test set, which significantly decreased to 0.60 (SD: ± 0.05) on the external holdout test set comprising 198 patients, including 7 (3.5%) with AL. Conversely, the Logistic Regression model demonstrated more consistent AUC-ROC values of 0.69 (SD: ± 0.01) on the internal set and 0.61 (SD: ± 0.05) on the external set. Accuracy measures for Random Forest were 0.82 (SD: ± 0.04) internally and 0.87 (SD: ± 0.08) externally, while Logistic Regression achieved accuracies of 0.81 (SD: ± 0.10) and 0.88 (SD: ± 0.15). F1 Scores for Random Forest moved from 0.58 (SD: ± 0.03) internally to 0.51 (SD: ± 0.03) externally, with Logistic Regression maintaining more stable scores of 0.53 (SD: ± 0.04) and 0.51 (SD: ± 0.02). Conclusion: In this pilot study, we evaluated ML-based prediction models for AL post-colorectal surgery and identified ten patient-related risk factors associated with AL. Highlighting the need for multicenter data, external validation, and larger sample sizes, our findings emphasize the potential of ML in enhancing surgical outcomes and inform future development of a web-based application for broader clinical use. [ABSTRACT FROM AUTHOR]

In this study, an artificial neural network (ANN) model was developed based on molecular descriptors to predict the surface tension of liquids. A dataset containing various features was constructed by collecting experimental data from 25 different fluids and extracting molecular structural descriptors. Feature selection was performed using the forward search wrapper method based on Random Forest, identifying 7 significant features (Temperature, MinAbsEStateIndex, LabuteASA, MolMR, Chi1v, qed and FpDensityMorgan3) for surface tension prediction. Subsequently, an ANN model was constructed with the selected features as inputs to predict the surface tension of liquids. The derived model demonstrates high accuracy with a correlation coefficient (R) exceeding 0.999 and a notably low mean square error (MSE = 1.843e−5). Moreover, the ANN model exhibited a total average absolute deviation (AAD) of 0.98 %, comparable to that of the REFPROP, which had a total AAD of 1.26 %. This quantitative model serves an easy tool for gaining insights into the molecular underpinnings of surface tension and predicting its value across various fluids. [ABSTRACT FROM AUTHOR]

The present study analyzes morphological differences in the pelvis of Japanese and Western Australian individuals and investigates the feasibility of population affinity classification based on computed tomography (CT) data. The Japanese and Western Australian samples comprise CT scans of 207 (103 females; 104 males) and 158 (78 females; 80 males) adult individuals, respectively. Following volumetric reconstruction, a total of 19 pelvic landmarks were obtained on each sample, and 11 measurements, including two angles, were calculated. Machine learning methods (random forest modeling [RFM] and support vector machine [SVM]) were used to classify population affinity. Classification accuracy of the two-way models was approximately 80% for RFM: the two-way sex-specific and sex-mixed models for SVM achieved > 90% and > 85%, respectively. The sex-specific models had higher accurate classification rates than the sex-mixed models, except for the Japanese male sample. The classification accuracy of the four-way sex and population affinity model had an overall classification accuracy of 76.71% for RFM and 87.67% for SVM. All the correct classification rates were higher in the Japanese relative to the Western Australian sample. Our data suggest that pelvic morphology is sufficiently distinct between Japanese and Western Australian individuals to facilitate the accurate classification of population affinity based on measurements acquired in CT images. To the best of our knowledge, this is the first study investigating the feasibility of population affinity estimation based on CT images of the pelvis, which appears as a viable supplement to traditional approaches based on cranio-facial morphology. [ABSTRACT FROM AUTHOR]

Background and aim: Seed inoculation with effective rhizobia (Rh) strains and phosphorus (P) application are sustainable agricultural practices that enhance soybean grain yield. However, adoption of these practices in northern Ghana is hindered by unpredictable yield responses and poor understanding of the factors that cause yield variation. We compared soybean yield from inputs managed trials such as Rh inoculants and P sources. Methods: Soybean trials were conducted on-station and on-farm with one of three Rh inoculants [Bradyrhizobium japonicum (Bj), B. elkanii (Be), and B. diazoefficiens (Bd)], and with one of three P sources [no-P, Triple Superphosphate (TSP), and a mixture of rock-P (low solubility) and TSP (water-soluble P)]. Yield was compared at harvest. The random forest (RF) model was used to predict yield, and factors of yield variation were determined using forward redundancy analysis (RDA). Results: The soybean yield was consistently higher in the Rh × P treatments compared to the control in on-station and on-farm experiments. The treatments that received P fertilizer (RP and TSP) showed a higher grain yield than the control. The on-station experiment yielded more accurate results than the on-farm dataset, with a trained coefficient of determination (R2) of 0.95 and 0.87, respectively. The yield variation was explained by the Rh × P fertilizer, P sources, Rh strains, and exchangeable soil Mg. Conclusion: The exchangeable magnesium in the soil contributed significantly to the variation in yield. Further research is recommended to study the mechanisms that underline the impact of magnesium on soybean growth. [ABSTRACT FROM AUTHOR]

Project risk management which has been rarely considered, especially in software projects, is a crucial process to complete projects successfully. This paper aims to propose a novel project risk management framework both to evaluate the project risks and to predict the success or failure of software projects based on their risk. This new framework uses Machine Learning (ML) and Multi-Attribute Decision-Making under a cloud model to effectively manage uncertainty. Based on the proposed framework, in the first stage, the important risks of the software projects are identified by an organized approach. Then, the risks are evaluated based on their probability and impacts on time, cost, and quality. In the second stage, the obtained results of the previous stage are entered into a new COPRAS method under the cloud model to rank the risks. Then, the risks are classified into various groups according to their rank. It helps project managers to gain a profound awareness of their high-priority project risks. In this paper, data on risks for fifty software projects has been collected. All the steps of the second stage are implemented on these projects in order to assess their risks. As a result, a data set whose features are nine types of software project risks and the label of success or failure of the projects is created. To recognize the pattern between risks' values and the success or failure of the projects, various efficient ML algorithms such as Naive Bayes, Logistic regression, Decision Tree, Bagging, Random Forest, and AdaBoost are applied. This framework can predict the success or failure of software projects based on their risks with good accuracy. The results depict that the Naive Bayes algorithm has the best results compared to others. [ABSTRACT FROM AUTHOR]

The current study aims to investigate the applicability of ensemble modeling in improving the simulation of water surface evaporation (EW) in the Three Gorges Reservoir. To achieve this objective, a sensitivity analysis is performed to determine the most influential model inputs. Various models are employed for the simulation of EW, including empirical models such as Stelling (STE), Thornthwaite Holzman (T-H), and Ryan Harleman (R-H); statistical models including multiple-linear regression (MLR), Ridge regression (Ridge), and Lasso regression (Lasso); and different Artificial Intelligence (AI) techniques such as Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and General Regression Neural Network (GRNN). The performance of these models is evaluated. Additionally, three ensemble methods, namely simple averaging, weighted averaging, and neural ensemble, are utilized in strategy 1 for empirical models, strategy 2 for statistical models, strategy 3 for AI models, and strategy 4 for multi-class mixed models, with the aim of improving the simulation performance. The results indicate that the dominant parameters are water surface temperature, water surface area, relative humidity, temperature difference of vapor, and wind speed. For the single model, empirical and statistical models can yield valuable results, while most AI models suffer from overfitting issues. Among the ensemble models, the neural ensemble method outperforms the simple averaging and weighted averaging methods. The multi-class mixed ensemble model exhibits the highest simulation accuracy, with NSE values of 0.95 and 0.86 in the training and validation phases, respectively. Compared to the best single model, the ensemble approaches proposed in this study improve the performance of single models in the validation phase by up to 11.63%, 8.21%, 6.88%, and 6.96% for strategies 1 ~ 4, respectively. Furthermore, the results demonstrate that the multi-class mixed ensemble modeling approach is preferable over empirical, statistical, and AI ensemble modeling. [ABSTRACT FROM AUTHOR]

Urbanization is a complex phenomenon that has a profound impact on natural resources, economies, climate dynamics, and community structures. This study sets out to analyze the effects of urban sprawl on key urban regions in Bangladesh, covering urban, suburban, and rural landscapes. We used Landsat images from 1992 to 2022, processed and classified using the Google Earth Engine and Random Forest algorithm. Spatial metrics, Tupu method, ANOVA, and spatial regression models helped us quantify and explain land cover changes, their drivers, and consequences. Our findings show that the total built-up area in the five districts expanded from 1,200 km2 in 1992 to 2,800 km2 in 2022, with an overall yearly expansion rate of 3.5%. Meanwhile, vegetation cover declined from 2,500 km2 in 1992 to 1,500 km2 in 2022, with an average yearly loss rate of 2.5%. We identified population growth, economic development, and infrastructure expansion as the main drivers of urbanization, while land degradation, air and water pollution, and loss of biodiversity were the main consequences. Urbanization also had a significant impact on environmental functions, including carbon sequestration, water flow control, and urban heat mitigation, with varying effects across different urban areas and land use types. For instance, we found that carbon storage capacity decreased by 20% in urban areas, while it increased by 10% in rural areas. Our research highlights the crucial role of preserving and enhancing green spaces, including vegetation cover and water bodies, to mitigate urban heat and promote urban resilience. The findings underscore the need for sustainable urbanization, balancing economic growth with environmental protection and climate resilience, using geospatial and remote sensing techniques. [ABSTRACT FROM AUTHOR]

Purpose: Posterior circulation ischemic stroke (PCIS) possesses unique features. However, previous studies have primarily or exclusively relied on anterior circulation stroke cases to build machine learning (ML) models for predicting onset time. To date, there is no research reporting the effectiveness and stability of ML in identifying PCIS onset time. We aimed to build diffusion-weighted imaging-based ML models to identify the onset time of PCIS patients. Methods: Consecutive PCIS patients within 24 h of definite symptom onset were included (112 in the training set and 49 in the independent test set). Images were processed as follows: volume of interest segmentation, image feature extraction, and feature selection. Five ML models, naïve Bayes, logistic regression, tree ensemble, k-nearest neighbor, and random forest, were built based on the training set to estimate the stroke onset time (binary classification: ≤ 4.5 h or > 4.5 h). Relative standard deviations (RSD), receiver operating characteristic (ROC) curves, and the calibration plot was performed to evaluate the stability and performance of the five models. Results: The random forest model had the best performance in the test set, with the highest area under the curve (AUC, 0.840; 95% CI: 0.706, 0.974). This model also achieved the highest accuracy, sensitivity, specificity, positive predictive value, and negative predictive value (83.7%, 64.3%, 91.4%, 75.0%, and 86.5%, respectively). Furthermore, the model had high stability (RSD = 0.0094). Conclusion: The PCIS case-based ML model was effective for estimating the symptom onset time and achieved considerably high specificity and stability. [ABSTRACT FROM AUTHOR]

Variations of picoplankton groups were investigated over a one-month period in Daya Bay and Sanya Bay, in the northern South China Sea. The two coastal regions exhibited different variation patterns in physicochemical parameters. Moreover, the diel variations of picoplankton groups were different between the two bays. The abundance of the picoplankton in Sanya Bay displayed a pronounced diel variation, while it was not significant in Daya Bay. In addition, some similar patterns of picoplankton abundance were discovered. In the two bays, virioplankton exhibited the smallest fluctuation range, whereas picocyanobacteria fluctuated most markedly. The fluctuation range of picoplankton groups was larger in spring tide than in neap tide, especially in Sanya Bay. Random forest model analysis demonstrated that the variation of picoplankton groups was attributed to physical and chemical factors in Sanya Bay and Daya Bay, respectively. Therefore, our findings suggest that virioplankton abundance can persist more stably in response to changing environmental conditions compared to bacterioplankton and picophytoplankton. [ABSTRACT FROM AUTHOR]

Background: Treatment and prevention of intracranial hypertension (IH) to minimize secondary brain injury are central to the neurocritical care management of traumatic brain injury (TBI). Predicting the onset of IH in advance allows for a more aggressive prophylactic treatment. This study aimed to develop random forest (RF) models for predicting IH events in TBI patients. Methods: We analyzed prospectively collected data from patients admitted to the intensive care unit with invasive intracranial pressure (ICP) monitoring. Patients with persistent ICP > 22 mmHg in the early postoperative period (first 6 h) were excluded to focus on IH events that had not yet occurred. ICP-related data from the initial 6 h were used to extract linear (ICP, cerebral perfusion pressure, pressure reactivity index, and cerebrospinal fluid compensatory reserve index) and nonlinear features (complexity of ICP and cerebral perfusion pressure). IH was defined as ICP > 22 mmHg for > 5 min, and severe IH (SIH) as ICP > 22 mmHg for > 1 h during the subsequent ICP monitoring period. RF models were then developed using baseline characteristics (age, sex, and initial Glasgow Coma Scale score) along with linear and nonlinear features. Fivefold cross-validation was performed to avoid overfitting. Results: The study included 69 patients. Forty-three patients (62.3%) experienced an IH event, of whom 30 (43%) progressed to SIH. The median time to IH events was 9.83 h, and to SIH events, it was 11.22 h. The RF model showed acceptable performance in predicting IH with an area under the curve (AUC) of 0.76 and excellent performance in predicting SIH (AUC = 0.84). Cross-validation analysis confirmed the stability of the results. Conclusions: The presented RF model can forecast subsequent IH events, particularly severe ones, in TBI patients using ICP data from the early postoperative period. It provides researchers and clinicians with a potentially predictive pathway and framework that could help triage patients requiring more intensive neurological treatment at an early stage. [ABSTRACT FROM AUTHOR]

Introduction: Previous studies have explored the connections between various ocular biological parameters with myopia. Our previous study also found that pupil data can predict the myopic progression during the interventions for myopia. However, studies exploring the association between pupil diameter and myopia in preschoolers with myopia were lacking. Hence this study was aimed to investigate the association between pupil diameter and myopia in preschoolers with myopia based on a real-world, large-scale dataset. Methods: Data containing 650,671 preschoolers were collected from a total of 1943 kindergartens in Shenzhen, China. Refraction and pupil parameters were collected. After data filtering, the occurrence of myopia and its association with age, gender, pupil diameter, and other variables, were analyzed. Random forest (RF) and eXtreme gradient boosting (XGBoost) were selected from seven machine learning algorithms to build the model. The mean decrease accuracy (MDA), mean decrease Gini (MDG), and gain feature importance (GFI) techniques were employed to quantify the importance of pupil diameter and other features. Results: After the assessments, 51,325 valid records with complete pupil data were included, and 3468 (6.76%) were identified as myopia based on the calculated cycloplegic refraction. Preschoolers with myopia presented reduced pupil diameter and greater variation (5.00 ± 0.99 mm) compared to non-myopic preschoolers (6.22 ± 0.67 mm). A nonlinear relationship was found according to the scatterplots between pupil diameter and refraction (R2 = 0.14). Especially preschoolers with myopia had reduced pupil diameter compared to emmetropic preschoolers, but hyperope did not experience additional pupil enlargement. After adjusting for other covariates, this relationship is still consistent (P < 0.001). XGBoost and RF algorithms presented the highest performance and validated the importance of pupil diameter in myopia. Conclusions: Based on a real-world large-scale dataset, the current study illuminated that preschoolers with myopia had a reduced pupil diameter compared to emmetropic preschoolers with a nonlinear pattern. Machine learning algorithms visualized and validated the pivotal role of pupil diameter in myopia. Trial Registration: chictr.org Identifier: ChiCTR2200057391. [ABSTRACT FROM AUTHOR]

Introduction: The study aims to demonstrate and estimate the prevalence of clinical corneal ectasia and keratoconus (KC) in patients with relatively low keratometry (low-K KC). Methods: In a retrospective, analytical, and non-interventionist study, one eye was randomly selected from 1054 patients from the original Tomographic Biomechanical Index (TBIv1) study and the external validation (from Rio de Janeiro, Brazil, and Milan, Italy clinics). Patients were stratified into three groups. Group 1 included 736 normal patients, and groups 2 and 3 included 318 patients with clinical KC in both eyes, divided into low-K KC (90 patients) and high-K KC (228 patients), respectively. All patients underwent a comprehensive ophthalmological evaluation along with Pentacam and Corvis ST (Oculus, Wetzlar, Germany) examinations. Cases with maximum mean zone 3 mm keratometry (Kmax zone mean 3 mm) lower than 47.6 diopters (D) were considered as low-keratometry keratoconus, and cases with Kmax zone mean 3 mm higher than 47.6 D were regarded as high-keratometry keratoconus. Results: Ninety (28.30%) of the 318 KC group presented ectasia with low-keratometric values (low-Kmax). The average age in the normal group was 39.28 years (range 6.99–90.12), in the low-Kmax KC group it was 37.49 (range 13.35–78.45), and in the high-Kmax KC group it was 34.22 years (range 12.7–80.34). Mean and SD values and median (range), respectively, of some corneal tomographic and biomechanical parameters evaluated from the low-Kmax KC group were as follows: Belin-Ambrósio enhanced ectasia display (BAD-D) 3.79 ± 1.62 and 3.66 (0.83–9.73); Pentacam random forest index (PRFI) 0.78 ± 0.25 and 0.91 (0.05–1); corneal biomechanical index (CBI) 0.58 ± 0.43 and 0.75 (0–1); TBI 0.93 ± 0.17 and 1 (0.35–1); and stiffness parameter at A1 (SP-A1) 86.16 ± 19.62 and 86.05 (42.94–141.66). Conclusion: Relatively low keratometry, with a Kmax lower than 47.6 D, can occur in up to 28.30% of clinical keratoconus. These cases have a less severe presentation of the disease. Future studies involving larger populations and prospective designs are necessary to confirm the prevalence of keratoconus with low keratometry and define prognostic factors in such cases. [ABSTRACT FROM AUTHOR]

The evaluation of the load-settlement behavior of piles is crucial in meeting the strength and serviceability criteria for pile analysis and design. The most reliable approach for estimating this behavior is by conducting pile load tests. However, due to the considerable expense and time requirements of these tests, the load-transfer methods were used routinely in practice. The objective of this study is to explore the potential application of several machine learning (ML) algorithms to evaluate the load-settlement behavior of axially loaded single square precast prestressed concrete from cone penetration test (CPT) data. Several ML models such as artificial neural network (ANN), random forest (RF), and gradient boosted tree (GBT), were developed to estimate the load-settlement behavior from CPT data (corrected cone tip resistance, qt, and sleeve friction, fs). A database of load-settlement curves of 64 static pile load tests and corresponding CPT data were compiled and used for the development of these ML models. The developed ANN, RF, and GBT models are evaluated based on several statistical criteria. The load-settlement curves predicted using the developed ML models were compared with the measured curves from pile load tests and the load-settlement curves predicted using the conventional load-transfer methods. The results of this study demonstrated the great potential of using ML models to predict the load-settlement behavior of axially loaded piles from CPT data. The comparison clearly shows that ML models outperformed the load-transfer methods. The results showed that both the GBT and ANN algorithms demonstrated to be the best-performing ML models. [ABSTRACT FROM AUTHOR]

Machine learning has grown in popularity in the past few years for susceptibility and hazard mapping tasks. Necessary steps for the generation of a susceptibility or hazard map are repeatedly implemented in new studies. We present a Random Forest classifier-based landslide susceptibility and hazard mapping framework to facilitate future mapping studies using machine learning. The framework, as a piece of software, follows the FAIR paradigm, and hence is set up as a transparent, reproducible and modularly extensible workflow. It contains pre-implemented steps from conceptualisation to map generation, such as the generation of input datasets. The framework can be applied to different areas of interest using different environmental features and is also flexible in terms of the desired scale and resolution of the final map. To demonstrate the functionality and validity of the framework, and to explore the challenges and limitations of Random Forest-based susceptibility and hazard mapping, we apply the framework to a test case. This test case conveys the influence of the training dataset on the generated susceptibility maps in terms of feature combination, influence of non-landslide instances and representativeness of the training data with respect to the area of interest. A comparison of the test case results with the literature shows that the framework works reliably. Furthermore, the results obtained in this study complement the findings of previous studies that demonstrate the sensitivity of the training process to the training data, particularly in terms of its representativeness. [ABSTRACT FROM AUTHOR]

Understanding the extent of landslide damage is important for reducing the impact of landslides, which can cause great losses of life and property. Although numerous studies have been done on landslide disaster susceptibility, they have been limited by an unreasonable negative sample selection strategy or the absence of subjective environmental information of the study area in a single machine learning evaluation model. To evaluate landslide susceptibility based on sample optimization, we propose an analytic hierarchy process (AHP) method weighted by an improved random forest (RF) model. Based on the density analysis of landslide data, this method employs the certainty factor (CF) method to generate negative sample data. Correspondingly, ADB_RF, an enhanced RF model based on adaptive boosting (AdaBoost) is proposed to obtain objective weights, which are then combined with subjective weights obtained by the AHP (CF-combination). Additionally, a case study on the evaluation of landslide disasters was conducted in the Chuxiong Autonomous Prefecture of Yunnan, China. The results show the following: (1) the proposed landslide susceptibility evaluation method could objectively reflect the area prone to landslides with a high degree of accuracy and efficacy. (2) The area under the curve (AUC) of the CF-combination model reached 96.1%, indicating a high degree of accuracy. (3) In the northwestern region of Chuxiong Prefecture, more extremely high-risk areas were found than in the southeast; therefore, it has a high likelihood of experiencing another landslide disaster, which requires special attention. Accordingly, the research findings have significant reference value for preventing disasters and mitigating losses. [ABSTRACT FROM AUTHOR]

A set of quantum chemical descriptors (molecular polarization, heat capacity, entropy, Mulliken net charge of the most positive hydrogen atom, APT charge of the most negative atom and APT charge of the most positive atom with hydrogen summed into heavy atoms) was successfully used to establish the classification models for the toxicity pLC50 of pesticides in Americamysis bahia. The optimal random forest model (Class Model A) yielded predictive accuracy of 100% (training set of 217 pesticides), 95.8% (test set of 72 pesticides) and 99.0% (total set of 289 pesticides), which were very satisfactory, compared with previous classification models reported for the toxicity of compounds in aquatic organisms. Therefore, it is reasonable to apply the quantum chemical descriptors associated with molecular structural information on molecular bulk, chemical reactivity and weak interactions, to develop classification models for the toxicity pLC50 of pesticides in A. bahia. [ABSTRACT FROM AUTHOR]

This paper presents an integrated blockchain solution for health monitoring of livestock and leverages the Directed Acyclic Graph (DAG) based blockchain. The novel Social Relation Based DAG Blockchain (SRBDB) system which exploits the social relation of nodes, is implemented for lameness detection data storage of dairy cattle. A Friend Cluster is formed based on the social relations of blockchain nodes. The SRBDB system proposes to manage the data regarding the lameness detection of cattle. Data inclusive of standing time, number of steps, and lying time for the cows is captured and pre-processed using Local Outlier Factor (LOF) algorithm and then classified using a novel weighted voting-based random forest classification algorithm. Extensive experimentation reveals that weighted voting based random forest algorithm shows accuracy of 98%. Further, data storage at a standard node and average latency are reduced by 92.85% and 21.4% respectively. The SRBDB approach, compared with other schemes, shows better performance regarding data storage requirement and average latency. [ABSTRACT FROM AUTHOR]

Dealing with network attacks is becoming more uphill as we go further due to the complexity of computer networks. Among all the network attacks, DDoS attacks are widespread and challenging to detect. Because launching these attacks requires no vulnerability in the target network and they are like legitimate traffic, there is no certain solution for detecting them. Analyzing network users' behavior can be a well-founded solution for detecting anomalies in network resource usage. Since, in most networks, the users' behavior differs at different times of the day, in this paper, we proposed a DDoS attack detection method that clusters the network users' behavior based on adaptive time intervals in a single day. Our contribution is introducing the Timestamp feature as a primary indicator of normal behavior during different times of the day. Time intervals are computed adaptively by clustering the network IP flow using DBSCAN. This process leads to the extraction of a new feature that helps to detect DDoS attacks more accurately. To demonstrate the importance and impact of our new feature, several attack classification models have been trained using prevalent shallow machine algorithms such as Support Vector Machine (SVM), Random Forest (RF), and XGBoost. The method is also validated with the CICDDoS2019 and the CICIoT2023 datasets, which are the most popular and latest DDoS attack datasets. The results showed that our new feature has improved the evaluation metrics impressively with both datasets. [ABSTRACT FROM AUTHOR]

Sports analytics (SA) incorporate machine learning (ML) techniques and models for performance prediction. Researchers have previously evaluated ML models applied on a variety of basketball statistics. This paper aims to benchmark the forecasting performance of 14 ML models, based on 18 advanced basketball statistics and key performance indicators (KPIs). The models were applied on a filtered pool of 90 high-performance players. This study developed individual forecasting scenarios per player and experimented using all 14 models. The models' performance ranking was developed using a bespoke evaluation metric, called weighted average percentage error (WAPE), formulated from the weighted mean absolute percentage error (MAPE) evaluation results of each forecasted statistic and model. Moreover, we employed a comprehensive forecasting approach to improve KPI's results. Results showed that Tree-based models, namely Extra Trees, Random Forest, and Decision Tree, are the best performers in most of the forecasted performance indicators, with the best performance achieved by Extra Trees with a WAPE of 34.14%. In conclusion, we achieved a 3.6% MAPE improvement for the selected KPI with our approach on unseen data. [ABSTRACT FROM AUTHOR]

We use monthly data covering a century-long sample period (1915–2021) to study whether geopolitical risk helps to forecast subsequent gold volatility. We account not only for geopolitical threats and acts, but also for 39 country-specific sources of geopolitical risk. The response of subsequent volatility is heterogeneous across countries and nonlinear. We find that accounting for geopolitical risk at the country level improves forecast accuracy, especially when we use random forests to estimate our forecasting models. As an extension, we report empirical evidence on the predictive value of the country-level sources of geopolitical risk for two other candidate safe-haven assets, oil and silver, over the sample periods 1900–2021 and 1915–2021, respectively. Our results have important implications for the portfolio and risk-management decisions of investors who seek a safe haven in times of heightened geopolitical tensions. [ABSTRACT FROM AUTHOR]