Ensemble Of Machine Learning Algorithms Research Articles

Longitudinal patterns of natural hazard exposures and anxiety and depression symptoms among young adults in four low- and middle-income countries

We estimated the effect of community-level natural hazard exposure during prior developmental stages on later anxiety and depression symptoms among young adults and potential differences stratified by gender. We analyzed longitudinal data (2002–2020) on 5585 young adults between 19 and 26 years in Ethiopia, India, Peru, and Vietnam. A binary question identified community-level exposure, and psychometrically validated scales measured recent anxiety and depression symptoms. Young adults with three exposure histories (“time point 1,” “time point 2,” and “both time points”) were contrasted with their unexposed peers. We applied a longitudinal targeted minimum loss-based estimator with an ensemble of machine learning algorithms for estimation. Young adults living in exposed communities did not exhibit substantially different anxiety or depression symptoms from their unexposed peers, except for young women in Ethiopia who exhibited less anxiety symptoms (average causal effect [ACE] estimate = − 8.86 [95% CI: − 17.04, − 0.68] anxiety score). In this study, singular and repeated natural hazard exposures generally were not associated with later anxiety and depression symptoms. Further examination is needed to understand how distal natural hazard exposures affect lifelong mental health, which aspects of natural hazards are most salient, how disaster relief may modify symptoms, and gendered, age-specific, and contextual differences.

Variation and attribution of probable maximum precipitation of China using a high-resolution dataset in a changing climate

Abstract. Accurate assessment of the probable maximum precipitation (PMP) is crucial in assessing the resilience of high-risk water infrastructures, water resource management, and hydrological hazard mitigation. Conventionally, PMP is estimated based on a static climate assumption and is constrained by the insufficient spatial resolution of ground observations, thus neglecting the spatial heterogeneity and temporal variability of climate systems. Such assumptions are critical, especially for China, which is highly vulnerable to global warming in ∼ 100 000 existing reservoirs. Here, we use the finest-spatiotemporal-resolution (1 d and 1 km) precipitation dataset from an ensemble of machine learning algorithms to present the spatial distribution of 1 d PMP based on the improved Hershfield method. Current reservoir design values, a quasi-global satellite-based PMP database, and in situ precipitation are used to benchmark against our results. The 35-year running trend from 1961–1995 to 1980–2014 is quantified and partitioned, followed by future projections using the Coupled Model Inter-comparison Project Phase 6 simulations under two scenarios. We find that the national PMP generally decreases from southeast to northwest and is typically dominated by the high variability of precipitation extremes in northern China and high intensity in southern China. Though consistent with previous project design values, our PMP calculations present underestimations by comparing them with satellite and in situ results due to differences in spatial scales and computation methods. Interannual variability, instead of the intensification of precipitation extremes, dominates the PMP running trends on a national scale. Climate change, mainly attributed to land–atmosphere coupling effects, leads to a widespread increase (> 20 %) in PMP across the country under the SSP126 scenario, which is projected to be higher along with the intensification of CO2 emissions. Our observation- and modeling-based results can provide valuable implications for water managers under a changing climate.

Daily reservoir inflow prediction using stacking ensemble of machine learning algorithms

ABSTRACT The present study aims to evaluate the potentiality of Bidirectional Long Short-Term Memory (Bi-LSTM), Convolutional Neural Networks (CNNs), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LGBM), and Random Forest (RF) for predicting daily inflows to the Sri Ram Sagar Project (SRSP), Telangana, India. Inputs to the model are rainfall, evaporation, time lag inflows, and climate indices. Seven combinations (S1–S7) of inputs were made. Fifteen and a half years of data were considered, out of which 11 years were used for training. Hyperparameter tuning is performed with the Tree-Structured Parzen Estimator. The performance of the algorithms is assessed using Kling–Gupta efficiency (KGE). Results indicate that Bi-LSTM with combination S7 performed better than others, as evident from KGE values of 0.92 and 0.87 during the training and testing, respectively. Furthermore, the Stacking Ensemble Mechanism (SEM) has also been employed to ascertain its efficacy over other chosen algorithms, resulting in KGE values of 0.94 and 0.89 during training and testing. It has also been able to simulate peak inflow events satisfactorily. Thus, SEM is a better alternative for reservoir inflow predictions.

Learning vs. understanding: When does artificial intelligence outperform process-based modeling in soil organic carbon prediction?

In recent years, machine learning (ML) algorithms have gained substantial recognition for ecological modeling across various temporal and spatial scales. However, little evaluation has been conducted for the prediction of soil organic carbon (SOC) on small data sets commonly inherent to long-term soil ecological research. In this context, the performance of ML algorithms for SOC prediction has never been tested against traditional process-based modeling approaches. Here, we compare ML algorithms, calibrated and uncalibrated process-based models as well as multiple ensembles on their performance in predicting SOC using data from five long-term experimental sites (comprising 256 independent data points) in Austria. Using all available data, the ML-based approaches using Random forest and Support vector machines with a polynomial kernel were superior to all process-based models. However, the ML algorithms performed similar or worse when the number of training samples was reduced or when a leave-one-site-out cross validation was applied. This emphasizes that the performance of ML algorithms is strongly dependent on the data-size related quality of learning information following the well-known curse of dimensionality phenomenon, while the accuracy of process-based models significantly relies on proper calibration and combination of different modeling approaches. Our study thus suggests a superiority of ML-based SOC prediction at scales where larger datasets are available, while process-based models are superior tools when targeting the exploration of underlying biophysical and biochemical mechanisms of SOC dynamics in soils. Therefore, we recommend applying ensembles of ML algorithms with process-based models to combine advantages inherent to both approaches.

Remotely sensed desertification modeling using ensemble of machine learning algorithms

Due to having a sensitive and fragile ecosystem, dry areas are constantly exposed to land degradation and desertification. Therefore, it is necessary to formulate appropriate strategies for quantitative assessment of desertification that are highly accurate. In this research, desertification of the region was first evaluated using MEDALUS model, then according to the results of MEDALUS model and reviewing the results of other researchers, 8 indicators remote sensing that had the highest correlation with field data were selected for modeling. Four machine learning methods Support Vector Machine (SVM), Gradient Boosting Machine (GBM), Generalized Linear Models (GLM) and Random Forests (RF) were used to model the risk of desertification in northeastern Iran. Finally, the weighted average of the ensemble model in the SDM statistical package was used to predict the desertification of the region. Based on the results obtained from MEDALUS model, the indicators of drought resistance (score 162), conservation operations (score 158) and soil salinity (score 155), in the working units of abandoned lands, wetland lands, and Salty lands located in the north East of the region, have increased the process of desertification. The results of modeling using machine learning methods showed that in 2002, the SVM model (AUC = 0.91, TSS = 0.93, and Kappa = 0.86) and in 2021, the RF model (AUC = 0.94, TSS = 0.94, and Kappa = 0.90) have performed best. The forecast of the combined model for desertification in 2021 in the studied area showed that the northeastern and sporadically in the central parts of the studied area are affected by the progress of the desertification process. Therefore, by considering the results of the combined model (as a model with the least uncertainty), it is possible to reduce the progress of the desertification process by planning, optimal management and applying corrective methods in the areas affected by desertification.

An efficient smart grid stability prediction system based on machine learning and deep learning fusion model

<span>A smart grid is a modern power system that allows for bidirectional communication, driven mostly by the idea of demand responsiveness. Predicting the stability of the smart grid is necessary for improving its dependability and maximizing the efficacy and regularity of electricity delivery. Predicting smart grid stability is difficult owing to the various elements that impact it, including consumer and producer engagement, which may contribute to smart grid stability. This research work proposes machine learning (ML) and deep learning (DL) approaches for predicting smart grid sustainability. Five ML algorithms, namely support vector machine (SVM), decision tree (DT), K-nearest neighbor (KNN), random forest (RF), and logistic regression (LR), were applied for the prediction of smart grid stability. Later, the stacking ensemble and voting ensemble of ML algorithms were also applied for prediction. To further increase accuracy, a novel fusion model with DL artifical neural networks (ANN) and ML SVM was applied and achieved an accuracy of 98.92%. The experiment results show that the proposed model outperformed existing models for smart grid stability prediction.</span>

Securing cloud-enabled smart cities by detecting intrusion using spark-based stacking ensemble of machine learning algorithms

<abstract> <p>With the use of cloud computing, which provides the infrastructure necessary for the efficient delivery of smart city services to every citizen over the internet, intelligent systems may be readily integrated into smart cities and communicate with one another. Any smart system at home, in a car, or in the workplace can be remotely controlled and directed by the individual at any time. Continuous cloud service availability is becoming a critical subscriber requirement within smart cities. However, these cost-cutting measures and service improvements will make smart city cloud networks more vulnerable and at risk. The primary function of Intrusion Detection Systems (IDS) has gotten increasingly challenging due to the enormous proliferation of data created in cloud networks of smart cities. To alleviate these concerns, we provide a framework for automatic, reliable, and uninterrupted cloud availability of services for the network data security of intelligent connected devices. This framework enables IDS to defend against security threats and to provide services that meet the users' Quality of Service (QoS) expectations. This study's intrusion detection solution for cloud network data from smart cities employed Spark and Waikato Environment for Knowledge Analysis (WEKA). WEKA and Spark are linked and made scalable and distributed. The Hadoop Distributed File System (HDFS) storage advantages are combined with WEKA's Knowledge flow for processing cloud network data for smart cities. Utilizing HDFS components, WEKA's machine learning algorithms receive cloud network data from smart cities. This research utilizes the wrapper-based Feature Selection (FS) approach for IDS, employing both the Pigeon Inspired Optimizer (PIO) and the Particle Swarm Optimization (PSO). For classifying the cloud network traffic of smart cities, the tree-based Stacking Ensemble Method (SEM) of J48, Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) are applied. Performance evaluations of our system were conducted using the UNSW-NB15 and NSL-KDD datasets. Our technique is superior to previous works in terms of sensitivity, specificity, precision, false positive rate (FPR), accuracy, F1 Score, and Matthews correlation coefficient (MCC).</p> </abstract>

Periconceptional Dietary Patterns and Adverse Pregnancy and Birth Outcomes

BackgroundThe periconceptional period is a critical window for the origins of adverse pregnancy and birth outcomes, yet little is known about the dietary patterns that promote perinatal health. ObjectiveWe used machine learning methods to determine the effect of periconceptional dietary patterns on risk of preeclampsia, gestational diabetes, preterm birth, small-for-gestational-age (SGA) birth, and a composite of these outcomes. MethodsWe used data from 8259 participants in the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-Be (8 US medical centers, 2010‒2013). Usual daily periconceptional intake of 82 food groups was estimated from a food frequency questionnaire. We used k-means clustering with a Euclidean distance metric to identify dietary patterns. We estimated the effect of dietary patterns on each perinatal outcome using targeted maximum likelihood estimation and an ensemble of machine learning algorithms, adjusting for confounders including health behaviors and psychological, neighborhood, and sociodemographic factors. ResultsThe 4 dietary patterns that emerged from our data were identified as “Sandwiches and snacks” (34% of the sample); “High fat, sugar, and sodium” (29%); “Beverages, refined grains, and mixed dishes” (21%); and “High fruits, vegetables, whole grains, and plant proteins” (16%). One-quarter of pregnancies had preeclampsia (8% incidence), gestational diabetes (5%), preterm birth (8%), or SGA birth (8%). Compared with the “High fat, sugar, and sodium” pattern, there were 3.3 to 4.3 fewer cases of the composite adverse outcome per 100 pregnancies among participants following the “Beverages, refined grains and mixed dishes” pattern (risk difference −0.043; 95% confidence interval −0.078, −0.009), “High fruits, vegetables, whole grains and plant proteins” pattern (−0.041; 95% confidence interval −0.078, −0.004), and “Sandwiches and snacks” pattern (−0.033; 95% confidence interval −0.065, −0.002). ConclusionsOur results highlight that there are a variety of periconceptional dietary patterns that are associated with perinatal health and reinforce the negative health implications of diets high in fat, sugars, and sodium.

Ensemble of Machine Learning Algorithms for Rice Grain Yield Prediction Using UAV-Based Remote Sensing

Ensemble of Machine Learning Algorithms for Rice Grain Yield Prediction Using UAV-Based Remote Sensing

UPDATING ABOVEGROUND BIOMASS AT A PAN-EUROPEAN SCALE THROUGH SATELLITE DATA AND ARTIFICIAL INTELLIGENCE

Abstract. In this work, an ensemble of machine learning algorithms was trained using stratified sampling from an existing European-scale biomass map from 2018 to predict an updated version for 2020. The objective of stratification is to make sure that the full range of biomass values is represented. The sampled biomass values from 2018 were filtered to remove areas that did were subject to forest disturbances between 2018 and 2020. This information was available from forest cover/loss/gain maps derived from satellite imagery. We train using a total of 49 features derived from the following sources: bioclimatic data, maps of land-cover, tree cover, tree height, annual composites of vegetation indices per pixel (EVI and NDVI) obtained from Sentinel-2, radar backscatter median annual values from Sentinel-1 and ALOS-2, and the ALOS DSM (3D) elevation grid. A model was created dividing Europe into 19 tiles to limit variability due to very different bioclimatic zones. The result is a raster with 100 m × 100 m resolution and an estimated value of biomass (Mg ha−1) at each node. Overall results on validation data over Europe report a root mean square error (RMSE) of 32.4 Mg ha−1 and a mean absolute error (MAE) of 21.5 Mg ha−1; when considering single tiles, the largest RMSE was 54.7 Mg ha−1 in tile D2, which can be explained by the very high variance of climate, environment, terrain topography and biomass values as the tile enclosed the Alpine region and the western part of Eastern Europe.

Predicting priority management areas for land use/cover change in the transboundary Okavango basin based on machine learning

Remote sensing and modelling of land use/land cover (LULC) change is useful to reveal the extent and spatial patterns of landscape changes at various environments and scales. Predicting susceptibility to LULC change is crucial for policy formulation and land management. However, the use of machine learning (ML) for modelling LULC change is limited. This study modelled LULC change susceptibility in the Okavango basin using ML techniques. Areas with high LULC change susceptibility are termed priority management areas (PMAs) in this study. Trajectories of LULC change between 1996 and 2020 are derived from existing LULC change maps of the Okavango basin. Overlay analysis is then used to detect patches of LULC change transitions. Three LULC transitional categories are adopted for modelling PMAs, namely 1) from natural to anthropogenic classes (Category A); 2) from anthropogenic to natural classes (Category B); and 3) from natural to another natural class (Category C). An ensemble of ML algorithms is calibrated with categories of LULC change and social-ecological drivers of change to produce maps showing the susceptibility of LULC change in the basin. Thereafter, thresholding is done on probability maps of susceptibility to LULC change based on the maximum sum of sensitivity and specificity (max SSS) to delineate PMAs. Results for trajectories of LULC change indicate that anthropogenic activities (croplands, built-up areas, and barelands) generally expanded, displacing natural areas (wetlands, woodlands, water, and shrubland) from 1996 to 2020. Regarding PMAs, anthropogenic-related PMAs (Category A ∼34 560 km2) covered a larger area compared to the natural ones (Categories B∼33 407 km2) and (Categories C∼15 040 km2). The findings of this study emphasize the value of ensemble ML modelling in identifying PMAs and guiding transboundary land use planning. Overall, this study highlights the role of anthropogenic activities in driving land use changes in Transboundary Drainage Basins (TDBs) and suggests a need to promote sustainable practices in predicted PMAs through comprehensive planning to ensure water availability in the Okavango basin.

A machine learning ensemble approach for predicting growth of abalone reared in land-based aquaculture in Hokkaido, Japan

Land-based aquaculture is an ideal aquaculture solution for creating high-quality seafoods and providing optimal conditions for maximizing growth of seafood production because environmental factors are well controlled. Predicting the growth of indoor-cultured abalone is meaningful because it facilitates evaluation of the effectiveness of this type of farming and understanding of the effects of controllable environmental factors on abalone growth. In this study, such predictions were made using an ensemble of machine learning algorithms: the random forest, gradient boosting, support vector machine, and neural network algorithms. Data were collected in the town of Fukushima, Hokkaido, Japan, and the increase in the weight of abalone was hypothesized from independent variables, including air and water temperature, loss of individuals caused by mortality or emigration, flow speed, age, and growth period between two measurements. The results showed that the ensemble method predicts growth well, with a low mean absolute error and mean square error. Temperature adjustment can make a strong contribution to increasing the weight of abalone, where a stable and warm temperature enhances growth. Moreover, the age of abalone is closely related to growth. Abalone size increased strongly in the early stages but decreased slightly once near market size.

FLIGHT FARE PREDICTION USING MACHINE LEARNING

ABSTRACT The Flight Fare Prediction System is a comprehensive solution aimed at accurately forecasting flight ticket prices, providing travelers with valuable insights for better planning and decision-making. Nowadays, airline ticket prices can vary dynamically for the same flight. From the customers perspective, they want to save money, so I have proposed a model that predicts the approximate ticket price. This system leverages machine learning algorithms and historical flight data to generate accurate fare predictions. The system utilizes a vast dataset comprising historical flight fares, including factors such as travel dates, destinations, airlines, departure times, and various other relevant variables. By analyzing this data using advanced machine learning techniques, the system learns patterns and relationships, enabling it to make reliable predictions about future flight fares. An ensemble of machine learning algorithms, including regression-based models like Random Forest, Gradient Boosting, and Support Vector Regression, is employed to capture complex patterns and relationships within the data. This system will give people an idea of the trends the prices follow and also provide the predicted value of the price, which they can check before booking flights to save money. This kind of system or service can be provided to customers through flight booking companies to help them book tickets. KEYWORDS - Flight Fare Prediction, Machine Learning, Historical Flight data, Random Forest.

Improving Forecasting Accuracy of Multi-Scale Groundwater Level Fluctuations Using a Heterogeneous Ensemble of Machine Learning Algorithms

Accurate groundwater level (GWL) forecasts are crucial for the efficient utilization, strategic long-term planning, and sustainable management of finite groundwater resources. These resources have a substantial impact on decisions related to irrigation planning, crop selection, and water supply. This study evaluates data-driven models using different machine learning algorithms to forecast GWL fluctuations for one, two, and three weeks ahead in Bangladesh’s Godagari upazila. To address the accuracy limitations inherent in individual forecasting models, a Bayesian model averaging (BMA)-based heterogeneous ensemble of forecasting models was proposed. The dataset encompasses 1807 weekly GWL readings (February 1984 to September 2018) from four wells, divided into training (70%), validation (15%), and testing (15%) subsets. Both standalone models and ensembles employed a Minimum Redundancy Maximum Relevance (MRMR) algorithm to select the most influential lag times among candidate GWL lags up to 15 weeks. Statistical metrics and visual aids were used to evaluate the standalone and ensemble GWL forecasts. The results consistently favor the heterogeneous BMA ensemble, excelling over standalone models for multi-step ahead forecasts across time horizons. For instance, at GT8134017, the BMA approach yielded values like R (0.93), NRMSE (0.09), MAE (0.50 m), IOA (0.96), NS (0.87), and a-20 index (0.94) for one-week-ahead forecasts. Despite a slight decline in performance with an increasing forecast horizon, evaluation indices confirmed the superior BMA ensemble performance. This ensemble also outperformed standalone models for other observation wells. Thus, the BMA-based heterogeneous ensemble emerges as a promising strategy to bolster multi-step ahead GWL forecasts within this area and beyond.

Validating instructional design and predicting student performance in histology education: Using machine learning via virtual microscopy.

As a part of modern technological environments, virtual microscopy enriches histological learning, with support from large institutional investments. However, existing literature does not supply empirical evidence of its role in improving pedagogy. Virtual microscopy provides fresh opportunities for investigating user behavior during the histology learning process, through digitized histological slides. This study establishes how students' perceptions and user behavior data can be processed and analyzed using machine learning algorithms. These also provide predictive data called learning analytics that enable predicting students' performance and behavior favorable for academic success. This information can be interpreted and used for validating instructional designs. Data on the perceptions, performances, and user behavior of 552 students enrolled in a histology course were collected from the virtual microscope, Cytomine®. These data were analyzed using an ensemble of machine learning algorithms, the extra-tree regression method, and predictive statistics. The predictive algorithms identified the most pertinent histological slides and descriptive tags, alongside 10 types of student behavior conducive to academic success. We used these data to validate our instructional design, and align the educational purpose, learning outcomes, and evaluation methods of digitized histological slides on Cytomine®. This model also predicts students' examination scores, with an error margin of <0.5 out of 20 points. The results empirically demonstrate the value of a digital learning environment for both students and teachers of histology.

Multiclass Classification of Brain MRI through DWT and GLCM Feature Extraction with Various Machine Learning Algorithms

This study delves into the domain of medical diagnostics, focusing on the crucial task of accurately classifying brain tumors to facilitate informed clinical decisions and optimize patient outcomes. Employing a diverse ensemble of machine learning algorithms, the paper addresses the challenge of multiclass brain tumor classification. The investigation centers around the utilization of two distinct datasets: the Brats dataset, encompassing cases of High-Grade Glioma (HGG) and Low-Grade Glioma (LGG), and the Sartaj dataset, comprising instances of Glioma, Meningioma, and No Tumor. Through the strategic deployment of Discrete Wavelet Transform (DWT) and Gray-Level Co-occurrence Matrix (GLCM) features, coupled with the implementation of Support Vector Machines (SVM), k-nearest Neighbors (KNN), Decision Trees (DT), Random Forest, and Gradient Boosting algorithms, the research endeavors to comprehensively explore avenues for achieving precise tumor classification. Preceding the classification process, the datasets undergo pre-processing and the extraction of salient features through DWT-derived frequency-domain characteristics and texture insights harnessed from GLCM. Subsequently, a detailed exposition of the selected algorithms is provided and elucidates the pertinent hyperparameters. The study's outcomes unveil noteworthy performance disparities across diverse algorithms and datasets. SVM and Random Forest algorithms exhibit commendable accuracy rates on the Brats dataset, while the Gradient Boosting algorithm demonstrates superior performance on the Sartaj dataset. The evaluation process encompasses precision, recall, and F1-score metrics, thereby providing a comprehensive assessment of the classification prowess of the employed algorithms.

Is the Association Between Fruits and Vegetables and Preeclampsia Due to Higher Dietary Vitamin C and Carotenoid Intakes?

BackgroundDiets dense in fruits and vegetables are associated with a reduced risk of preeclampsia, but pathways underlying this relationship are unclear. Dietary antioxidants may contribute to the protective effect. ObjectiveWe determined the extent to which the effect of dietary fruit and vegetable density on preeclampsia is because of high intakes of dietary vitamin C and carotenoids. Methods: We used data from 7572 participants in the Nulliparous Pregnancy Outcomes Study: monitoring mothers-to-be (8 United States medical centers, 2010‒2013). Usual daily periconceptional intake of total fruits and total vegetables was estimated from a food frequency questionnaire. We estimated the indirect effect of ≥2.5 cups/1000 kcal of fruits and vegetables through vitamin C and carotenoid on the risk of preeclampsia. We estimated these effects using targeted maximum likelihood estimation and an ensemble of machine learning algorithms, adjusting for confounders, including other dietary components, health behaviors, and psychological, neighborhood, and sociodemographic factors. ResultsParticipants who consumed ≥2.5 cups of fruits and vegetables per 1000 kcal were less likely than those who consumed <2.5 cups/1000 kcal to develop preeclampsia (6.4% compared with 8.6%). After confounder adjustment, we observed that higher fruit and vegetable density was associated with 2 fewer cases of preeclampsia (risk difference: –2.0; 95% CI: –3.9, –0.1)/100 pregnancies compared with lower density diets. High dietary vitamin C and carotenoid intake was not associated with preeclampsia. The protective effect of high fruit and vegetable density on the risk of preeclampsia and late-onset preeclampsia was not mediated through dietary vitamin C and carotenoids. ConclusionsEvaluating other nutrients and bioactives in fruits and vegetables and their synergy is worthwhile, along with characterizing the effect of individual fruits or vegetables on preeclampsia risk.

Fault detection on the 3-D printed objective surface by using the SVM algorithm

The concept of 3D-printed items has gained traction in the age of Industry 4.0, proving helpful in terms of money and time spent. Assemble the products like this are physically in stages according to computerised CAD data. However, faults still exist in 3D printed objects owing to the diversity in characteristics and structure, which degrades printed quality. Must uncover these mistakes at each development stage of the product. This study Using an ensemble of machine learning algorithms and pre-trained models, details a procedure for detecting anomalies at each layer. For defect detection, the recommended combination is learned in advance offline and then applied in real-time online. This research compares several pre-trained models using SVM, machine learning methods for monitoring and defect detection in Fused Deposition Modeling in an experimental setting (FDM). According to the findings, the best accuracy was achieved by combining Alexnet with the SVM algorithm. As a result of its minimal experimental and computational costs, apply the suggested fault detection technique for real-time defect detection.

Benchmarking of Machine Learning classifiers on plasma proteomic for COVID-19 severity prediction through interpretable artificial intelligence

The SARS-CoV-2 pandemic highlighted the need for software tools that could facilitate patient triage regarding potential disease severity or even death. In this article, an ensemble of Machine Learning (ML) algorithms is evaluated in terms of predicting the severity of their condition using plasma proteomics and clinical data as input.An overview of AI-based technical developments to support COVID-19 patient management is presented outlining the landscape of relevant technical developments. Based on this review, the use of an ensemble of ML algorithms that analyze clinical and biological data (i.e., plasma proteomics) of COVID-19 patients is designed and deployed to evaluate the potential use of AI for early COVID-19 patient triage. The proposed pipeline is evaluated using three publicly available datasets for training and testing. Three ML “tasks” are defined, and several algorithms are tested through a hyperparameter tuning method to identify the highest-performance models. As overfitting is one of the typical pitfalls for such approaches (mainly due to the size of the training/validation datasets), a variety of evaluation metrics are used to mitigate this risk.In the evaluation procedure, recall scores ranged from 0.6 to 0.74 and F1-score from 0.62 to 0.75. The best performance is observed via Multi-Layer Perceptron (MLP) and Support Vector Machines (SVM) algorithms. Additionally, input data (proteomics and clinical data) were ranked based on corresponding Shapley additive explanation (SHAP) values and evaluated for their prognosticated capacity and immuno-biological credence. This “interpretable” approach revealed that our ML models could discern critical COVID-19 cases predominantly based on patient's age and plasma proteins on B cell dysfunction, hyper-activation of inflammatory pathways like Toll-like receptors, and hypo-activation of developmental and immune pathways like SCF/c-Kit signaling. Finally, the herein computational workflow is corroborated in an independent dataset and MLP superiority along with the implication of the abovementioned predictive biological pathways are corroborated.Regarding limitations of the presented ML pipeline, the datasets used in this study contain less than 1000 observations and a significant number of input features hence constituting a high-dimensional low-sample (HDLS) dataset which could be sensitive to overfitting. An advantage of the proposed pipeline is that it combines biological data (plasma proteomics) with clinical-phenotypic data. Thus, in principle, the presented approach could enable patient triage in a timely fashion if used on already trained models. However, larger datasets and further systematic validation are needed to confirm the potential clinical value of this approach. The code is available on Github: https://github.com/inab-certh/Predicting-COVID-19-severity-through-interpretable-AI-analysis-of-plasma-proteomics.

AGRN: accurate gene regulatory network inference using ensemble machine learning methods.

Biological processes are regulated by underlying genes and their interactions that form gene regulatory networks (GRNs). Dysregulation of these GRNs can cause complex diseases such as cancer, Alzheimer's and diabetes. Hence, accurate GRN inference is critical for elucidating gene function, allowing for the faster identification and prioritization of candidate genes for functional investigation. Several statistical and machine learning-based methods have been developed to infer GRNs based on biological and synthetic datasets. Here, we developed a method named AGRN that infers GRNs by employing an ensemble of machine learning algorithms. From the idea that a single method may not perform well on all datasets, we calculate the gene importance scores using three machine learning methods-random forest, extra tree and support vector regressors. We calculate the importance scores from Shapley Additive Explanations, a recently published method to explain machine learning models. We have found that the importance scores from Shapley values perform better than the traditional importance scoring methods based on almost all the benchmark datasets. We have analyzed the performance of AGRN using the datasets from the DREAM4 and DREAM5 challenges for GRN inference. The proposed method, AGRN-an ensemble machine learning method with Shapley values, outperforms the existing methods both in the DREAM4 and DREAM5 datasets. With improved accuracy, we believe that AGRN inferred GRNs would enhance our mechanistic understanding of biological processes in health and disease. https://github.com/DuaaAlawad/AGRN. Supplementary data are available at Bioinformatics online.