Hybrid Model Research Articles

Measuring the effectiveness of hybrid diabetes care over 90 days through continuous data monitoring in type 2 diabetic patients

BackgroundDiabetes Mellitus, a global health challenge, affects 537 million individuals. Traditional management relies on periodic clinic visits, but technological advancements, including remote monitoring, offer transformative changes. Telemedicine enhances access, convenience, adherence, and glycemic control. Challenges include trust-building and limitations in face-to-face interactions. Integrating remote monitoring with in-person healthcare creates a hybrid approach. This study evaluates the impact on Type 2 Diabetes patients over 3 months.MethodsA retrospective case-control observational study. Inclusion criteria involved previous Type 2 Diabetes diagnosis and a minimum 3-month GluCare model period with two physical visits. Patients in the case group had in-clinic visits, bi-weekly app engagement, and monthly body weight readings. Control group had in-clinic visits only. Outcomes measured included HbA1c, lipid profile, CV risk, eGFR, urine Albumin/Creatinine Ratio, Uric Acid, and CRP.ResultsCase group showed significant HbA1c improvements (-2.19%), especially in higher baseline levels. Weight, BMI, LDL, total cholesterol, and CVD risk also improved. Controls showed smaller improvements. Higher digital interactions correlated with better outcomes. Patients with ≥11 interactions showed significant reductions in HbA1c (-2.38%) and weight (-6.00 kg).ConclusionThe GluCare.Health hybrid model demonstrates promising outcomes in Type 2 diabetes management. The integration of in-clinic consultations with continuous remote monitoring leads to substantial improvements in glycemic control and clinical parameters. The study highlights the importance of patient engagement in achieving positive outcomes, with higher digital interactions associated with greater reductions in HbA1c and weight. The hybrid approach proves more effective than digital-only interventions, emphasizing the need for comprehensive, end-to-end solutions in diabetes care.

A Hybrid Model for Carbon Price Forecasting Based on Improved Feature Extraction and Non-Linear Integration

Accurately predicting the price of carbon is an effective way of ensuring the stability of the carbon trading market and reducing carbon emissions. Aiming at the non-smooth and non-linear characteristics of carbon price, this paper proposes a novel hybrid prediction model based on improved feature extraction and non-linear integration, which is built on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), fuzzy entropy (FuzzyEn), improved random forest using particle swarm optimisation (PSORF), extreme learning machine (ELM), long short-term memory (LSTM), non-linear integration based on multiple linear regression (MLR) and random forest (MLRRF), and error correction with the autoregressive integrated moving average model (ARIMA), named CEEMDAN-FuzzyEn-PSORF-ELM-LSTM-MLRRF-ARIMA. Firstly, CEEMDAN is combined with FuzzyEn in the feature selection process to improve extraction efficiency and reliability. Secondly, at the critical prediction stage, PSORF, ELM, and LSTM are selected to predict high, medium, and low complexity sequences, respectively. Thirdly, the reconstructed sequences are assembled by applying MLRRF, which can effectively improve the prediction accuracy and generalisation ability. Finally, error correction is conducted using ARIMA to obtain the final forecasting results, and the Diebold–Mariano test (DM test) is introduced for a comprehensive evaluation of the models. With respect to carbon prices in the pilot regions of Shenzhen and Hubei, the results indicate that the proposed model has higher prediction accuracy and robustness. The main contributions of this paper are the improved feature extraction and the innovative combination of multiple linear regression and random forests into a non-linear integrated framework for carbon price forecasting. However, further optimisation is still a work in progress.

Developing interpretable machine learning model for evaluating young modulus of cemented paste backfill

Cemented paste backfill (CPB), a mixture of wet tailings, binding agent, and water, proves cost-effective and environmentally beneficial. Determining the Young modulus during CPB mix design is crucial. Utilizing machine learning (ML) tools for Young modulus evaluation and prediction streamlines the CPB mix design process. This study employed six ML models, including three shallow models Extreme Gradient Boosting (XGB), Gradient Boosting (GB), Random Forest (RF) and three hybrids Extreme Gradient Boosting-Particle Swarm Optimization (XGB-PSO), Gradient Boosting-Particle Swarm Optimization (GB-PSO), Random Forest-Particle Swarm Optimization (RF-PSO). The XGB-PSO hybrid model exhibited superior performance (coefficient of determination R2 = 0.906, root mean square error RMSE = 19.535 MPa, mean absolute error MAE = 13.741 MPa) on the testing dataset. Shapley Additive Explanation (SHAP) values and Partial Dependence Plots (PDP) provided insights into component influences. Cement/Tailings ratio emerged as the most crucial factor for enhancing Young modulus in CPB. Global interpretation using SHAP values identified six essential input variables: Cement/Tailings, Curing age, Cc, solid content, Fe2O3 content, and SiO2 content.

Predicting Ili River streamflow change and identifying the major drivers with a novel hybrid model

Study regionIli River, the main river in the Lake Balkhash basin. Study focusThis study introduces a novel hybrid model by coupling the Soil and Water Assessment Tool (SWAT) hydrological model with machine learning algorithms. It aims to simulate and forecast the streamflow of the Ili River, clearly delineating the roles of meteorological factors and anthropogenic factors in the process of streamflow change. New hydrological insights for the regionDuring the period 1960–2020, the contribution of climate change to streamflow variation was 104.59%-113.07%, the contribution of land use ranged from −10.75% to −4.59%, and the impact of reservoir construction was −2.27%. The predictive outcomes of the hybrid model indicate that, under the SSP2–4.5 and SSP5–8.5 scenarios, the streamflow of the Ili River is projected to increase by 12.8% and 14.3% respectively in the future period (2021–2100), in comparison to the historical period (1960–2020). The warming and humidification from 2021 to 2100 will lead to changes in the streamflow components of the Ili River, with a decrease in the proportion of groundwater flow and an increase in the proportion of surface flow. The results of this study provide a reference for the rational utilization and scientific management of water resources in the Ili River Basin.

Interpretable and explainable hybrid model for daily streamflow prediction based on multi-factor drivers.

Streamflow time series data typically exhibit nonlinear and nonstationary characteristics that complicate precise estimation. Recently, multifactorial machine learning (ML) models have been developed to enhance the performance of streamflow predictions. However, the lack of interpretability within these ML models raises concerns about their inner workings and reliability. This paper introduces an innovative hybrid architecture, the TCN-LSTM-Multihead-Attention model, which combines two layers of temporal convolutional networks (TCN) followed by one layer of long short-term memory (LSTM) units, integrated with a Multihead-Attention mechanism for predicting streamflow with streamflow causation-driven prediction samples (RCDP), employing local and global interpretability studies through Shapley values and partial dependency analysis. The find_peaks method was used to identify peak flow events in the test dataset, validating the model's generality and uncovering the physical causative patterns of streamflow. The results show that (1) compared to the LSTM model with the same hyperparameter settings, the proposed TCN-LSTM-Multihead-Attention hybrid model increased the R2 by 52.9%, 2.5%, 43.1%, and 10.7% respectively at four stations in the test set predictions using RCDP samples. Moreover, comparing the prediction results of the hybrid model under different samples in Hengshan station, the R2 for RCDP increased by 5.06% and 1.22% compared to streamflow autoregressive prediction samples (RAP) and meteorological-soil volumetric water content coupled autoregressive prediction samples (MCSAP) respectively. (2) Historical streamflow data from the preceding 3days predominantly influences predictions due to strong autocorrelation, with flow quantity (Q) typically emerging as the most significant feature alongside precipitation (P), surface soil moisture (SSM), and adjacent station flow data. (3) During periods of low and normal flow, historical data remains the most crucial factor; however, during flood periods, the roles of upstream inflow and precipitation become significantly more pronounced. This model facilitates the identification and quantification of various hydrodynamic impacts on flow predictions, including upstream flood propagation, precipitation, and soil moisture conditions. It also elucidates the model's nonlinear relationships and threshold responses, thereby enhancing the interpretability and reliability of streamflow predictions.

Robustness and Interpretability of Machine Learning Models in Financial Forecasting

Purpose: The practice of financial forecasting is an essential component of contemporary financial management and the process of making decisions on investments. As a result of the complexity and volatility of financial markets, traditional financial forecasting techniques often fail to adequately capture the situation. Methodology/ Findings: One of the most effective methods for improving the precision and effectiveness of financial forecasting is the use of machine learning techniques. Through an examination of its potential, limitations, and future possibilities. They take linear and nonlinear possibilities into account in their study. Our focus is on linear methods, namely penalized regressions, and ensemble of models. The study considers a range of nonlinear techniques, including tree-based methods like boosted trees and random forests, and deep and shallow neural networks in both feed-forward and recurrent forms. They also think about ensemble and hybrid models, which combine features from several kinds of alternatives. Implications to Theory, Practice and Policy: A brief overview of the tests used to measure outstanding predictive ability is provided. In the last part of this article, they focus on the possible applications of machine learning in the fields of economics and finance, and we provide an example that makes use of high-frequency financial data (Benti, Chaka, and Semie, 2023).

Towards decision-making support for complex socio-technical system safety assessment: A hybrid model combining FRAM and dynamic Bayesian networks

Effectively managing system safety and resilience in critical infrastructures requires addressing emergent risks and critical resonances. This study introduces a quantitative model that merges Monte Carlo Simulation (MCS) of the Functional Resonance Analysis Method (FRAM) and Dynamic Bayesian Network (DBN) to assess technological, organizational, and human performance variability in complex social-technical systems. FRAM identifies system taxonomy and functional variability, while MCS pinpoints critical coupling, supplying prior probabilities for functional resonance. This information feeds into the DBN, facilitating the modeling of causal relationships and probabilistic inferences regarding risk uncertainties and system resonances. The model underwent rigorous testing and validation on a preheater cyclone system within the cement industry. This process involved the utilization of historical field data gathered from six prominent cement industries and input from thirty subject matter experts. The integrated approach deeply analyzes emerging risk indicators, unveiling interactions within organizational, human, and technical subsets, alongside performance variability. Furthermore, the model incorporates system learning parameters, aiding decision-making under uncertainty. These findings advance system safety and resilience management, offering insights for risk assessment and accident prevention across diverse scenarios in complex socio-technical systems.

Fake News Detection Using Hybrid Approach

Over the last few years, fake news has dramatically increased on social media. Fake news can originate from any number of sources and is shared across different social platforms. This type of information is used to spread for fun or economic gain. Our goal is to stop distributing this type of misleading information on social media or any other platform. In this paper, we have proposed a hybrid model (RoBERTa and BERT) to detect fake news. Our proposed architecture is based on the LIAR multi-label dataset. Our model shows promising results.

Pangenome and multi-tissue gene atlas provide new insights into the domestication and highland adaptation of yaks

BackgroundThe genetic diversity of yak, a key domestic animal on the Qinghai-Tibetan Plateau (QTP), is a vital resource for domestication and breeding efforts. This study presents the first yak pangenome obtained through the de novo assembly of 16 yak genomes.ResultsWe discovered 290 Mb of nonreference sequences and 504 new genes. Our pangenome-wide presence and absence variation (PAV) analysis revealed 5,120 PAV-related genes, highlighting a wide range of variety-specific genes and genes with varying frequencies across yak populations. Principal component analysis (PCA) based on binary gene PAV data classified yaks into three new groups: wild, domestic, and Jinchuan. Moreover, we proposed a ‘two-haplotype genomic hybridization model’ for understanding the hybridization patterns among breeds by integrating gene frequency, heterozygosity, and gene PAV data. A gene PAV-GWAS identified a novel gene (BosGru3G009179) that may be associated with the multirib trait in Jinchuan yaks. Furthermore, an integrated transcriptome and pangenome analysis highlighted the significant differences in the expression of core genes and the mutational burden of differentially expressed genes between yaks from high and low altitudes. Transcriptome analysis across multiple species revealed that yaks have the most unique differentially expressed mRNAs and lncRNAs (between high- and low-altitude regions), especially in the heart and lungs, when comparing high- and low-altitude adaptations.ConclusionsThe yak pangenome offers a comprehensive resource and new insights for functional genomic studies, supporting future biological research and breeding strategies.

Transfer learning based hybrid model for power demand prediction of large-scale electric vehicles

Accurately predicting the power demand of large-scale electric vehicles (EVs) is one of the key tasks of power grid operation optimization. However, this task is difficult to complete due to insufficient data and high randomness of power demand. To address this issue, this paper proposes a transfer learning based hybrid method for power demand prediction of large-scale EVs. Firstly, the linear trend of power demand is extracted by Multiple Linear Regression (MLR), and the nonlinear residual error is obtained by removing the trend from the original data. Secondly, the residual error is predicted via the bidirectional long short-term memory network (BiLSTM). Meanwhile, transfer learning is employed to improve the prediction accuracy of BiLSTM. The BiLSTM is pre-trained using the residual error data from building energy consumption and fine-tuned by the residual error data from EV power demand. Finally, the extracted trend of large-scale EVs’ power demand and the predicted residual errors are combined to obtain the final predicted results. To validate the proposed method, the power demand of a real EV charging station is predicted using the proposed method, and fifteen models are taken for comparison (including the most popular data-driven models). The experimental result indicates that the proposed method can improve the predictive performance by at least 14.23% compared with the comparative models in terms of symmetric mean absolute percentage error, effectively avoids negative transfer results, and significantly enhance the confidence level.

Enhancing Coronary Artery Disease Detection with a Hybrid Machine Learning Approach: Integrating K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) Algorithms

Recent studies have identified coronary artery disease (CAD) as a leading cause of death globally. Early detection of CAD is crucial for reducing mortality rates. However, accurately predicting CAD poses challenges, particularly in treating patients effectively before a heart attack occurs due to the complexity of data and relationships in traditional methodologies. This research has successfully developed a machine learning model for CAD prediction by combining K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) Classifier techniques. The model, trained and tested on a dataset of 918 samples (508 with cardiac issues and 410 healthy cases), achieved an accuracy of 82% for KNN, 84.3% for SVM, and 88.7% for the hybrid model after rigorous training and testing.

Forecast combination using grey relational analysis and Choquet fuzzy integral for container throughput forecasting

Utilizing the accurate and reliable demand forecasting model for the port logistics industry decision makers, can prevent misguided decisions ensuing from inaccurate forecasting. Current forecasting models of container throughput forecasting focus on the development of single or related hybrid models. However, these studies focus on improving the forecast accuracy, while neglecting the risk of failure. Combination forecasting can improve the forecast accuracy and reduce the risks from model selection failure. The previous research on port container throughput (PCT) combination forecasting focused on using a small number of linear models to generate the combined values, and neglected the nonlinear combined models reflecting the interaction between variables. Given the significance of combination forecasting for the PCT prediction, this study establishes a nonlinear combined model using the grey relational analysis (GRA) and Choquet fuzzy integral. The empirical results indicate that the proposed combined model outperforms both the other considered linear and nonlinear models. The findings assist the port decision-makers in both the public and private sectors for understanding the future PCT, and devising timely response plans aligned with actual circumstances and trends, besides formulating the appropriate development plans and decisions to modify the current situation or sustain competitive advantages.

Enhancing Cybersecurity Through AI-Driven Threat Detection: A Transfer Learning Approach

In today's digital landscape, fortifying cyber security is of utmost importance. This research introduces an innovative strategy that relies on transfer learning within deep neural networks to combat evolving threats, with a specific focus on phishing URLs and Malicious links, a major vector for cyber-attacks. We meticulously curate a diverse dataset of phishing and legitimate URLs, subjecting it to rigorous pre-processing. Departing from traditional methods, we leverage transfer learning to extract intricate patterns within URLs and their content. Our unique approach integrates transfer learning into a hybrid model, combining deep learning techniques with the power of transfer learning. This hybrid model employs soft and hard voting to optimize phishing threat detection accuracy and efficiency. We fine-tune our models with advanced feature selection and hyper parameter optimization, using rigorous evaluation metrics to assess performance

A cyber defense system against phishing attacks with deep learning game theory and LSTM-CNN with African vulture optimization algorithm (AVOA)

AbstractPhishing attacks pose a significant threat to online security, utilizing fake websites to steal sensitive user information. Deep learning techniques, particularly convolutional neural networks (CNNs), have emerged as promising tools for detecting phishing attacks. However, traditional CNN-based image classification methods face limitations in effectively identifying fake pages. To address this challenge, we propose an image-based coding approach for detecting phishing attacks using a CNN-LSTM hybrid model. This approach combines SMOTE, an enhanced GAN based on the Autoencoder network, and swarm intelligence algorithms to balance the dataset, select informative features, and generate grayscale images. Experiments on three benchmark datasets demonstrate that the proposed method achieves superior accuracy, precision, and sensitivity compared to other techniques, effectively identifying phishing attacks and enhancing online security.

Revealing forest structural "fingerprints": An integration of LiDAR and deep learning uncovers topographical influences on Central Amazon forests

Amazon forests are characterized by rich structural diversity. However, the influence of factors such as topography, soil attributes, and external disturbances on structural variability is not always well characterized, and traditional structural metrics may be inadequate to capture this type of complexity. While LiDAR offers expanded structural insights, traditional parameters used in LiDAR analysis, such as mean or maximum canopy height, are not always well directly linked to environmental variables like topography. Emerging approaches merge LiDAR with machine learning to uncover deeper structural complexities. However, work to date may fail to fully utilize the potential of fine-scale LiDAR information. Here we introduce a novel approach, leveraging 2D point cloud images derived from a profiling canopy LiDAR (PCL). The technique targets intricate details within LiDAR point clouds by using deep learning algorithms. With a dataset from the Central Amazon comprising 18 multitemporal transects of 450 m in length, our objective was to detect structural "fingerprints" of varied topographical types along a hillslope, comprising: Riparian, White-sand, and Plateau, and to detect any gradient of structural shifts based on terrain variations here represented by the height above the nearest drainage (HAND). The dataset was trained and tested using a leave-one-group-out approach (LOGO) in which, for each iteration, a complete 450 m multitemporal transect was excluded from training and tested after each iteration. The fast.ai platform and a ResNet-34 architecture, coupled with transfer learning, were used to perform a classification to distinguish between three topographical types. Furthermore, a hybrid model combining a Convolutional Autoencoder, and Partial Least Square (PLS) regression was designed to detect forest structural gradient correlations with HAND variation. Cross-validation achieved a promising high weighted F1 score of 0.83 to classify forests based on the topographical types. Additionally, a combined Convolutional Autoencoder and PLS regression revealed a strong correlation (R2 = 0.76) between actual and predicted HAND. Innovatively combining deep learning with ground-based PCL LiDAR, our study revealed unique Amazon Forest structures connected to topographic variation. Our findings underscore the transformative potential of such integrative approaches for investigating forest dynamics and promise a powerful new tool for understanding climate-related forest structure change.

Hybrid Transfer Learning and Support Vector Machine Models for Asphalt Pavement Distress Classification

Pavement condition evaluation plays a crucial role in assisting with the management of the highway infrastructure. However, the current methods used for assessing pavement conditions are costly, time-consuming, and subjective. There is a growing need to automate these assessment tactics and leverage low-cost technologies to enable widespread deployment. This study aims to develop robust and highly accurate models for classifying asphalt pavement distresses using transfer learning (TL) techniques based on pretrained deep learning (DL) networks. This topic has gained considerable attention in the field since 2015 when DL became the mainstream choice for various computer vision tasks. While progress has been made in TL model development, challenges persist in areas of accuracy, repeatability, and training cost. To tackle these challenges, this study proposes hybrid models that combine DL networks with support vector machines (SVMs). Three strategies were evaluated: single DL models using transfer learning (TLDL), hybrid models combining DL and SVM (DL+SVM), and hybrid models combining TLDL and SVM (TLDL+SVM). The performance of each strategy was assessed using statistical metrics based on the confusion matrix. Results consistently showed that the TLDL+SVM strategy outperformed the other approaches in accuracy and F1 scores, regardless of the DL network type. On average, the hybrid models achieved an accuracy of 95%, surpassing the 80% accuracy of the best single model and the 55% accuracy for DL+SVM without TL. The results clearly indicate that employing transfer-learned models as feature extractors, in combination with SVM as the classifier, consistently achieves exceptional performance.

Hybrid modeling of an ultracentrifugation process for separation of full and empty adeno-associated virus particles.

Ultracentrifugation is an attractive method for separating full and empty capsids, exploiting their density difference. Changes of the serotype/capsid, density of loading material, or the genetic information contained in the adeno-associated viruses (AAVs) require the adaptation of the harvesting parameters and the density gradient loaded onto the centrifuge. To streamline these adaptations, a mathematical model could support the design and testing of operating conditions.Here, hybrid models, which combine empirical functions with artificial neural networks, are proposed to describe the separation of full and empty capsids as a function of material and operational parameters, i.e., the harvest model. In addition, critical quality attributes are estimated by a quality model which is operating on top of the harvest model. The performance of these models was evaluated using test data and two additional blind runs. Also, a "what-if" analysis was conducted to investigate whether the models' predictions align with expectations.It is concluded that the models are sufficiently accurate to support the design of operating conditions, though the accuracy and applicability of the models can further be increased by training them on more specific data with higher variability.

Development of Hybrid Models Based on AlexNet and Machine Learning Approaches for Strip Steel Surface Defect Classification

Development of Hybrid Models Based on AlexNet and Machine Learning Approaches for Strip Steel Surface Defect Classification

Crystal Phase Ionic Liquids for Energy Applications: Heat Capacity Prediction via a Hybrid Group Contribution Approach.

In the selection and design of ionic liquids (ILs) for various applications, including heat transfer fluids, thermal energy storage materials, fuel cells, and solvents for chemical processes, heat capacity is a key thermodynamic property. While several attempts have been made to develop predictive models for the estimation of the heat capacity of ILs in their liquid phase, none so far have been reported for the ILs' solid crystal phase. This is particularly important for applications where ILs will be used for thermal energy storage in the solid phase. For the first time, a model has been developed and used for the prediction of crystal phase heat capacity based on extending and modifying a previously developed hybrid group contribution model (GCM) for liquid phase heat capacity. A comprehensive database of over 5000 data points with 71 unique crystal phase ILs, comprising 42 different cations and 23 different anions, was used for parameterization and testing. This hybrid model takes into account the effect of the anion core, cation core, and subgroups within cations and anions, in addition to the derived indirect parameters that reflect the effects of branching and distribution around the core of the IL. According to the results, the developed GCM can reliably predict the crystal phase heat capacity with a mean absolute percentage error of 6.78%. This study aims to fill this current gap in the literature and to enable the design of ILs for thermal energy storage and other solid phase applications.

Psychometric evaluation of the PCL-5: assessing validity, diagnostic utility, and bifactor structures

ABSTRACT Background: The changes DSM-5 brought to the diagnostic criteria for posttraumatic stress disorder (PTSD) resulted in revising the most widely used instrument in assessing PTSD, namely the Posttraumatic Checklist for DSM-5 (PCL-5). Objective: This study examined the psychometric properties of the Romanian version of the PCL-5, tested its diagnostic utility against the Structured Clinical Interview for DSM-5 (SCID-5), and investigated the latent structure of PTSD symptoms through correlated symptom models and bifactor modelling. Method: A total sample of 727 participants was used to test the psychometric properties and underlying structure of the PCL-5 and 101 individuals underwent clinical interviews using SCID-5. Receiver operating characteristic curve (ROC) analyses were performed to test the diagnostic utility of the PCL-5 and identify optimal cut-off scores based on Youden's J index. Confirmatory Factor Analyses (CFAs) and bifactor modelling were performed to investigate the latent structure of PTSD symptoms. Results: Estimates revealed that the PCL-5 is a valuable tool with acceptable diagnostic accuracy compared to SCID-5 diagnoses, indicating a cut-off score of >47. The CFAs provide empirical support for Anhedonia, Hybrid, and bifactor models. The findings are limited by using retrospective, self-report data and the high percentage of female participants. Conclusions: The PCL-5 is a psychometrically sound instrument that can be useful in making provisional diagnoses within community samples and improving trauma-informed practices.