Artificial Neural Network Model Research Articles (Page 1)

Objectives: To investigate the effect of lime and quarry dust, both individually and in combination, on the geotechnical properties of Puducherry inland clay using Artificial Neural Network (ANN) modelling. Methods: Clay samples were treated individually with varying proportions of lime and QD (7%, 14%, 21%, and 28% by dry weight) as well as in combination to assess improvements in geotechnical behaviour. Laboratory tests, including Atterberg limits, Free Swell Index (FSI), compaction characteristics, direct shear test, and Unconfined Compressive Strength (UCS), were conducted to evaluate changes in soil properties. To forecast the parameters of stabilized soil, the ANN Simulink model was simulated using a neural network fitting tool after training. Findings: The experimental findings showed that the plasticity index was reduced by 25% and 37% with lime and QD stabilization, respectively. Lime- and QD-stabilized clay reduced the optimum moisture content by 20% and 35%, while maximum dry density increased by 10% and 35%, respectively. Cohesion was reduced by 28% in both cases. Regarding UCS, lime-stabilized clay showed an increase up to 21% addition before declining, whereas QD-stabilized clay showed continuous strength gain. FSI decreased by 35% and 28% in lime- and QD-stabilized clay, respectively. The combination of both lime and QD showed superior performance due to synergistic effects. ANN modelling with statistical indicators (R2: 0.95–0.99, RMSE <30%, MAPE <20%) effectively predicted geotechnical properties with less than 25% error. Novelty: Utilizing QD provides a sustainable alternative to lime while improving the geotechnical performance of clay soil comparable to lime. Using QD as a stabilizer also helps in addressing environmental waste disposal issues. Keywords: Stabilization, Artificial neural network, Lime, Quarry dust, Simulink model

Abstract In the last few years, hybrid photovoltaic-thermal (PVT) collectors have become an attractive subject of research because of their ability to convert solar radiation into both electrical and thermal energies. Nonlinear relationships among their control variables, such as design parameters, climatic conditions, heat transfer fluid type, and electrical and thermal performances, require advanced modeling methodologies. This review examines the application of machine learning, especially artificial neural networks (ANNs), in photovoltaic-thermal systems. The paper begins with the state of the art in PVT systems, covering types, applications, recent developments, and more. It then presents a detailed analysis of ANN models, including the General Regression Neural Network (GRNN), Elman Neural Network (ENN), Radial Basis Function Network (RBFN), Multilayer Perceptron (MLP), and Adaptive Neuro-Fuzzy Inference Systems (ANFIS). Furthermore, the review highlights the roles that these models have played in enhancing PVT system performance in previous studies and includes a literature analysis to identify research gaps in this field. According to the literature, ANNs are valuable tools for predicting and optimizing the performance of PVT collectors; however, further exploration of alternative ANN models in novel PVT designs, combined with optimization algorithms, is necessary.

In agricultural systems, soil pH and electrical conductivity (EC) are crucial chemical properties that directly affect nutrient availability and microbial activity, but the challenging environment of the Peruvian Andes has limited research on their estimation. This study aimed to develop an ensemble learning method to predict soil pH and EC in Andean agroecosystems using environmental predictors. By using simple and weighted averaging, we developed a heterogeneous ensemble learning approach that integrates machine learning (ML) algorithms, including Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest (RF), and Extreme Gradient Boosting (XGBoost). The weighted ensemble assigns weights to models based on their predictive accuracy, measured by R² from spatial cross-validation. Spatial patterns are noticeable, and pH displays greater spatial clustering than EC. Elevation was the most important predictor in ML models for both parameters. Ensemble models significantly outperformed individual models, with the weighted ensemble achieving R² >0.93 and reducing RMSE by approximately 72%. Among standalone models, RF and XGBoost performed best for pH, while SVM performed the best for EC. ANN models were the least effective. Uncertainty analysis indicated high confidence in pH predictions but moderate to high uncertainty in EC predictions, suggesting that EC is more challenging to predict. Ensemble models with optimized weighting provide robust and accurate mapping of spatially autocorrelated soil properties. The high-confidence pH maps are reliable for soil management decisions, while EC predictions, though more uncertain, effectively identify priority areas for future sampling and investigation.

Older adults with head and neck squamous cell carcinoma (HNSCC) are underrepresented in clinical trials, limiting evidence-based treatment decisions. Artificial neural networks (ANNs) have demonstrated the ability to personalize treatment recommendations using patient-specific characteristics. To develop and externally validate ANNs for overall survival (OS) and progression-free survival (PFS) in older adults with HNSCC undergoing definitive chemoradiation. This international cohort study included retrospective clinical data from 19 academic cancer centers across Germany, Switzerland, Czech Republic, Cyprus, and the US from the SENIOR registry. ANNs were developed and validated using data from patients 65 years and older with locoregionally advanced HNSCC treated with definitive chemoradiation. Exclusion criteria included induction or adjuvant chemotherapy, history of head and neck cancer, and metastatic disease at treatment initiation. Data were collected from January 2021 to December 2023, and data were analyzed from December 2023 to April 2025. All patients received definitive radiotherapy with concurrent systemic therapy between 2005 and 2019. OS and PFS were predicted using 2 separate ANN models. Patients were classified as high or low risk based on median prediction thresholds. Model performance was assessed with receiver operating characteristic (ROC) area under the curve (AUC) and precision recall AUC. Model explainability was assessed with Shapley additive explanations values. Of 898 patients included in the OS analysis (738 in training cohort and 160 in testing cohort), 665 (74.1%) were male, and the median (IQR) age was 71 (68-76) years. Of 945 included in the PFS analysis (770 in training cohort and 175 in testing cohort), 696 (73.7%) were male, and the median (IQR) age was 71 (68-76) years. The OS ANN stratified patients into high-risk and low-risk groups with significantly different survival, achieving an ROC-AUC of 0.68 (95% CI, 0.60-0.76). The PFS ANN showed similar discrimination, with an ROC-AUC of 0.64 (95% CI, 0.56-0.72). Human papillomavirus status, kidney function (estimated glomerular filtration rate), Eastern Cooperative Oncology Group Performance Status score, and nodal classification were among the most predictive features. In this study, ANN-based models using routine clinical data effectively stratified older adults with HNSCC into prognostic groups. Integration of ANNs into clinical workflows could support personalized treatment decisions for this vulnerable population.

The objective of this study was to analyze the impact of post-mastectomy radiotherapy (PMRT) in male breast cancer (MBC) patients and develop an artificial neural network (ANN) model to identify a potential PMRT benefit population. Data from a total of 2,247 MBC patients with T1-2N0-1M0 who underwent total mastectomy between 1998 and 2016 were enrolled from the SEER database. Propensity score matching was used to reduce covariate imbalances. Cox regression analysis was conducted to compare overall survival (OS) between the PMRT and no-PMRT groups. The hypothesis was that patients who had undergone PMRT and lived longer than the median OS of the no-PMRT group could benefit from PMRT. An ANN model was then developed to predict PMRT benefit population. Multivariate Cox regression analysis demonstrated better OS in the PMRT group compared to the no-PMRT group of matched patients. This survival benefit was particularly significant in patients with grade III or T2N0 and T2N1 disease, while no significant difference was observed in patients with grade I/II or T1N0 and T1N1 disease. An ANN model was established to predict PMRT benefit population based on patients with T2N0/T2N1. The optimal cut-off value for the model predicted probability was 0.51. Survival curves indicated that a score of 0.51 could accurately distinguish potential PMRT benefit population. For MBC patients with T2N0, T2N1, and grade III, PMRT would improve survival. The ANN model would be used to identify patients who are likely to benefit from PMRT and aid in clinical decision-making.

The growing interest in sustainable construction materials has prompted the investigation of alternative resources and sophisticated predictive techniques to enhance material performance. Waste foundry sand (WFS), a secondary product resulting from the metal casting procedure, present a viable alternative to natural aggregates, while the cement strength class (CSC) plays a crucial role in determining the properties of mortar. Although considerable research has been conducted on these elements separately, their combined influence on the compressive strength of mortar has not been thoroughly examined. This study aims to explore the interactive effects of varying percentages of WFS and different CSCs on the compressive strength of cement mortar, utilizing Gene Expression Programming (GEP), a cutting-edge machine learning approach. Compared to Artificial Neural Network (ANN) and other Machine Learning (ML) models, GEP offers enhanced transparency and robust predictive accuracy, making it more suitable for data-driven decision-making in sustainable construction. A comprehensive experimental dataset was created by varying WFS percentages (0%, 10%, 20%, 30%, 40% and 50%) and CSCs (32.5, 42.5, 52.5MPa). The mix designs were evaluated under two conditions: random and sorted data modes, both with and without CSC as an input variable. GEP models were constructed to forecast compressive strength, incorporating WFS percentage, sand/cement ratio (S/C), water/cement ratio (W/C), and CSC as primary inputs. The addition of CSC as an input variable significantly improved predictive accuracy, achieving a high correlation coefficient (R = 0.99) and a low root mean square error (RMSE = 2.3). The results underscore the necessity of considering both WFS and CSC in tandem within predictive models to effectively optimize mortar mix designs. By merging sustainable materials with advanced modeling methodologies, this research aids in resource conservation and the creation of high-performance, eco-friendly construction materials. The study provides a solid framework for engineers and researchers to advance material design and sustainability within the construction sector.

This research focuses on introducing the gas sweetening unit and identifying its optimal operating conditions using advanced modeling techniques. By analyzing experimental data collected over 1227 days from a gas refinery's sweetening unit, the study investigates factors influencing energy consumption. Two methods, response surface methodology (RSM) and artificial neural networks (ANNs), were employed to model the process, with the ANNs utilizing a multilayer perceptron (MLP) and a radial basis function (RBF). The R² value for RSM was 0.930, whereas the ANN models achieved higher accuracy, with R² values of 0.981 for RBF and 0.986 for MLP. Performance metrics also favored MLP, which recorded a lower error value of 0.002 compared to RBF (0.0051), making MLP the preferred method. Using the optimized MLP model, it was predicted that with an input feed of 8.3 MMSCM, 30.3% DEA amine, and 12.7% methyl diethanol amine (MDEA), fuel consumption could be reduced to 17380.2 SCM - a saving of 12710 SCM. Finally, the accuracy of the MLP model's predictions was validated through simulations in Aspen HYSYS, further confirming its effectiveness in optimizing energy consumption.

Purpose This study aims to develop a robust and generalizable hybrid strategy for investigating impact behaviors of high-performance composites, significantly advancing composite impact engineering and demonstrating strong potential for applications in protective and structural systems. Design/methodology/approach With the hybrid framework that integrates experimental testing, finite element (FE) modeling and machine learning (ML), the study on the low-velocity impact behavior of Kevlar fiber-reinforced composites with thermoplastic polyurethane matrix was carried out: with the validation of the FE model by experiments, the numerical model was used to produce data to train the ML methods. Findings The best-performing Levenberg–Marquardt artificial neural network model achieved excellent agreement with FE simulation data, yielding a correlation coefficient R > 0.98 and a low mean squared error, which was also proven through experimental validation with satisfactory accuracy. Originality/value In the current work, the combined method with the FE model, experiments and ML was developed for low-velocity impact of thermoplastic composite materials. The damage process was investigated, while the accuracy of the proposed methodology was verified when compared to experimental outcomes.

Artificial neural network modeling of heat transfer in bioconvective Oldroyd-B nanofluids with nonlinear radiation and activation energy

A cerebral aneurysm may present irregularities associated with rupture risks. However, conventional morphological parameters are limited in evaluating the aneurysm irregularity. Although the mass moment of inertia has been devised for the irregularity evaluation, its performance still needs to be improved. In this study, three novel morphological indexes (NMIs) were devised based on the mass moment of inertia (ANI, aneurysm-to-neck index; AVI, aneurysm-to-vessel index; AII, aneurysm irregularity index) to effectively describe aneurysm irregularities. 456 patients with cerebral aneurysms (367 unruptured and 89 ruptured) were enrolled and their NMIs and the conventional morphological parameters were calculated for comparison. Artificial neural networks (ANNs) were trained with each parameter and then used to predict rupture risk. All NMIs were significantly higher in ruptured cases than in unruptured cases (p-values for :ANI, :AVI, and :AII were < 0.001, <0.001, and < 0.001, respectively). The highest performance for rupture risk prediction (sensitivity, 92.9%; specificity, 92.0%; and area under the receiver operating characteristic curve, 0.951) was obtained when the NMIs were considered in the ANN model. In particular, the :AII effectively described the aneurysm irregularities that could not be evaluated using conventional morphological parameters. The NMIs were effective in evaluating aneurysm irregularities, enabling timely prediction of an aneurysm rupture.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-22582-2.

Purpose The purpose of this research is to investigate the peristaltic transport of a chemically reactive nanofluid under dual-diffusive convection within an inclined flexible microchannel. The goal is to accurately predict coupled heat and mass transfer behaviors through an artificial neural network (ANN)-based predictive framework, offering a computationally efficient alternative for modeling environmentally responsive fluidic systems. Design/methodology/approach A Casson nanofluid model is formulated, incorporating magnetohydrodynamic forces, thermal radiation, viscous dissipation, double diffusion and cross-diffusion effects. The governing equations are simplified using lubrication theory, transformed into non-dimensional form and solved numerically via a bvp5c solver in MATLAB. Numerical datasets are then used to train a bayesian-optimized ANN (BAN-ANN) model across nine key parameter variations. Model performance is validated through regression metrics, error histograms, fitness plots and mean squared error evaluation. Findings Stronger magnetic fields and higher solutal Grashof numbers reduce fluid velocity, while temperature profiles are strongly influenced by Brownian motion and Dufour effects. Increasing the porosity parameter from 0.1 to 0.3 raises skin friction by 36.57%, whereas raising the Brinkman number from 0.1to 0.2 results in a 23.4% reduction in the Nusselt number. ANN predictions for heat and mass transfer rates closely align with numerical results, demonstrating excellent accuracy and generalization capability. Practical implications The study offers a framework for ANN-driven optimization of microscale fluid systems, contributing to improved design strategies for pollutant removal, energy-efficient cooling technologies and bioinspired microdevices operating under multiphysics conditions. Originality/value This work presents a novel BAN-ANN approach to simulate and predict chemically reactive nanofluid transport in magneto-thermal environments.

Abstract The ionosphere significantly impacts Global Navigation Satellite Systems (GNSS) positioning accuracy, particularly in regions with pronounced ionospheric irregularities and high solar activity, such as Egypt. These regions face more challenges in ionospheric modeling compared to the higher-latitude areas. This study provides a detailed assessment of different approaches to Total Electron Content (TEC) estimation in Egypt and their impact on GNSS positioning accuracy using single-frequency receivers. The approaches include the traditional Klobuchar model, Global Ionospheric Maps (GIMs), and the Artificial Neural Network (ANNTEC) model. This ANN model was developed to predict the TEC over Egypt using 10 years of GNSS observations and ionospheric data. The assessment process is performed in two scenarios: static and kinematic positioning. The positioning accuracy employing each approach is evaluated relative to reference coordinates tied to the International Terrestrial Reference Frame 2020 (ITRF-2020) through the nearest International GNSS Service (IGS) stations. The results indicate that the ANNTEC model outperforms the other approaches in both kinematic and static scenarios. In the static mode, the Root Mean Square Error (RMSE) value of the horizontal positions can be improved by 46.6 % and 24.5 % compared to using the Klobuchar model and the GIMs, respectively. In addition, in the kinematic mode, the RMSE value was reduced by 54.7 % and 25.8 % compared to using the Klobuchar model and the GIMs, respectively. These results demonstrate the potential of employing the ANNTEC model to enhance single-frequency GNSS positioning accuracy in Egypt.

Myocardial ischemia is a core pathological mechanism in diverse fatal diseases and can be triggered by multiple factors. Diagnosing early myocardial ischemia (EMI) caused by nontraditional factors (e.g., drugs or stress) remains challenging due to subtle histological changes and limited clinical awareness. Label-free infrared spectroscopic imaging of myocardial tissue enables the revelation of convergent ischemic signatures across diverse etiologies. Here, we present an artificial intelligence (AI)-based analytical strategy to investigate the molecular mechanisms underlying EMI, enabling effective diagnosis of myocardial ischemia triggered by multiple factors. The artificial neural network (ANN) model developed using infrared spectroscopic data enabled accurate diagnosis of EMI caused by traditional factors, such as obstructive coronary artery disease. The accuracy, precision, sensitivity, and the area under the curve (AUC) were 97.45%, 99.82%, 95.24%, and 0.9993, respectively. For the first time, the model's precise diagnostic capabilities were extended to nontraditional forms of ischemia, including drug-induced Kounis Syndrome (KS) and stress-induced Takotsubo Syndrome (TTS), with prediction scores greater than 84%. This etiology-agnostic strategy captures trigger-independent biomolecular signatures, overcomes the limitations of conventional histology, and enables diagnosis of a broader range of ischemic diseases. Our method highlights the potential of spectral histopathology with AI in diagnosing diverse diseases with similar pathological features, not only providing valuable insights into the application of AI in data analysis but also demonstrating distinctive advantages of infrared spectroscopic imaging in mechanistic investigations and disease diagnosis, thereby greatly advancing the field of spectral histopathological analysis.

Background: Heart failure has traditionally been classified as systolic vs diastolic, however acute heart failure (AHF) hospitalizations, often has various outcomes seen in bedside clinical medicine, driven by comorbidities, hemodynamic states, and initial presentation. Research: Question Can unsupervised machine learning (ML) identify distinct clinical phenotypes among adults hospitalized with acute heart failure and guide management strategies? Methods: We analyzed non-elective adult hospitalizations with AHF from the 2022 National Inpatient Sample (NIS). Five machine learning models were employed for prediction: Logistic Regression, Naive Bayes, Random Forest, XGBoost, and an artificial neural network (ANN) model. Unsupervised clustering was performed using K-means after principal component analysis (PCA) (k=4, Fig 1). Cluster visualization was enhanced via t-distributed stochastic neighbor embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) on a 10,000-patient subsample (Fig 2). Clusters were profiled by demographic and clinical characteristics.All analysis was conducted using Python 3.11. Results: Four distinct patient clusters were identified amongst heart failure admissions. Cluster 0 (47.1%) comprised younger healthier patients (mean age 28.8 years) and lowest in-hospital mortality (1%) representing AHF cases with non-chronic triggers. Cluster 1(2.9%) consisted of older individuals, with coronary artery disease and interventions, such as CABG, PCI (mean age 74.1 years) and moderate mortality (3%) confirming an ischemic heart disease phenotype . Cluster 2 (35.3%) exhibited a cardiopulmonary/ metabolic profile with moderate mortality (3%, mean age 66.1 years) indicating a common heart failure phenotype with systemic involvement . Cluster 3 (14.6%) was dominated by arrhythmia and older population (mean age 74.5 years) and had the highest in-hospital mortality(6%) suggesting a high-risk arrhythmia-dominant phenotype (p &lt;0.001 for all).(Image 3) Conclusions: Unsupervised machine learning identified four unique and clinically relevant phenotypes among adults hospitalized with AHF, differentiated by comorbidity burden and in-hospital mortality. Our study highlights the potential of ML driven phenotyping to inform risk stratification, and clinical decisions in acute heart failure management.

The composite materials are widely used across industries however, these materials are prone to damages like cracking and delamination due to its complexity. The Electromechanical Impedance (EMI) technique offers a reliable non-destructive solution for detecting such damage using piezoelectric sensors and enabling effective structural health monitoring and enhancing safety and durability. This study explores the application of the EMI technique for monitoring damages in composite fibre concrete specimens. The specimens were prepared using Ordinary Portland Cement (OPC), fly ash, and polypropylene, glass fiber mixture, water, fine and coarse aggregates. The Piezoelectric sensors were employed to record conductance and susceptance signatures, enabling early detection and quantification of damages. The severity of damages were assessed using statistical indices such as Root Mean Square Deviation (RMSD), Mean Absolute Percentage Deviation (MAPD), and Correlation Coefficient (CC) revealing higher sensitivity. A notable leftward shift in EMI signatures with increasing damage was confirmed progressive structural degradation. Additionally, structural parameters equivalent stiffness and equivalent damping were evaluated, demonstrating a decrease in stiffness and an increase in damping with greater damage depth. Temperature effects on EMI responses were also investigated, necessitating compensation for reliable analysis. An Artificial Neural Network (ANN) model was trained using Levenberg-Marquardt (LM) algorithm and implemented to predict conductance values and damage depth. The developed ANN showed high accuracy, with strong agreement between experimental and predicted results. Overall, the findings confirm the EMI technique’s potential for SHM of composite fiber concrete and integration with machine learning for improved predictive its durability assessment.

Due to their complex and unpredictable nature, stock market movements were always challenging to predict. Factors like economic indicators, market sentiment, and political and global events significantly contribute to stock price unpredictability. There are different methods to analyze risks, returns, and average price movements, based on which investors make assumptions. Identifying patterns and making the right decision on large amounts of data is very difficult, but nowadays, with the advancement of neural networks, we can solve prediction problems by identifying patterns of high-dimensional sequential data. We will analyze and compare five neural network architectures, including Recurrent Neural Networks (RNNs), Long Short-Term Memory networks (LSTMs), Gated Recurrent Units (GRUs), Convolutional Neural Networks (CNNs), and Artificial Neural Networks (ANNs), to try to predict stock prices using historical data taken from Yahoo Finance API, which is widely used and reliable for financial data analysis. We will separate historical data into two parts, 80% of which will be trained and 20% will be tested. For each model, we will use different hyperparameters we selected as the most effective training. Popular Python libraries such as TensorFlow, Keras, and NumPy are used for efficient implementation. Additionally, we used preprocessing for data, such as data cleaning and normalization, to avoid errors and enhance model performance. The models are evaluated based on prediction accuracy using metrics like Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R-squared (R²). Additionally, we use classification metrics such as the confusion matrix and Receiver Operating Characteristic - Area Under the Curve (ROC-AUC) to analyze each model’s performance in predicting price movement directions. We concluded that the GRU model achieves the highest accuracy and reliability in our analysis, with notable performance in classification metrics. Conversely, the simple ANN model shows the worst results, highlighting the variability in predictive capabilities across different neural network architectures.

New artificial neural network models for risk assessment of skin sensitization using amino acid derivative assay, KeratinoSens™, human cell line activation test and in silico structural alert parameter.

Development of MgFe layered double hydroxide-modified biochar for efficient nitrate adsorption and potential slow-release fertilizer: Mechanism and artificial neural network modeling.

Predicting microplastic accumulation zones and shoreline changes along the Kelantan coast, Malaysia, using integrated GIS and ANN models.

This study presents an integrated nonlinear, memory-driven model to analyze the hemodynamic and thermal behavior of blood flow through multi-stenosed arteries under physiological conditions. Blood is modeled as a viscoelastic, non-Newtonian fluid using the fractional Jeffery fluid model with Caputo–Fabrizio derivatives to capture memory-dependent characteristics. A ternary hybrid nanofluid with gold (Au), silver (Ag), and multi-walled carbon nanotubes is employed to enhance thermal performance, with physiological factors, such as magnetic fields, thermal radiation, body acceleration, and metabolic heat generation incorporated to ensure biological relevance. The governing equations are solved analytically via Laplace and Hankel transforms to derive exact expressions for velocity and temperature fields. To complement and validate the analytical model, an artificial neural network (ANN) trained using the Levenberg–Marquardt algorithm is employed to predict key hemodynamic parameters, including skin friction and Nusselt number, under varying conditions. The ANN model is rigorously assessed through K-fold cross-validation, demonstrating high accuracy and generalization. Results reveal a significant influence of fractional-order and viscoelastic parameters on flow resistance and heat transfer, highlighting the potential of hybrid modeling in cardiovascular diagnostics and thermal therapies. This study exemplifies the synergy of theoretical modeling and intelligent data-driven methods in addressing complex problems in biomedical engineering.