Single Machine Learning Model Research Articles

EcGBMsub: an integrative stacking ensemble model framework based on eccDNA molecular profiling for improving IDH wild-type glioblastoma molecular subtype classification.

IDH wild-type glioblastoma (GBM) intrinsic subtypes have been linked to different molecular landscapes and outcomes. Accurate prediction of molecular subtypes of GBM is very important to guide clinical diagnosis and treatment. Leveraging machine learning technology to improve the subtype classification was considered a robust strategy. Several single machine learning models have been developed to predict survival or stratify patients. An ensemble learning strategy combines several basic learners to boost model performance. However, it still lacked a robust stacking ensemble learning model with high accuracy in clinical practice. Here, we developed a novel integrative stacking ensemble model framework (ecGBMsub) for improving IDH wild-type GBM molecular subtype classification. In the framework, nine single models with the best hyperparameters were fitted based on extrachromosomal circular DNA (eccDNA) molecular profiling. Then, the top five optimal single models were selected as base models. By randomly combining the five optimal base models, 26 different combinations were finally generated. Nine different meta-models with the best hyperparameters were fitted based on the prediction results of 26 different combinations, resulting in 234 different stacked ensemble models. All models in ecGBMsub were comprehensively evaluated and compared. Finally, the stacking ensemble model named "XGBoost.Enet-stacking-Enet" was chosen as the optimal model in the ecGBMsub framework. A user-friendly web tool was developed to facilitate accessibility to the XGBoost.Enet-stacking-Enet models (https://lizesheng20190820.shinyapps.io/ecGBMsub/).

Inferring molecular inhibition potency with AlphaFold predicted structures

Even though in silico drug ligand-based methods have been successful in predicting interactions with known target proteins, they struggle with new, unassessed targets. To address this challenge, we propose an approach that integrates structural data from AlphaFold 2 predicted protein structures into machine learning models. Our method extracts 3D structural protein fingerprints and combines them with ligand structural data to train a single machine learning model. This model captures the relationship between ligand properties and the unique structural features of various target proteins, enabling predictions for never before tested molecules and protein targets. To assess our model, we used a dataset of 144 Human G-protein Coupled Receptors (GPCRs) with over 140,000 measured inhibition constants (Ki) values. Results strongly suggest that our approach performs as well as state-of-the-art ligand-based methods. In a second modeling approach that used 129 targets for training and a separate test set of 15 different protein targets, our model correctly predicted interactions for 73% of targets, with explained variances exceeding 0.50 in 22% of cases. Our findings further verified that the usage of experimentally determined protein structures produced models that were statistically indistinct from the Alphafold synthetic structures. This study presents a proteo-chemometric drug screening approach that uses a simple and scalable method for extracting protein structural information for usage in machine learning models capable of predicting protein-molecule interactions even for orphan targets.

Machine learning forecasting of solar PV production using single and hybrid models over different time horizons

This study uses operational data from a 180 kWp grid-connected solar PV system to train and compare the performance of single and hybrid machine learning models in predicting solar PV production a day-ahead, a week-ahead, two weeks ahead and one month-ahead. The study also analyses the trend in solar PV production and the effect of temperature on solar PV production. The performance of the models is evaluated using R2 score, mean absolute error and root mean square error. The findings revealed the best-performing model for the day ahead forecast to be Artificial Neural Network. Random Forest gave the best performance for the two-week and a month-ahead forecast, while a hybrid model composed of XGBoost and Random Forest gave the best performance for the week-ahead prediction. The study also observed a downward trend in solar PV production, with an average monthly decline of 244.37 kWh. Further, it was observed that an increase in the module temperature and ambient temperature beyond 47 °C and 25 °C resulted in a decline in solar PV production. The study shows that machine learning models perform differently under different time horizons. Therefore, selecting suitable machine learning models for solar PV forecasts for varying time horizons is extremely necessary.

Development of an integrated machine learning model to improve the secondary inorganic aerosol simulation over the Beijing–Tianjin–Hebei region

Secondary inorganic aerosols (sulfate, nitrate, and ammonium, SNA) are the key components of PM2.5 in China. Accurate and seamless SNA concentration data therefore are important for pollution controls and environmental health risk studies. However, the spatiotemporal characters and changes of SNA are still difficult to accurately estimate due to the shortages of SNA datasets with high accuracy and resolution, as well as the uncertainties in SNA simulations arising from air-quality numerical models. In this study, we developed an integrated model to improve hourly SNA simulations with 5 km spatial resolution over the Beijing–Tianjin–Hebei (BTH) region by integrating the random forest (RF) and Light Gradient Boost Machine (LGBM) using Stacked Generalization. Our model fuses Three-dimensional (3D) numerical simulations from the Weather Research and Forecast model (WRF) and Nested Air Quality Prediction Model System (NAQPMS) with observations of 16 sites from the BTH monitoring network. Three months’ data from January to March 2020 were employed to evaluate the model performance using the cross-validation (CV) method. The results showed that the integrated model provide more accurate simulations of SNA than the 3D numerical model does, with root mean square errors (RMSE) decreased by 33%, 45%, and 35%; correlation coefficient (R) increased by 61%, 28%, and 34%; and Taylor skill score (TSS) increased by 331%, 85%, and 65% for sulfate, nitrate, and ammonium respectively. Moreover, the integrated model showed higher evaluation criteria and more accurate spatiotemporal characteristic compared with the single machine learning (ML) model, especially in heavily polluted area. This study provides a new approach to improve SNA simulations and reveals the potential of ML models for improving aerosol modeling when observational data are scarce.

High-resolution estimation of PM2.5 concentrations across China using multiple machine learning approaches and model fusion

Accurate estimation of fine particulate matter (PM2.5) concentrations can provide necessary data sources for epidemiological studies and environmental management decisions. However, the different sources and resolutions of satellite data and the limited number of monitoring stations decrease the accuracy of PM2.5 concentration estimation, which subsequently complicates the continuous estimation of high-resolution PM2.5. In this study, we evaluate and optimize five different machine learning methods, including cubist, random forest (RF), support vector machine (SVM), the generalized additive model (GAM), and Extreme Gradient Boosting (XGBoost), to perform spatial high-resolution (1 km) estimations of PM2.5 concentrations. Based on the optimization, two model fusion methods, i.e., the Bayesian model averaging and Granger–Ramanathan averaging algorithms, are applied and compared for estimating PM2.5 concentrations. Among the five single machine learning models, the RF model performed the best, with a coefficient of determination (R2) of 0.84, root mean square error (RMSE) of 5.09 μg/m3, and ratio of performance to interquartile distance (RPIQ) of 3.40. The cubist model ranks second in terms of performance (R2 = 0.82, RMSE = 5.37 μg/m3, and RPIQ = 3.23). The estimation accuracies of the five models decrease in the following order: RF > cubist > XGBoost > SVM > GAM. The fusion model performed better than the single model and reliably estimated the PM2.5 concentration (R2 = 0.86, RMSE = 5.04 μg/m3, and RPIQ = 3.44). The results show that the model fusion algorithms improved the overall performance of fine-level and long-term PM2.5 concentration estimation, and that the accuracies of the fusion models are significantly higher than those of the five machine learning models used individually. Fusion model is used to generate an annual PM2.5 concentration distribution map of China from 2017 to 2021. This study provides a reference for estimating PM2.5 concentrations and contributes to the formulation of targeted policies.

Combination prediction method of students' performance based on ant colony algorithm.

Students' performance is an important factor for the evaluation of teaching quality in colleges. The prediction and analysis of students' performance can guide students' learning in time. Aiming at the low accuracy problem of single model in students' performance prediction, a combination prediction method is put forward based on ant colony algorithm. First, considering the characteristics of students' learning behavior and the characteristics of the models, decision tree (DT), support vector regression (SVR) and BP neural network (BP) are selected to establish three prediction models. Then, an ant colony algorithm (ACO) is proposed to calculate the weight of each model of the combination prediction model. The combination prediction method was compared with the single Machine learning (ML) models and other methods in terms of accuracy and running time. The combination prediction model with mean square error (MSE) of 0.0089 has higher performance than DT with MSE of 0.0326, SVR with MSE of 0.0229 and BP with MSE of 0.0148. To investigate the efficacy of the combination prediction model, other prediction models are used for a comparative study. The combination prediction model with MSE of 0.0089 has higher performance than GS-XGBoost with MSE of 0.0131, PSO-SVR with MSE of 0.0117 and IDA-SVR with MSE of 0.0092. Meanwhile, the running speed of the combination prediction model is also faster than the above three methods.

Comparative analysis of single and hybrid machine learning models for daily solar radiation

This study investigates the estimation of daily solar radiation (SR) through various machine learning (ML) models, including the k-nearest neighbor algorithm (KNN), support vector regression (SVR), and random forest (RF), both individually and in combination with the wavelet transform (WT). The assessment of these models is based on meteorological data spanning three decades (1981–2010) from the province of Kütahya in Türkiye. To address the inherent uncertainty in these data-driven models, the quantile regression method is employed for uncertainty analysis. Statistical metrics, such as mean absolute error (MAE), root mean square error (RMSE), coefficient of determination (R2), mean prediction interval (MPI), and prediction interval coverage probability (PICP), are utilized to evaluate the effectiveness and uncertainties of the models. The SVR and KNN models exhibit comparable performances concerning both predictive accuracy and uncertainty levels. However, hybrid models, such as KNN-WT, RF-WT, and SVR-WT, display enhanced accuracy compared to individual ML models, as indicated by statistical performance criteria. Notably, the SVR-WT model, incorporating inputs such as sunshine duration, air temperature, wind speed, and relative humidity, outperforms other models in terms of RMSE (2.174 MJ/m2), MAE (1.721 MJ/m2), R2 (0.923), MPI (28.55), and PICP (0.80) for the testing dataset. In conclusion, the integration of WT significantly improves the performance of ML models, providing valuable insights for the design and operation of solar energy systems, where precise daily SR estimation is critical for optimal and cost-efficient operation.

Fully automated learning and predict price of aquatic products in Taiwan wholesale markets using multiple machine learning and deep learning methods

The aquatic market price is an essential indicator for fishermen in making decisions regarding conveyance, fish-catching, and wholesale strategies. As a guide for the aquaculture industry, understanding market prices, the geographical distribution of fish prices, and the prediction of fish prices are vital. In Taiwanese fisheries, however, predicting aquatic prices is challenging due to their drastic fluctuations caused by tropical and subtropical climate variations, export/import fish quantity changes, and political and economic situation uncertainty. Therefore, instead of using a single machine learning model to address the price fluctuations, we propose a hybrid fish price prediction model integrating multiple machine learning and deep learning models to identify and predict fish price dynamics in various wholesale markets. The extensive experimental results on actual market data from the Taiwan Council of Agriculture (COA) show that the proposed models can achieve >90% accuracy in predicting future aquatic prices. Additionally, we developed a fully automated learning and prediction system architecture on the cloud, allowing us to acquire data automatically and fine-tune the AI models continuously to achieve better performance with long-term operability.

Optimization of recycled rubber self-compacting concrete: Experimental findings and machine learning-based evaluation

This research aims to assess the rheological and mechanical characteristics of Self-compacting concrete (SCC) incorporating waste tire rubber aggregates (WRTA) as an interim substitute for coarse aggregates. However, the standard experimental modeling approach has significant obstacles when it comes to overcoming the nonlinearity and environmental susceptibility of concrete parts. Therefore, linear regression (LR) and extreme gradient boosting (XGBoost) were used as two standard single machine learning (ML) models to predict the aforementioned rubberized SCC features. In this study, conventional coarse aggregates were supplanted with WRTA at 0%, 5%, 10%, and 20% to uncover the optimal proportion of coarse aggregates substituting rubber. To find the optimum amount of WRTA to use as a substitute, the study follows the impacts of rubber on the self-compacting rubberized concrete's (SCRC) rheological and mechanical characteristics. The consequences on fresh properties were investigated by the slump flow, J-ring, and V-funnel tests, while compressive and splitting tensile strengths tests were conducted to assess mechanical properties. Increasing WRTA test outputs indicated a deterioration in workability and hardened qualities. While a 10% swapping ratio is deemed feasible for producing SCRC, optimal results were achieved by reducing environmental impacts and efficiently managing a significant volume of rubber tire waste with a 5% substitution of rubber within the coarse aggregates. The research findings indicated a noticeable decrease in fresh properties as the WRTA content increased. Notably, after 28 days, a 10% WRTA substitution led to a 34% reduction in compressive strength and a 28% decrease in splitting tensile strength, satisfying ACI standards. Furthermore, XGBoost demonstrated superior predictive performance with the highest R2 values, outperforming the LR model and affirming its efficacy in delivering more accurate predictions.

Towards a Reliable Design of Geopolymer Concrete for Green Landscapes: A Comparative Study of Tree-Based and Regression-Based Models

The design of geopolymer concrete must meet more stringent requirements for the landscape, so understanding and designing geopolymer concrete with a higher compressive strength challenging. In the performance prediction of geopolymer concrete compressive strength, machine learning models have the advantage of being more accurate and faster. However, only a single machine learning model is usually used at present, there are few applications of ensemble learning models, and model optimization processes is lacking. Therefore, this paper proposes to use the Firefly Algorithm (AF) as an optimization tool to perform hyperparameter tuning on Logistic Regression (LR), Multiple Logistic Regression (MLR), decision tree (DT), and Random Forest (RF) models. At the same time, the reliability and efficiency of four integrated learning models were analyzed. The model was used to analyze the influencing factors of geopolymer concrete and determine the strength of their influencing ability. According to the experimental data, the RF-AF model had the lowest RMSE value. The RMSE value of the training set and test set were 4.0364 and 8.7202, respectively. The R value of the training set and test set were 0.9774 and 0.8915, respectively. Therefore, compared with the other three models, RF-AF has a stronger generalization ability and higher prediction accuracy. In addition, the molar concentration of NaOH was the most important influencing factors, and its influence was far greater than the other possible factors including NaOH content. Therefore, it is necessary to pay more attention to NaOH molarity when designing geopolymer concrete.

Daily Runoff Prediction with a Seasonal Decomposition-Based Deep GRU Method

Accurately predicting hydrological runoff is crucial for water resource allocation and power station scheduling. However, there is no perfect model that can accurately predict future runoff. In this paper, a daily runoff prediction method with a seasonal decomposition-based-deep gated-recurrent-unit (GRU) method (SD-GRU) is proposed. The raw data is preprocessed and then decomposed into trend, seasonal, and residual components using the seasonal decomposition algorithm. The deep GRU model is then used to predict each subcomponent, which is then integrated into the final prediction results. In particular, the hyperparameter optimization algorithm of tree-structured parzen estimators (TPE) is used to optimize the model. Moreover, this paper introduces the single machine learning model (including multiple linear regression (MLR), back propagation (BP), long short-term memory neural network (LSTM) and gate recurrent unit (GRU)) and a combination model (including seasonal decomposition–back propagation (SD-BP), seasonal decomposition–multiple linear regression (SD-MLR), along with seasonal decomposition–long-and-short-term-memory neural network (SD-LSTM), which are used as comparison models to verify the excellent prediction performance of the proposed model. Finally, a case study of the Qingjiang Shuibuya test set, which considers the period 1 January 2019 to 31 December 2019, is conducted. Case studies of the Qingjiang River show the proposed model outperformed the other models in prediction performance. The model achieved a mean absolute error (MAE) index of 38.5, a Nash-Sutcliffe efficiency (NSE) index of 0.93, and a coefficient of determination (R2) index of 0.7. In addition, compared to the comparison model, the NSE index of the proposed model increased by 19.2%, 19.2%, 16.3%, 16.3%, 2.2%, 2.2%, and 1.1%, when compared to BP, MLR, LSTM, GRU, SD-BP, SD-MLR, SD-LSTM, and SD-GRU, respectively. This research can provide an essential reference for the study of daily runoff prediction models.

Unifying supervised learning and VAEs: coverage, systematics and goodness-of-fit in normalizing-flow based neural network models for astro-particle reconstructions

Neural-network based predictions of event properties in astro-particle physics are getting more and more common. However, in many cases the result is just utilized as a point prediction. Statistical uncertainties, coverage, systematic uncertainties or a goodness-of-fit measure are often not calculated. Here we describe a certain choice of training and network architecture that allows to incorporate all these properties into a single network model. We show that a KL-divergence objective of the joint distribution of data and labels allows to unify supervised learning and variational autoencoders (VAEs) under one umbrella of stochastic variational inference. The unification motivates an extended supervised learning scheme which allows to calculate a goodness-of-fit p-value for the neural network model. Conditional normalizing flows amortized with a neural network are crucial in this construction. We discuss how to calculate coverage probabilities without numerical integration for specific “base-ordered” contours that are unique to normalizing flows. Furthermore we show how systematic uncertainties can be included via effective marginalization during training. The proposed extended supervised training incorporates (1) coverage calculation, (2) systematics and (3) a goodness-of-fit measure in a single machine-learning model. There are in principle no constraints on the shape of the involved distributions, in fact the machinery works with complex multi-modal distributions defined on product spaces like Rn×Sm\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${\\mathbb {R}}^n \ imes {\\mathbb {S}}^m$$\\end{document}. The coverage calculation, however, requires care in its interpretation when the distributions are too degenerate. We see great potential for exploiting this per-event information in event selections or for fast astronomical alerts which require uncertainty guarantees.

Estimating dynamic compressive strength of rock subjected to freeze-thaw weathering by data-driven models and non-destructive rock properties

ABSTRACT The dynamic compressive strength (DCS) of frozen-thawed rock influences the stability of rock mass in cold regions, especially when rock masses are possibly disturbed by dynamic loads. Laboratory freeze-thaw weathering treatment is usually time-consuming, and the dynamic strength test is destructive. Therefore, this paper attempts to quickly predict the DCS of frozen-thawed sandstones using data-driven methods, non-destructive rock properties, and basic environmental parameters. The sparrow search algorithm (SSA), gorilla troops optimiser, and dung beetle optimiser were chosen to develop two hyperparameters in the random forest (RF). The classic RF, back propagation neural network, and support vector regression models were taken as the control group. These six models were developed to predict the DCS. Their prediction results were compared. Finally, the sensitivity analysis was carried out to assess the significance of all input variables. The results indicate that the SSA – RF model yields the best prediction result, and three optimised models have better performance than single machine-learning models. Strain rate, dry density, and wave velocity are found to be the three most important parameters in DCS prediction, which further indicates that there is also a strong correlation between the characteristic impedance of the rock and the DCS.

Performance comparison of machine learning models used for predicting subclinical mastitis in dairy cows: Bagging, boosting, stacking, and super-learner ensembles versus single machine learning models

Mastitis has a substantial impact on the dairy industry across the world, causing dairy producers to suffer losses due to the reduced quality and quantity of produced milk. A further problem, related to this issue, is the excessive use of antibiotics that leads to the development of resistance in different bacterial strains. The growing consumer awareness oriented toward food safety and rational use of antibiotics has promoted the search for new methods of early identification of cows that may be at risk of developing the disease. Subclinical mastitis does not cause any visible changes to the udder or milk, and therefore it is more difficult to detect than clinical mastitis. The collection of large amounts of data related to milk performance of cows allows using machine learning (ML) methods to build models that could be used for classifying cows into healthy and at risk of subclinical mastitis. The data used for the purpose of this study included information from routine milk recording procedures. The data set consisted of 19,856 records of 2,227 Polish Holstein-Friesian cows from 3 herds. The authors decided to use the approach of building ensemble ML models, in particular bagging, boosting, stacking and super learner models, and comparing them for accuracy of identification of disease-affected cows against single ML models based on the Support Vector Machines, Logistic Regression, Gaussian Naïve Bayes, k-Nearest Neighbors and Decision Tree algorithms. The models were trained and evaluated based on the information recorded for herd 1 and using an 80:20 train-test split ratio according to animal ID (to avoid data leakage). The information recorded for herds 2 and 3 was only used to evaluate on unseen data models developed using the herd 1 data set. Among the single ML models, the Support Vector Machines model was found to be the most accurate in predicting subclinical mastitis at subsequent test-day when used both for the training set (mean F1-score of 0.760) and the testing sets containing data for herds 1, 2 and 3 (F1-score of 0.778, 0.790 and 0.741 respectively). The Gradient Boosting model was found to be the best performing model among the ensemble ML models (F1-score of 0.762, 0.779, 0.791 and 0.723 for the training set and the testing sets respectively). The super learner model, featuring the most advanced design and Logistic Regression in the meta layer, achieved the highest mean F1-score of 0.775 during the cross-validation, however, it was characterized by a slightly worse prediction accuracy of the testing sets (mean F1-score of 0.768, 0.790 and 0.693 for herds 1, 2 and 3 respectively). The study findings confirm the promising role of ensemble ML methods that were found to be slightly superior with respect to most of the single ML models.

Experimental analysis of intrusion detection systems using machine learning algorithms and artificial neural networks

Since the invention of the internet for military and academic research purposes, it has evolved to meet the demands of the increasing number of users on the network, who have their scope beyond military and academics. As the scope of the network expanded maintaining its security became a matter of increasing importance. With various users and interconnections of more diversified networks, the internet needs to be maintained as securely as possible for the transmission of sensitive information to be one hundred per cent safe; several anomalies may intrude on private networks. Several research works have been released around network security and this research seeks to add to the already existing body of knowledge by expounding on these attacks, proffering efficient measures to detect network intrusions, and introducing an ensemble classifier: a combination of 3 different machine learning algorithms. An ensemble classifier is used for detecting remote to local (R2L) attacks, which showed the lowest level of accuracy when the network dataset is tested using single machine learning models but the ensemble classifier gives an overall efficiency of 99.8%.

Exchange rate prediction research based on LSTM-ELM hybrid model

The fluctuation of exchange rates holds paramount importance for a country's economic and trade activities. Due to the non-stationary and nonlinear structural characteristics of exchange rate time series, accurately predicting exchange rate movements is a challenging task. Single machine learning models often exhibit lower precision in exchange rate prediction compared to combined machine learning models. Hence, employing a combined model approach aims to enhance the predictive performance of exchange rate models. Both Long Short-Term Memory (LSTM) and Extreme Learning Machine (ELM) exhibit intricate structures, making their direct integration challenging. To address this issue, an innovative weighted approach is adopted in this study, combining LSTM and ELM models and further refining the combination weights using an improved Marine Predators Algorithm. This paper encompasses both univariate and multivariate prediction scenarios, employing two distinct allocation strategies for training and testing datasets. This is done to investigate the influence of different dataset allocations on exchange rate prediction. Finally, the proposed LSTM-ELM weighted combination exchange rate prediction model is compared with SVM, Random Forest, ELM, LSTM, and LSTM-ELM average combination models. Experimental results demonstrate that the LSTM-ELM weighted combination exchange rate prediction model outperforms the others in both univariate and multivariate prediction settings, yielding higher predictive accuracy and superior fitting performance. Consequently, the LSTM-ELM weighted combination prediction model proves to be effective in exchange rate forecasting.

Understanding imbalanced data: XAI & interpretable ML framework

There is a gap between current methods that explain deep learning models that work on imbalanced image data and the needs of the imbalanced learning community. Existing methods that explain imbalanced data are geared toward binary classification, single layer machine learning models and low dimensional data. Current eXplainable Artificial Intelligence (XAI) techniques for vision data mainly focus on mapping predictions of specific instances to inputs, instead of examining global data properties and complexities of entire classes. Therefore, there is a need for a framework that is tailored to modern deep networks, that incorporates large, high dimensional, multi-class datasets, and uncovers data complexities commonly found in imbalanced data. We propose a set of techniques that can be used by both deep learning model users to identify, visualize and understand class prototypes, sub-concepts and outlier instances; and by imbalanced learning algorithm developers to detect features and class exemplars that are key to model performance. The components of our framework can be applied sequentially in their entirety or individually, making it fully flexible to the user’s specific needs (https://github.com/dd1github/XAI_for_Imbalanced_Learning).

Enhancing Solar Radiation Forecasting in Diverse Moroccan Climate Zones: A Comparative Study of Machine Learning Models with Sugeno Integral Aggregation

Hourly solar radiation (SR) forecasting is a vital stage in the efficient deployment of solar energy management systems. Single and hybrid machine learning (ML) models have been predominantly applied for precise hourly SR predictions based on the pattern recognition of historical heterogeneous weather data. However, the integration of ML models has not been fully investigated in terms of overcoming irregularities in weather data that may degrade the forecasting accuracy. This study investigated a strategy that highlights interactions that may exist between aggregated prediction values. In the first investigation stage, a comparative analysis was conducted utilizing three different ML models including support vector machine (SVM) regression, long short-term memory (LSTM), and multilayer artificial neural networks (MLANN) to provide insights into their relative strengths and weaknesses for SR forecasting. The comparison showed the proposed LSTM model had the greatest contribution to the overall prediction of six different SR profiles from numerous sites in Morocco. To validate the stability of the proposed LSTM, Taylor diagrams, violin plots, and Kruskal–Wallis (KW) tests were also utilized to determine the robustness of the model’s performance. Secondly, the analysis found coupling the models outputs with aggregation techniques can significantly improve the forecasting accuracy. Accordingly, a novel aggerated model that integrates the forecasting outputs of LSTM, SVM, MLANN with Sugeno λ-measure and Sugeno integral named (SLSM) was proposed. The proposed SLSM provides spatially and temporary interactions of information that are characterized by uncertainty, emphasizing the importance of the aggregation function in mitigating irregularities associated with SR data and achieving an hourly time scale forecasting accuracy with improvement of 11.7 W/m2.

A bearing fault diagnosis method with improved symplectic geometry mode decomposition and feature selection

A rolling bearing fault diagnosis method based on improved symplectic geometry mode decomposition (SGMD) and feature selection was proposed to solve the problem of low fault identification due to the influence of noise on early bearing fault features. First, the SGMD SGMD is improved to enhance its robustness in decomposing signals with noise, then the time domain, frequency domain, and time-frequency features of each symplectic geometric component are extracted as feature vectors. Second, a comprehensive feature selection strategy is proposed to select the optimal subset of features that are conducive to fault classification. Finally, considering the problem of low classification accuracy of a single machine learning model, the AdaBoost-WSO-SVM model is constructed for fault classification using the AdaBoost algorithm of integrated learning. Experimental decomposition of complex signals with noise indicates that the improved SGMD is more effective compared to traditional SGMD. Subsequently, multiple experiments were conducted using the bearing datasets from Case Western Reserve University (CWRU) and Jiangnan University (JNU). The experimental results reveal that, after comprehensive feature selection and ensemble learning pattern recognition experiments on the CWRU dataset, the average accuracy of fault diagnosis can reach 99.67%. On the JNU dataset, the proposed fault diagnosis method achieves an average accuracy of 95.03%. This suggests that, compared to other feature selection methods and classification models, the proposed approach in this paper exhibits higher accuracy and generalization capabilities.

Fatigue life prediction of the FCC-based multi-principal element alloys via domain knowledge-based machine learning

We propose a machine learning (ML) model to predict the fatigue life of multi-principal element alloys (MPEAs) by extracting features from empirical formulas. The model is built on XGBoost and GBDT, and outperforms the single ML model, with almost all predictions lying in the ± 2 error bands and the relative error not exceeding 0.16 in the extrapolation test. Feature analysis shows that for the nine explored MPEAs systems, their S–N curves are more suitable to be fitted by logN=a+blogσmax. Interpretable analysis indicates that for the explored alloys, elongation > 47% benefits the increase of fatigue life; if their yield strength is less than 720 MPa, improving strength will favor improvement in lifetime, otherwise improving ductility will favor lengthening their lifetime. It provides a fast and low-cost method to predict the fatigue life of those FCC-based MPEAs, which guides designing alloys with longer fatigue life.