Multilayer Perceptron Research Articles

Advancing the prediction of crystalline phases in glass-ceramics via machine learning

Lithium aluminosilicate (LAS) glass-ceramics are widely utilized in diverse application, owing to their outstanding properties such as high transparency, high fracture toughness and ultra-low thermal expansion. The development of LAS glass-ceramics has traditionally relied on the phase diagram to identify primary crystalline phases. However, this approach is limited by constrained composition ranges and unknown heating treatment parameters, hindering the efficient development of high-performance glass-ceramics. In this study, we establish a comprehensive small-scale database of LAS glass-ceramics, comprising 751 samples characterized by 27 compositions, nucleation temperature, nucleation time, crystallization temperature, crystallization time and 13 crystalline phases. We employ five algorithms, i.e. Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Classification and Regression Trees (CART), K-Nearest Neighbors (K-NN) and Multi-Layer Perceptron (MLP) Classifier to predict the potential crystalline phases. Our results demonstrate that RF achieves the best overall performance, with the highest accuracy of 0.8609, the lowest hamming loss of 0.0142, and the highest micro F1 score of 0.9234. This work advances the understanding and prediction of crystalline phases in LAS glass-ceramics, providing valuable insights for the development and optimization of these materials.

Self-powered flexible wearable wireless sensing for outdoor work heatstroke prevention and health monitoring

Outdoor workers are susceptible to heatstroke in hot weather. Currently, there are no wearable devices on the market to assess the risk of heatstroke. This paper proposes a self-powered flexible wearable wireless sensing system for outdoor work heatstroke prevention and health monitoring (SWSM). SWSM consists of two parts, one part is a wearable heatstroke prevention patch based on flexible wireless sensing (WPFS) to monitor the wearer’s physiological indicators continuously. The other part is a self-powered device based on solar energy (SPDS), which can generate milliamp level current for powering the WPFS. With these indicators, the wearer’s activity category is confirmed by the multi-layer perceptron (MLP) model. WPFS contains a PVDF-HFD nanofibre cooling film (PNCF), flexible triboelectric nanogenerator (TENG)-based button (B-TENG), and functional sensing components. Temperatures under PNCF coverage can be up to 12% lower than environmental temperatures. B-TENG is used to adjust the mode. Based on the wearer’s physiological indicators a heatstroke risk assessment equation can be created to assess the potential risk of heatstroke. Finally, the data is displayed on the Internet of Things (IoT) platform. The system is wireless, self-powered, flexible wearable, and can be used for real-time prevention of heatstroke risk and health monitoring of outdoor workers.

Non-destructive detection of male and female information in ducklings based on near-infrared spectral wavelength selection and deep learning

The technology for sex identification in ducklings can contribute to increased revenue and cost savings in modern duck farming. However, traditional manual identification techniques require high skill levels and involve significant labor intensity. In this study, a non-destructive, user-friendly, and efficient duckling sex identification technique was proposed using near-infrared spectroscopy and deep learning algorithms. Spectral data from 600 groups of newly hatched ducklings were collected. These data were divided into training, testing, and validation sets in a ratio of 7:2:1. The raw spectral data was preprocessed using the Savitzky-Golay convolution derivative method, which was employed for subsequent spectral feature wavelength extraction and modeling. The characteristic wavelengths were extracted using competitive adaptive reweighted sampling (CARS), successive projections algorithm (SPA), and uninformative variable elimination combined with SPA (UVE-SPA). Conventional machine learning methods − support vector machine (SVM) and different deep learning models, including multilayer perceptron (MLP), mobile network version 2 (MobileNetV2), and residual neural network (ResNet), were studied and compared. The experiment showed that deep learning algorithms outperform traditional spectral analysis models in terms of classification performance. Furthermore, conducting feature wavelength extraction before constructing the classification model could reduce the model’s testing time and even improve its classification performance. Finally, the four models with better classification performance were validated using a validation set, and the combination of MobileNetV2 model UVE-SPA was selected as the optimized model for ducklings’ gender determination, with a classification accuracy of 98.3 % and an average validation time of 1.1 ms. In summary, the detection model established using near-infrared spectroscopy and MobileNetV2 can achieve non-destructive identification of the gender of ducklings. The findings can provide a preliminary research foundation and technical support for the subsequent design of related online intelligent detection systems.

Hyperspectral image classification with token fusion on GPU

Hyperspectral images capture material nuances with spectral data, vital for remote sensing. Transformer has become a mainstream approach for tackling the challenges posed by high-dimensional hyperspectral data with complex structures. However, a major challenge they face when processing hyperspectral images is the presence of a large number of redundant tokens, which leads to a significant increase in computational load, adding to the model’s computational burden and affecting inference speed. Therefore, we propose a token fusion algorithm tailored to the operational characteristics of the hyperspectral image and pure transformer network, aimed at enhancing the final accuracy and throughput of the model. The token fusion algorithm introduces a token merging step between the attention mechanism and the multi-layer perceptron module in each Transformer layer. Experiments on four hyperspectral image datasets demonstrate that our token fusion algorithm can significantly improve inference speed without any training, while only causing a slight decrease in the pure transformer network’s classification accuracy.

A neural network model for predicting the temperature of insulated overhead lines

In this work, we develop a neural network model based on experimental heating and cooling curves of insulated wires of overhead lines under changes in wind speed and its direction relative to the wire axis. Insulated SIP-3 wire used in overhead lines was investigated. A multilayer perceptron neural network was employed to model the process of heating and cooling of insulated wire under changes in wind speed and its direction. The average absolute error was taken as a criterion for evaluating the prediction accuracy of wire core and insulation temperatures. The main parameters of the developed neural network model include the number of hidden layers and neurons in each hidden layer, the degree of regulation, and the regulatory rigidity of the model. The modelled heating and cooling curves of insulated wire were compared with those obtained experimentally. The average absolute error was equal to 1.74 and –4.08°C for the predicted core and insulation temperatures, respectively. The difference between the heating curves at low wind speeds was found to range within 9°C. It was shown that an increase in wind speed leads to a decrease in the difference between the curves. Our analysis showed that neural network models used for predicting variations in the temperature of insulated overhead lines should be trained using a larger number of input parameters. This is the main prerequisite for high prediction accuracy of such models, when the difference between the simulated and experimental data does not exceed 5%.

Prediction of topside displacement of offshore platforms based on multi-source data fusion with multilayer perceptron neural network

Prediction of topside displacement of offshore platforms based on multi-source data fusion with multilayer perceptron neural network

Advancing source reactor-type discrimination using machine learning techniques and SFCOMPO-2.0 experimental database

In recent years, nuclear forensic analysis has become crucial due to the growing global threat of nuclear terrorism and smuggling. Since 2005, extensive research has been conducted on identifying the origin of spent nuclear fuel, focusing on the source reactor-type discrimination, 235U enrichment of the fresh fuel, and the fuel exposure in the reactor (known as burnup). However, the majority of research relies on computed databases, which may lead to tracing discrepancies compared with actual situations. The present study employs the isotopic measurements from the experimental SFCOMPO-2.0 database to predict nuclear reactor types using Factor Analysis (FA) and various machine learning classification algorithms. The results reveal that FA is an effective method for dimension reduction and visualization. The FA-KNN, Random Forest (RF), and Multilayer Perceptron (MLP) algorithms are applied using a consistent dataset partition to ensure unbiased comparisons. The prediction results based on 10-fold stratified cross-validation are quite promising and the Receiver Operating Characteristic (ROC) curves for multi-class classification confirm the excellent generalization ability of models. Therefore, the application of machine learning techniques is highly effective for reactor-type forensics analysis, especially for RF and MLP.

AI in the Courtroom: Enhancing Legal Decision-Making through Predictive Modelling

Legal outcome prediction or legal judgement prediction is the process of making predictions of the possible outcome of a legal proceeding based on its case file. This work focuses on the importance of allowing less privileged or marginalised groups to have better access to law services. This work discusses the implementation of 13 different deep learning models and the results obtained for each of them. The range of models employed encompasses a spectrum from basic perceptron models to more complex convolutional network models and encoder-based models. The models like multi-layer perceptron (MLP), convolutional neural network (CNN) model, BERT, LegalBERT, DistilBERT, RoBERTa and the different combinations of these encoder-based models are implemented. The input to these models is the case files and their corresponding judgement status. Four different classes of petitions and appeals, submitted to the court, were considered to train all the models. The efficiency of these models was tested in terms of various performance metrics like training accuracy, testing accuracy, training loss, validation loss and [Formula: see text]1 score.

CT-based Intra-thrombus and Peri-thrombus Radiomics for Prediction of Prognosis After Endovascular Thrombectomy: A Retrospective Study Across Two Centers.

Complications from endovascular thrombectomy (EVT) can negatively affect clinical outcomes, making the development of a more precise and objective prediction model essential. This research aimed to assess the effectiveness of radiomic features derived from pre-surgical CT scans in predicting the prognosis post- EVT in acute ischemic stroke patients. This investigation included 336 acute ischemic stroke patients from two medical centers, spanning from March 2018 to March 2024. The participants were split into a training cohort of 161 patients and a validation cohort of 175 patients. Patient outcomes were rated with the mRS: 0-2 for good, 3-6 for poor. A total of 428 radiomic features were derived from intra-thrombus and peri-thrombus regions in non-contrast CT and CT angiography images. Feature selection was conducted using a least absolute shrinkage and selection operator regression model. The efficacy of eight different supervised learning models was assessed using the area under the curve (AUC) of the receiver operating characteristic curve. Among all models tested in the validation cohort, the logistic regression algorithm for combined model achieved the highest AUC (0.87, with a 95% confidence interval of 0.81 to 0.92), outperforming other algorithms. The combined use of radiomic features from both the intra-thrombus and peri-thrombus regions significantly enhanced diagnostic accuracy over models using features from a single region (0.81 vs 0.70, 0.77), highlighting the benefit of integrating data from both regions for improved prediction. The findings suggest that a combined radiomics model based on CT imaging serves as a potent approach to assessing the prognosis following EVT. The logistic regression model, in particular, proved to be both effective and stable, offering critical insights for the management of stroke. AUC=area under the curve; EVT=endovascular thrombectomy; KNN=k-nearest neighbors; LASSO=least absolute shrinkage and selection operator; LightGBM=Light Gradient Boosting Machine; LR=logistic regression; MLP=multi-layer perceptron; RF=random forest; SVM=support vector machine; XGBoost=extreme gradient boosting.

New forest fire assessment model based on artificial neural network and analytic hierarchy process or fuzzy-analytic hierarchy process methodology for fire vulnerability map

Forest fires significantly disrupt global ecosystems. Many forecasting techniques predict fire activity and allocate prevention resources, but various factors are missing from the assessments, and multi-criteria decision approaches alone are insufficient. This work introduced a novel methodology combining artificial neural networks (ANN) with the analytical hierarchy process (AHP), fuzzy AHP multi-criteria methods, and spatial data to detect potential forest fire vulnerability areas using twenty variables. The results from AHP or fuzzy AHP, were processed using a multilayer perceptron with a backpropagation algorithm. The final ANN model has two objectives: first, to create a forest fire vulnerability with four classes (low, moderate, high, and very high), and second, to classify burned area sizes in regions highly or very highly vulnerable to fire. Evaluation metrics were also applied for validation. The fire model was tested using both literature review data and in situ observations. Comparative analysis showed that the burned area size model performed better than other machine learning methods, achieving an accuracy score of 89%. Meanwhile, the fire vulnerability model scored 82%. The study addresses the problem of prediction and provides an algorithm for classifying fire risk based on historical data in the Czech Republic, offering a master model for future forest management. Therefore, the ANN model, once trained, validated, and tested, does not require resetting and is effective for estimating forest fire vulnerability. Moreover, it produces results quickly, providing rapid insights into complex forest ecosystems, saving time and enhancing understanding for decision-makers.

Predicting the Risk of Maxillary Canine Impaction Based on Maxillary Measurements Using Supervised Machine Learning.

To predict palatally impacted maxillary canines based on maxilla measurements through supervised machine learning techniques. The maxilla images from 138 patients were analysed to investigate intermolar width, interpremolar width, interpterygoid width, maxillary length, maxillary width, nasal cavity width and nostril width, obtained through cone beam computed tomography scans. The predictive models were built using the following machine learning algorithms: Adaboost Classifier, Decision Tree, Gradient Boosting Classifier, K-Nearest Neighbours (KNN), Logistic Regression, Multilayer Perceptron Classifier (MLP), Random Forest Classifier and Support Vector Machine (SVM). A 5-fold cross-validation approach was employed to validate each model. Metrics such as area under the curve (AUC), accuracy, recall, precision and F1 Score were calculated for each model, and ROC curves were constructed. The predictive model included four variables (two dental and two skeletal measurements). The interpterygoid width and nostril width showed the largest effect sizes. The Gradient Boosting Classifier algorithm exhibited the best metrics, with AUC values ranging from 0.91 [CI95% = 0.74-0.98] for test data to 0.89 [CI95% = 0.86-0.94] for crossvalidation. The nostril width variable demonstrated the highest importance across all tested algorithms. The use of maxillary measurements, through supervised machine learning techniques, is a promising method for predicting palatally impacted maxillary canines. Among the models evaluated, both the Gradient Boosting Classifier and the Random Forest Classifier demonstrated the best performance metrics, with accuracy and AUC values exceeding 0.8, indicating strong predictive capability.

Forecasting interval carbon price through a multi-scale interval-valued decomposition ensemble approach

This paper proposes a novel Multi-scale Interval-valued Decomposition Ensemble (MIDE) framework for forecasting European Union Allowance (EUA) carbon futures prices, which integrates Noise-assisted Multivariate Empirical Mode Decomposition (NAMEMD), Interval-valued Vector Auto-Regressive (IVAR) model, Interval Event Analysis (IEA) method, and Interval Multi-Layer Perceptron (IMLP). First, the original interval-valued carbon prices with other interval-valued control variables are decomposed and integrated into high, medium, and low-frequency components by NAMEMD. Second, IVAR is used to investigate the dynamics of the interval-valued vector system in low-frequency components, while IMLP is employed to characterize the high-frequency components. Besides, the interval event analysis investigates typical events that significantly impact carbon prices in the medium-frequency component. Furthermore, empirical findings indicate that our proposed MIDE learning approach significantly outperforms some other benchmark models in out-of-sample forecasting.

Aeroacoustic analysis of fixed- and variable-pitch controlled multirotors and noise reduction via rotor phase control

This research conducts a comparative analysis of the aerodynamic and aeroacoustic characteristics of fixed-pitch and variable-pitch controlled multirotors (i.e., revolution per minute and collective pitch control). The study encompasses hovering and forward flight conditions, considering wind gust effects. Specifically, high-resolution time-frequency analysis is conducted to investigate the interference and modulation effects of multirotor noise. The analysis reveals significant variations in spectral characteristics depending on the control method. Notably, the frequencies of tonal noise from variable-pitch controlled multirotor do not exhibit short-periodic modulation, in contrast to those of fixed-pitch controlled multirotor. Considering the distinctive spectral characteristics of variable-pitch controlled multirotor, this study explores the effectiveness of noise reduction in variable-pitch controlled multirotor by implementing rotor phase control methods. To validate the noise reduction in time series simulations, a deep learning model is employed to determine the dynamically optimized rotor phases. A multilayer perceptron neural network was trained to derive the optimal rotor phase angles over time, considering the target observer points and flight conditions. The results demonstrate a substantial reduction in noise at the target observer points.

Collaborative Training of Data-Driven Remaining Useful Life Prediction Models Using Federated Learning

Remaining useful life prediction models are a central aspect of developing modern and capable prognostics and health management systems. Recently, such models are increasingly data-driven and based on various machine learning techniques, in particular deep neural networks. Such models are notoriously “data hungry”, i.e., to get adequate performance of such models, a substantial amount of diverse training data is needed. However, in several domains in which one would like to deploy data-driven remaining useful life models, there is a lack of data or data are distributed among several actors. Often these actors, for various reasons, cannot share data among themselves. In this paper a method for collaborative training of remaining useful life models based on federated learning is presented. In this setting, actors do not need to share locally held secret data, only model updates. Model updates are aggregated by a central server, and subsequently sent back to each of the clients, until convergence. There are numerous strategies for aggregating clients’ model updates and in this paper two strategies will be explored: 1) federated averaging and 2) federated learning with personalization layers. Federated averaging is the common baseline federated learning strategy where the clients’ models are averaged by the central server to update the global model. Federated averaging has been shown to have a limited ability to deal with non-identically and independently distributed data. To mitigate this problem, federated learning with personalization layers, a strategy similar to federated averaging but where each client is allowed to append custom layers to their local model, is explored. The two federated learning strategies will be evaluated on two datasets: 1) run-to-failure trajectories from power cycling of silicon-carbide metal-oxide semiconductor field-effect transistors, and 2) C-MAPSS, a well-known simulated dataset of turbofan jet engines. Two neural network model architectures commonly used in remaining useful life prediction, long short-term memory with multi-layer perceptron feature extractors, and convolutional gated recurrent unit, will be used for the evaluation. It is shown that similar or better performance is achieved when using federated learning compared to when the model is only trained on local data.

Machine learning based modeling for estimation of drug solubility in supercritical fluid by adjusting important parameters

Here, we employed machine learning models to predict how well Capecitabine drug would dissolve in supercritical carbon dioxide as the green solvent. The vision is to investigate the drug suitability for processing of nanodrugs with enhanced bioavailability in the body. In the employed data set, P (pressure) and T (temperature) serve as inputs, and Y, the solubility, is the only output for building the models. This study uses DT (Decision Tree) and MLP (Multilayer perceptron) as the core models. However, the raw and individual form of conventional algorithms may not provide accurate and general results. Ensemble methods like boosting improve the model performance. Also, single and ensemble models mounted on these models have hyper-parameters whose optimization affects the final models. Meta-heuristic algorithms are popular for tuning hyper-parameters. In this research, we used a new hybrid framework by coupling the basic models with the Adaboost algorithm (as an ensemble method) and PO and CS algorithms (as optimizers) to obtain four different models. The MLP model boosted with Adaboost and tuned with PO algorithm showed the best fitting accuracy among all models. This model reduces the RMSE error rate to 1.71, MSE to 2.92, and MAE to 1.42.

Optimal estimation of SoC, SoH using machine learning models and design of control algorithm for active cell balancing in lithium ion batteries

Abstract Electric vehicles transform the automotive industry by replacing traditional vehicles powered by fossil fuels with less polluting and efficient vehicles. They are powered by rechargeable Li-ion batteries. While there are drawbacks associated with Li-ion battery technology, it's important to note that these challenges can be addressed or mitigated effectively. A battery management system ensures the tracking of all functions performed by the battery. An advanced battery management system should accurately estimate the battery's state of health and state of charge. This paper aims to develop machine learning models for the estimation of the state of health and state of charge and to implement cell balancing in a battery management system. A dataset of battery charging and discharging profiles was used to train various machine learning models for estimation. These methods are computationally less complex than conventional methods. In addition, a cell balancing algorithm is implemented to control the charging and discharging of individual cells in a battery pack and balance the state of charge of each cell. The machine learning solution is created using several machine learning models, including Gradient Boosting, Random Forest, Linear Regression, AdaBoost, and Multi-layer Perceptron. Among the models Gradient Boosting and Random Forest provide good MSE and R2 score. The use of machine learning algorithms for the assessment of state of health and charge estimation combined with the design of an efficient control algorithm for active cell balancing offers significant advancements in the battery management system to optimise the performance, reliability, and useful life of Li-ion batteries. The paper also presents a case study utilising a combination of deep learning-based SoC, SoH estimation algorithms in a simulated data set. A well-designed control algorithm for active cell balancing presents a holistic and effective approach to optimise the performance, reliability, and lifespan of Li-ion batteries.

Explainable Machine Learning Model for Chronic Kidney Disease Prediction

More than 800 million people worldwide suffer from chronic kidney disease (CKD). It stands as one of the primary causes of global mortality, uniquely noted for an increase in death rates over the past twenty years among non-communicable diseases. Machine learning (ML) has promise for forecasting such illnesses, but its opaque nature, difficulty in explaining predictions, and difficulty in recognizing predicted mistakes limit its use in healthcare. Addressing these challenges, our research introduces an explainable ML model designed for the early detection of CKD. Utilizing a multilayer perceptron (MLP) framework, we enhance the model’s transparency by integrating Local Interpretable Model-agnostic Explanations (LIME), providing clear insights into the predictive processes. This not only demystifies the model’s decision-making but also empowers healthcare professionals to identify and rectify errors, understand the model’s limitations, and ascertain its reliability. By improving the model’s interpretability, we aim to foster trust and expand the utilization of ML in predicting CKD, ultimately contributing to better healthcare outcomes.

Enhancing membrane fouling control in wastewater treatment processes through artificial intelligence modeling: research progress and future perspectives

AbstractMembrane filtration processes have demonstrated remarkable effectiveness in wastewater treatment, achieving high contaminant removal and producing high-quality effluent suitable for safe reuse. Membrane technologies play a primary role in combating water scarcity and pollution challenges. However, the need for more effective strategies to mitigate membrane fouling remains a critical concern. Artificial intelligence (AI) modeling offers a promising solution by enabling accurate predictions of membrane fouling, thus supporting advanced fouling mitigation strategies.This review examines recent progress in the application of AI models, with a particular focus on artificial neural networks (ANNs), for simulating membrane fouling in wastewater treatment processes. It highlights the substantial potential of ANNs, particularly the widely studied multi-layer perceptron (MLP) and other emerging configurations, to accurately predict membrane fouling, thereby enhancing process optimization and fouling mitigation efforts. The review discusses both the potential benefits and current limitations of AI-based strategies, analyzing recent studies to offer valuable insights for designing ANNs capable of providing accurate fouling predictions. Specifically, it provides guidance on selecting appropriate model architectures, input/output variables, activation functions, and training algorithms. Finally, this review highlights the critical need to connect research findings with practical applications in full-scale wastewater treatment plants. Key steps crucial to address this challenge have been identified, emphasizing the potential of AI modeling to revolutionize process control and drive a paradigm shift toward more efficient and sustainable membrane-based wastewater treatment.

Robust unsupervised 5D seismic data reconstruction on regular and irregular grids

Seismic data reconstruction has become a central focus in seismic data processing, addressing challenges posed by sparse sampling due to physical and budgetary constraints. The advent of 5D acquisition methodologies marks a significant advancement in the quality and completeness of seismic data sets. Most traditional 5D reconstruction methods commonly use the fast Fourier transform (FFT), requiring regular grids and preliminary 4D binning before 5D interpolation. Discrete Fourier transform and nonequidistant FFT can honor the original irregular coordinates. However, when using exact locations, these methods become computationally expensive. We introduce an unsupervised deep-learning methodology to learn a continuous function across the sampling points in seismic data, facilitating reconstruction on regular and irregular grids. The network comprises a multilayer perceptron with linear layers and element-wise periodic activation functions. It excels at mapping the input coordinates to the corresponding seismic data amplitudes without relying on external training sets. The network’s intrinsic low-frequency bias is crucial in prioritizing acquiring self-similar features over high-frequency and incoherent ones during training. This characteristic mitigates incoherent noise in seismic data, such as random and erratic components. To assess the robustness of the unsupervised reconstruction technique, we conduct comprehensive evaluations using synthetic data examples sampled regularly and irregularly, as well as field-data examples with and without binning. The findings demonstrate the efficacy of our deep-learning framework in achieving resilient and accurate seismic data reconstruction across diverse sampling scenarios.

Machine learning-based classification of varicocoele grading: A promising approach for diagnosis and treatment optimization.

Varicocoele is a correctable cause of male infertility. Although physical examination is still being used in diagnosis and grading, it gives conflicting results when compared to ultrasonography-based varicocoele grading. We aimed to develop a multi-class machine learning model for the grading of varicocoeles based on ultrasonographic measurements. Between January and May 2024, we enrolled unilateral varicocoele patients at an infertility clinic, assessing their varicocoele stages using the Dubin and Amelar system. We measured vascular diameter and reflux time at the testicular apex and the subinguinal region ultrasonography in both the supine and standing positions. Using these measurements, we developed four multi-class machine learning models, evaluating their performance metrics and determining which patient position and projection were most influential in varicocoele grading. We included 248 patients with unilateral varicocoele in the study, their average age was 26.61±4.95 years old. Of these, 212 had left-sided and 36 had right-sided varicocoeles. According to the Dubin and Amelar system, there were 66 grade I, 96 grade II, and 86 grade III varicocoeles. Among the models we created, the random forest (RF) model performed best, with an overall accuracy of 0.81±0.06, an F1 score of 0.79±0.02, a sensitivity of 0.69±0.02, and a specificity of 0.8±0.03. Vascular diameter measurement at the testicular apex in the supine position had the most impact on grading across all models. In support vector machine and multi-layer perceptron models, reflux time measurements from the subinguinal projection in the standing position contributed the most, while in RF and k-nearest neighbors models, measurements from the subinguinal projection in the supine position were the most influential. Machine learning methods have demonstrated superior accuracy in predicting disease compared to traditional statistical regressions and nomograms. These advancements hold promise for clinically automated prediction of varicocoele grades in patients. Tailored varicocoele grading for individuals has the potential to enhance treatment effectiveness and overall quality of life.