Search

The Al 2 O 3 –SiO 2 porous ceramics (ASPC) prepared from phosphorus tailings usually showed a large fluctuation in material properties due to the complex composition of the raw material, which required extensive experimental exploration to optimize their properties. To solve the above problem, a novel deep learning strategy for automatically adjust hyperparameters had been proposed to predict the property of ASPC based on the small experimental datasets in this work. Initially, phosphate tailings and bauxite were used as raw materials. Active carbon, ranging from 0 to 15 wt%, was added as a pore-forming agent. After mixing the raw materials, they were shaped under a pressure of 5 MPa and sintered in the temperature range of 1000–1200 °C to produce ASPC, the density and compressive strength of the ASPC were measured to train the artificial neural network (ANN). An exhaustive search was conducted to identify the optimal combination of hyperparameters suitable for small datasets, then the influence of each hyperparameter on the training performance of ANN was analyzed. Subsequently, the trained model was used to predict the feature space of ASPC, and the key processes for regulating ASPC performance were explored, thereby the experimental schemes that meet target performance (density and compressive strength) were identified. Three groups of experimental schemes were predicted, whose properties were compared with the experimental ones. The results showed that the mean squared errors (MSE) of 0.03 and 0.85 for density and compressive strength, respectively. Furthermore, the ASPC exhibited excellent comprehensive properties when 70 wt% phosphorus tailings and 30 wt% bauxite were added, whose apparent porosity was 59.11 %, density was 1.16 g/cm3 and cold compressive strength was 6.49 MPa. And the predicted data trend influence of the pore-forming agent content and sintering temperature on the performance of ASPC is the same as the experimental data. This research provides valuable guidance for addressing the limited controllability of traditional experimental design processes and accelerate the preparation of ASPC from solid waste. [ABSTRACT FROM AUTHOR]

To identify patterns in big medical datasets and use Deep Learning and Machine Learning (ML) to reliably diagnose Cardio Vascular Disease (CVD), researchers are currently delving deeply into these fields. Training on large datasets and producing highly accurate validation results is exceedingly difficult. Furthermore, early and precise diagnosis is necessary due to the increased global prevalence of cardiovascular disease (CVD). However, the increasing complexity of healthcare datasets makes it challenging to detect feature connections and produce precise predictions. To address these issues, the Intelligent Cardiovascular Disease Diagnosis based on Ant Colony Optimisation with Enhanced Deep Learning (ICVD-ACOEDL) model was developed. This model employs feature selection (FS) and hyperparameter optimization to diagnose CVD. Applying a min–max scaler, medical data is first consistently prepared. The key feature that sets ICVD-ACOEDL apart is the use of Ant Colony Optimisation (ACO) to select an optimal feature subset, which in turn helps to upgrade the performance of the ensuring deep learning enhanced neural network (DLENN) classifier. The model reforms the hyperparameters of DLENN for CVD classification using Bayesian optimization. Comprehensive evaluations on benchmark medical datasets show that ICVD-ACOEDL exceeds existing techniques, indicating that it could have a significant impact on CVD diagnosis. The model furnishes a workable way to increase CVD classification efficiency and accuracy in real-world medical situations by incorporating ACO for feature selection, min–max scaling for data pre-processing, and Bayesian optimization for hyperparameter tweaking. [ABSTRACT FROM AUTHOR]

Solar radiation is a fundamental parameter for solar photovoltaic (PV) technology. Reliable solar radiation prediction has become valuable for designing solar PV systems, guaranteeing their performance, operational efficiency, safety in operations, grid dispatchment, and financial planning. However, high quality ground-based solar radiation measurements are scarce, especially for very short-term time horizons. Most existing studies trained machine learning (ML) models using datasets with time horizons of 1 h or 1 day, whereas very few studies reported using a dataset with a 1 min time horizon. In this study, a comprehensive evaluation of nine ensemble learning algorithms (ELAs) was performed to estimate solar radiation in Santo Domingo with a 1 min time horizon dataset, collected from a local weather station. The ensemble learning models evaluated included seven homogeneous ensembles: Random Forest (RF), Extra Tree (ET), adaptive gradient boosting (AGB), gradient boosting (GB), extreme gradient boosting (XGB), light gradient boosting (LGBM), histogram-based gradient boosting (HGB); and two heterogeneous ensembles: voting and stacking. RF, ET, GB, and HGB were combined to develop voting and stacking ensembles, with linear regression (LR) being adopted in the second layer of the stacking ensemble. Six technical metrics, including mean squared error (MSE), root mean squared error (RMSE), relative root mean squared error (rRMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and coefficient of determination (R2), were used as criteria to determine the prediction quality of the developed ensemble algorithms. A comparison of the results indicates that the HGB algorithm offers superior prediction performance among the homogeneous ensemble learning models, while overall, the stacking ensemble provides the best accuracy, with metric values of MSE = 3218.27, RMSE = 56.73, rRMSE = 12.700, MAE = 29.87, MAPE = 10.60, and R2 = 0.964. [ABSTRACT FROM AUTHOR]

The influence of ionic liquid (IL) characteristics, lignocellulosic biomass (LCB) properties, and process conditions on LCB pretreatment is not well understood. In this study, a total of 129 experimental data on LCB (grass, agricultural, and forest residues) pretreatment using imidazolium, triethylamine, and choline-amino acid ILs were compiled to develop machine learning (ML) models for cellulose, hemicellulose, lignin, and solid recovery. Following data imputation, a bilayer artificial neural network (ANN) and random forest (RF) regression, the two most widely adopted ML models, were developed. The full-featured ANN following Bayesian hyperparameter (HP) optimisation offered excellent fit on training (R2: 0.936–0.994), though cross-validation (R2CV) performance remained marginally poor, i.e. between 0.547 and 0.761. The fitness of HP-optimised RF models varied between 0.824 and 0.939 for regression, and between 0.383 and 0.831 in cross-validation. Temperature and pretreatment time had been the most important predictors, except for hemicellulose recovery. Bayesian predictor selection combined with HP optimisation improved the R2CV boundary for ANN (0.555–0.825), as well as for RF models (0.474–0.824). As predictive performance of the models varied depending on target response, use of a larger homogeneous dataset may be warranted. The predictive modelling framework for LCB pretreatment, developed in this study, can be extended to similar biochemical process systems. [ABSTRACT FROM AUTHOR]

Abstract To identify patterns in big medical datasets and use Deep Learning and Machine Learning (ML) to reliably diagnose Cardio Vascular Disease (CVD), researchers are currently delving deeply into these fields. Training on large datasets and producing highly accurate validation results is exceedingly difficult. Furthermore, early and precise diagnosis is necessary due to the increased global prevalence of cardiovascular disease (CVD). However, the increasing complexity of healthcare datasets makes it challenging to detect feature connections and produce precise predictions. To address these issues, the Intelligent Cardiovascular Disease Diagnosis based on Ant Colony Optimisation with Enhanced Deep Learning (ICVD-ACOEDL) model was developed. This model employs feature selection (FS) and hyperparameter optimization to diagnose CVD. Applying a min–max scaler, medical data is first consistently prepared. The key feature that sets ICVD-ACOEDL apart is the use of Ant Colony Optimisation (ACO) to select an optimal feature subset, which in turn helps to upgrade the performance of the ensuring deep learning enhanced neural network (DLENN) classifier. The model reforms the hyperparameters of DLENN for CVD classification using Bayesian optimization. Comprehensive evaluations on benchmark medical datasets show that ICVD-ACOEDL exceeds existing techniques, indicating that it could have a significant impact on CVD diagnosis. The model furnishes a workable way to increase CVD classification efficiency and accuracy in real-world medical situations by incorporating ACO for feature selection, min–max scaling for data pre-processing, and Bayesian optimization for hyperparameter tweaking.

This article addresses the bathing problem in multi-machine collaborative operations, proposing a method based on improved self-organizing iterative clustering. This approach circumvents the issues of traditional manual parameter setting in the self-organizing iterative clustering algorithm that is often inconvenient and non-intuitive. The proposed method allows multiple machines to autonomously adjust the parameters involved in the clustering process, given a small number of intuitive hyperparameters. The ultimate goal is to iterate toward reasonable editing results. Initially, this article focuses on selecting feature vectors for the multi-machine collaborative confrontation situation. It applies standardization and principal component analysis to high-dimensional multi-machine situation information to confirm the new vector space. This space mainly encompasses position information in three dimensions and speed information. Subsequently, the paper introduces the concept of neighborhood density discrimination from density clustering. This improves the merging and splitting operations of the traditional self-organizing iterative clustering method. It optimizes and reduces the artificial parameters involved in these operations, enhancing the intelligent autonomy for batch clustering tasks. Before optimization, artificial parameters primarily include the number of expected clusters, minimum number of points within a class, number of iterations, upper limit of standard deviation that limits data distribution within a class, and an allowable shortest distance indicator between classes. Post optimization, the artificial parameters are limited to the expected cluster quantity, minimum number of points, and the number of iterations within a single classification. These optimized parameters are relatively intuitive, and the algorithm output does not strongly correlate with the input parameters. Ultimately, the paper selects algorithm evaluation indicators, including Dunn, Davies–Bouldin, silhouette coefficient, and Calinski–Harabasz. It uses these to evaluate the proposed algorithms ISODATA+ and K-MEANS+, along with the original ISODATA algorithm, against multiple artificially synthesized data sets (completely random data, Gaussian-generated data, and sin-type data) and real-world scenarios. The experimental results suggest that while KMEANS+ shows significant advantages owing to multiple manually set hyperparameters, it requires constant debugging when adjusting parameters, which increases the complexity of the task. Compared with the original self-organizing iterative algorithm ISODATA, statistical results show that the improved algorithm has equivalent capabilities to the original algorithm. This demonstrates that the ISODATA+ algorithm maintains good clustering capabilities even after removing some artificial parameters. The batching results from actual scenario tests further illustrate the effectiveness of the improved self-organizing iterative clustering algorithm in specific application scenarios, demonstrating its practicability for future real-world applications.

Abstract Sarcopenic obesity (SO) is characterized by concomitant sarcopenia and obesity and presents a high risk of disability, morbidity, and mortality among older adults. However, predictions based on sequential neural network SO studies and the relationship between physical fitness factors and SO are lacking. This study aimed to develop a predictive model for SO in older adults by focusing on physical fitness factors. A comprehensive dataset of older Korean adults participating in national fitness programs was analyzed using sequential neural networks. Appendicular skeletal muscle/body weight was defined as SO using an anthropometric equation. Independent variables included body fat (BF, %), waist circumference, systolic and diastolic blood pressure, and various physical fitness factors. The dependent variable was a binary outcome (possible SO vs normal). We analyzed hyperparameter tuning and stratified K-fold validation to optimize a predictive model. The prevalence of SO was significantly higher in women (13.81%) than in men, highlighting sex-specific differences. The optimized neural network model and Shapley Additive Explanations analysis demonstrated a high validation accuracy of 93.1%, with BF% and absolute grip strength emerging as the most influential predictors of SO. This study presents a highly accurate predictive model for SO in older adults, emphasizing the critical roles of BF% and absolute grip strength. We identified BF, absolute grip strength, and sit-and-reach as key SO predictors. Our findings underscore the sex-specific nature of SO and the importance of physical fitness factors in its prediction.

Accurate wind speed forecasting is crucial for efficiently integrating of wind power into the electrical grid and ensuring a stable power supply. However, wind speed is inherently noisy and unpredictable, making it challenging to forecast accurately. This study investigates the effects of hyperparameters of Discrete Wavelet Transform (DWT) and Singular Spectrum Analysis (SSA) on the accuracy of wind speed forecasting using various prediction models. Our proposed method focuses on the optimisation of hyperparameters within the existing models, suggesting that significant untapped potential remains. Our study examines a wide range of wavelet function orders for DWT and varying trend ratio parameter for SSA, and evaluates their impact on the prediction accuracy using real data from thirteen locations in Jordan. Particularly, our investigation reveals that high-order Daubechies wavelets in DWT outperform low-order wavelets. The study also illustrates that optimal hyperparameters must be modified when changing the prediction model and the combination of DWT and SSA enhances prediction performance when the trend ratio is set to 90%. Our results demonstrate that fine-tuning data preprocessing techniques is essential for accurate wind speed prediction since hyperparameter tuning results in greater improvements in prediction accuracy than sophisticated prediction models alone. Our findings underscore the importance of leveraging data preprocessing techniques and hyperparameter tuning for accurate wind speed forecasting.

Diabetic eye detection has become a major concern across the globe, which could be effectively addressed by automated detection using a deep convolutional neural network (DCNN). CNN models have better detection and classification accuracy than other state-of-theart models. In this paper, a differential evolution (DE)-optimized CNN has been proposed for the single-step classification of diabetic retinopathy (DR) and glaucoma images. DE has been used to find out the optimized values of four hyper-parameters of CNN, i.e., the number of filters in the first layer, the filter size, the number. of convolution layers, and the number of strides. Simulation has been done using three publicly available datasets, and the accuracy obtained is 87.8%, 92.3%, and 88.7%, respectively, which outperforms other models. No other state-of-the-art model has used DE for hyper-parameter tuning in CNN models. Also, no other additional segmentation approach or handcrafted features have been used. The model has been kept simple to reduce computational costs. [ABSTRACT FROM AUTHOR]

Brain tumors are a leading cause of death globally, with numerous types varying in malignancy, and only 12% of adults diagnosed with brain cancer survive beyond five years. This research introduces a hyperparametric convolutional neural network (CNN) model to identify brain tumors, with significant practical implications. By fine-tuning the hyperparameters of the CNN model, we optimize feature extraction and systematically reduce model complexity, thereby enhancing the accuracy of brain tumor diagnosis. The critical hyperparameters include batch size, layer counts, learning rate, activation functions, pooling strategies, padding, and filter size. The hyperparameter-tuned CNN model was trained on three different brain MRI datasets available at Kaggle, producing outstanding performance scores, with an average value of 97% for accuracy, precision, recall, and F1-score. Our optimized model is effective, as demonstrated by our methodical comparisons with state-of-the-art approaches. Our hyperparameter modifications enhanced the model performance and strengthened its capacity for generalization, giving medical practitioners a more accurate and effective tool for making crucial judgments regarding brain tumor diagnosis. Our model is a significant step in the right direction toward trustworthy and accurate medical diagnosis, with practical implications for improving patient outcomes. [ABSTRACT FROM AUTHOR]

Suspended sediment load estimation is vital for the development of river initiatives, water resources management, the ecological health of rivers, determination of the economic life of dams and the quality of water resources. In this study, the potential of Feed Forward Neural Network (FFNN), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Shuffled Frog Leaping Algorithm (SFLA) models was evaluated for suspended sediment load (SSL) estimation in Yeşilırmak River. The heat map of Pearson correlation values of meteorological and hydrological parameters in 1973–2021, which significantly impacted SSL estimation, was examined to estimate SSL values. As a result of the analysis it was developed a prediction model with three different combinations of precipitation, stream flow and past SSL values (M1: streamflow, M2: streamflow and precipitation, M3: streamflow, precipitation, and SSL). The prediction accuracy of the models was visually compared with the Coefficient of Determination (R2), Bias Factor (BF), Mean Absolute Error (MAE), Mean Bias Error (MBE), Root Mean Square Error (RMSE), Akaike Information Criterion (AIC), Kling-Gupta Efficiency (KGE) statistical criteria and Bland-Altan plot, boxplot, scatter plot and line plot. Based on the analyses, the PSO-ANN model in the M1 model combination showed good estimation performance with an RMSE of 1739.92, MAE of 448.56, AIC of 1061.55, R2 of 0.96, MBE of 448.56, and BF of 0.29. Similarly, the SFLA-ANN model in the M2 model combination had an RMSE of 1819.58, MAE of 520.64, AIC of 1069.9, R2 of 0.96, MBE of 520.64, and BF of 0.19. In the M3 model combination, the SFLA-ANN model achieved an RMSE of 1423.09, MAE of 759.88, AIC of 1071.9, R2 of 0.81, MBE of 411.31, and BF of -0.77. Overall, these models can be considered good estimators as their predicted values are generally close to the measured values. The study outputs can help ensure water structures' effective lifespan and operation and take precautions against sediment-related disaster risks. [ABSTRACT FROM AUTHOR]

Background: Legumes, such as lentils, field peas, Faba beans and chickpeas, are high in vitamins, fiber, important minerals and protein and can help avoid obesity and cardiovascular illnesses. They also contribute to ecosystem services, such as nitrogen fixation and resilience to environmental stresses. Despite a 60% increase in global pulse production from 2000 to 2021, a demand-supply gap, especially in South Asia, raises concerns about nutritional access. Since illnesses are currently an issue to the food security of faba beans, machine learning is required for efficient disease identification. Methods: This research employs Convolutional Neural Networks (CNNs) for robust Faba bean leaf disease identification. The CNN model is trained with diverse images representing specific diseases. The study focuses on diseases like Chocolate Spot, Faba Bean Gall, Rust and Healthy leaves. Image processing involves resizing, grayscale conversion and labeling. The CNN architecture includes eight convolutional layers, four max-pooling layers and three dropout layers. The model is trained using 80% of the dataset, validated using 20% and tested for accuracy. Result: The CNN model achieves an accuracy of 99.37% during training and 89.69% during validation after 75 epochs. Confusion matrix and classification report illustrate the model’s performance. It shows high precision, recall and F1 scores for each class, indicating balanced performance. Chocolate Spot and Rust exhibit the highest precision and F1 scores. The overall accuracy is 91%, comparable to other studies on Faba bean disease detection. The study presents a CNN-based disease identification system for Faba beans, demonstrating high accuracy and balanced performance across different diseases. The model’s effectiveness is comparable to other advanced techniques. The research highlights the potential of machine learning in optimizing disease management for Faba beans. Future work could explore a broader range of diseases and incorporate hybrid machine learning algorithms for further improvement. [ABSTRACT FROM AUTHOR]

High accuracy in breast cancer classification contributes to the effectiveness of early breast cancer detection. This study aimed to improve the multiview convolutional neural network (MVCNN) performance for classifying breast cancer based on the combined mediolateral (MLO) and craniocaudal (CC) views. The main contribution of this study is the development of a system, consisting of an effective image pre-processing method to create datasets using background removal techniques, and image enhancement. Also, a simplicity of preprocessing stage in the classifier machine, which does not require a feature extraction process. Furthermore, the performance of the classifier was improved by combining preprocessing dataset techniques and evaluating the best hyperparameter in MVCNN architecture. The digital dataset for screening mammography (DDSM) dataset was used for evaluation in this study. The best result from this proposed method achieved accuracy, precision, sensitivity, and specificity of 98.63%, 97.29%, 100%, and 97.29%. The evaluation results demonstrated the capability to improve classification performance. The method proposed in this work can be applied to the detection of breast cancer. [ABSTRACT FROM AUTHOR]

Tomatoes are recognized for their nutritional value in disease prevention, yet foliar infections often hinder their production. Monitoring and management of tomato leaf disease need much time, cost, and effort. Artificial Intelligence (AI), specifically Deep Neural Networks offers a promising solution for automating disease detection, such as Convolutional Neural Networks (CNN). CNN has several advantages over the traditional machine learning. Selecting the optimal CNN architecture is crucial for accurate feature extraction from image data. Numerous well-known CNN-based architectures have been proposed and used in numerous studies, including VGGNet and DenseNet. DenseNet121 has been proven to produce high classification accuracy. However, the DenseNet121 performance depends on the selection of optimal hyperparameter. This paper proposes an enhancing DenseNet121 performance through hyperparameter optimization using the Bat Algorithm (BA) for classifying tomato leaf disease. BA is used in this research since it has good performance and some advantages. The proposed method achieves better performance than the original DenseNet121 and has a competitive computational time. It produces an accuracy of 0.9465, macroaverage precision of 0.9498, macro-average recall of 0.9465, and macro-average f1-score of 0.9460. This method is prospective to be implemented in devices in the real problem with many various image conditions in the future. [ABSTRACT FROM AUTHOR]

This study investigates the performance of YOLOv8, a Convolutional Neural Network (CNN) architecture, for multi-crop classification in a mixed farm with Unmanned Aerial Vehicle (UAV) imageries. Emphasizing hyperparameter optimization, specifically batch size, the study's primary objective is to refine the model's batch size for improved accuracy and efficiency in crop detection and classification. Using the Google Colaboratory platform, the YOLOv8 model was trained over various batch sizes (10, 20, 30, 40, 50, 60, 70, 80, and 90) to automatically identify the five different classes (sugarcane, banana trees, spinach, pepper, and weeds) present on the UAV images. The performance of the model was assessed using classification accuracy, precision, and recall with the aim of identifying the optimal batch size. The results indicate a substantial improvement in classifier performance from batch sizes of 10 up to 60, while significant dips and peaks were recorded at batch sizes 70 to 90. Based on the analysis of the obtained results, Batch size 60 emerged with the best overall performance for automatic crop detection and classification. Although the F1 score was moderate, the combination of high accuracy, precision, and recall makes it the most balanced option. However, Batch Size 80 also shows very high precision (98%) and balanced recall (84%), which is suitable if the primary focus is on achieving high precision. The findings demonstrate the robustness of YOLOv8 for automatic crop identification and classification in a mixed crop farm while highlighting the significant impact of tuning to the appropriate batch size on the model's overall performance. [ABSTRACT FROM AUTHOR]

Traditional finite element analysis methods have the problem of expensive and unstable band structure diagram calculation when generating phononic crystals. Therefore, this study combines back propagation neural networks with ocean predator algorithms to optimize the geometric structure of phononic crystals. The results show that the coefficient of determination for the predicted bandgap width and lower bound of Model 3 is 1.00, which is better than the comparison model. However, Models 2 and 5 have poor predictive performance for bandgap width due to overfitting during actual training. Therefore, after using the ocean predator algorithm for hyperparameter adjustment, it is found that the maximum number of failures in the validation set of Model 5 is 30, the number of hidden layer nodes is 30, and after 500 experiments, the average error of the bandgap width is 5.26%, and the average error of the lower bound of the bandgap is 1.33%. The average error of the bandgap width after hyperparameter adjustment is 4.89%, and the average error of the lower bound of the bandgap is 1.21%, both of which are effectively reduced and better than the comparison model. Overall, the combination model has high computational efficiency and stability in phononic crystal optimization and can be practically applied in the design and optimization of phononic crystal geometric structures. [ABSTRACT FROM AUTHOR]

Gold is a formof investment known as a safe haven asset because of its stability in unstable market conditions. Gold price forecasting is important for investors as decisions making tool. This study aims to study the best long short–term memory (LSTM) hyperparameters (optimizer, learning rate, and epoch) from cross–validation for forecasting. LSTM, as part of deep learning methods, is developed based on a RNN widely used in time series forecasting. LSTM is superior compared to other methods for its ability to minimize errors and forecast for long–term periods. Walk–forward validation with sliding and extending window scenarios as a form of cross–validation is used to see the method’s accuracy. The used data is sourced from theWorld Gold Council with daily data periods for January 1, 2003, to December 31, 2023. The optimizer used is Adam and RMSProp, each with learning rate values of 0.01, 0.001, 0.0001, and epoch values of 100, 500, 1000. The best model uses the Adam optimizer, a learning rate of 0.01, and an epoch value of 100 with a MAPE value of 0.4867% in the validation process. Forecasting results show a tendency for gold prices to increase in the next eight years. [ABSTRACT FROM AUTHOR]

In the current network security framework, Intrusion Detection Systems (IDSs) happen to be among the major players in ensuring that the network activity is being monitored round the clock for any intrusions which may occur. The rising degree of cyber threats’ intricacy enforces the constant development of IDS methodologies to maintain effectiveness in detecting and reversing the emergence of any extra risks. Therefore, to settle the matter featured by, this research studies try to incorporate the most powerful metaheuristic algorithms, Lion Optimization Algorithm (LOA) and Grey Wolf Optimizer (GWO) in particular, to develop better detection accuracy and efficiency. The core obstacle recognized in this article is the fact that many systems of IDS send out false alarms and their mechanisms of detection of the true anomalies need to be improved immensely. In a nutshell, the change would unveil a fresh way of using LOA and GWO using them to promote the enhancement of internet defences systems in real-time. These schemes can discover previously unknown weaknesses or stealthy attacks. The core of this undertaking would consist in the conception and implementing of a Hybrid Network Intrusion Detection System, which will be created by blending the Lion Optimization Feature Selection (LOFS) and GWO smelters, denoted as LOFSGWO. Critically, the main purpose is to incorporate the GWO as a tool in the operations to cut down the dangerous parameters favourable towards an intrusion mechanism in the framework of a Hybrid CNN-LSTM Deep Learning system. Model tests reveal over 99.26% accuracy of low negative samples into out of a box that are served as testing as well as NSL-KDD dataset, which are similar to the simulation of WUSTL-EOM 2020 system. The obtained outcomes verify the relevance and efficiency of the suggested strategy, which may be used in the resolution of the issues faced in a network security today.

The novelty of this paper revolves around monitoring the condition of nitrogen-filled tires through the fusion of features and utilization of weightless neural networks. A tire pressure monitoring system (TPMS) plays a crucial role in ensuring vehicle safety and comfort. Reckless driving, poor road conditions, continual operation, higher road friction and excessive load are certain factors that can degrade the longevity of tires. Such conditions can result in instantaneous fault attacks in tires raising a concern for safety and comfort. To apply instantaneous and accurate fault diagnosis, the present study leverages machine learning techniques through the integration of an adaptive and robust algorithm, namely, the Wilkes, Stonham and Aleksander Recognition Device (WiSARD) classifier. The experiment uses three types of features namely, statistical, histogram and autoregressive moving average (ARMA) features. The J48 decision tree algorithm was used to pinpoint the key attributes crucial for classification. Following this, the identified attributes were segregated into training and testing datasets, facilitating the evaluation of the WiSARD classifier. Hyperparameter tuning was carried out to achieve optimal value for maximizing classification accuracy and minimizing computational time. Among the features considered, ARMA features delivered the best test set accuracy of about 96.18%.

AbstractThis paper describes deep neural network (DNN) models based on hyperparameter optimization for the prediction of the compressive strength of concrete. The novelty of this research lies in the implementation of optimized hyperparameters to train the DNN models with the aim to enhance their predictive accuracy. Utilizing the Keras Tuner library, the most effective hyperparameters for the DNN models were identified. These models are then trained and evaluated using an experimental dataset of 1030 instances, encompassing nine quantitative variables that include various concrete mix ingredients and age. The target variable for all models is the compressive strength of concrete. The assessment of model performance, based on metrics like mean squared error (MSE) and R2 values, was conducted on previously unseen test data. The optimal configuration for the hidden layers in the model was identified as five layers, containing 12, 16, 16, 40, and 26 neurons, respectively. The optimal learning rate for the model was 0.001. With this set of optimal hyperparameters, the best DNN model achieved an MSE of 28.76 and an R2 value of 0.89 on the testing data. The results demonstrate a significant improvement in the DNN models' performance when trained with the optimized hyperparameters. In comparison to the regression model, the performance of the DNN models was significantly better. Adopting the predictive models developed in this research offers the potential for substantial cost and time savings by circumventing the need for labour-intensive and time-consuming laboratory tests.

Reductions in energy consumption and greenhouse gas emissions are required globally. Under this background, the Multilayer Perceptron machine-learning algorithm was used to predict liquid natural gas consumption to improve energy consumption efficiency. Setting hyperparameters remains challenging in machine-learning-based prediction. Here, to improve prediction efficiency, hyperparameter autotuning via Bayesian optimization was used to identify the optimal combination of the eight key hyperparameters. The autotuned model was validated by comparing its predictive performance with that of a base model (with all hyperparameters set to the default values) using the coefficient of variation of root-mean-square error (CvRMSE) and coefficient of determination (R2) based on the Measurement and Verification Guideline evaluation metrics. To confirm the model's industrial applicability, its predictions were compared with values measured at a small-to-medium-sized food factory. The optimized model performed better than the base model, achieving a CvRMSE of 12.30% and an R2 of 0.94, and achieving a predictive accuracy of 91.49%. By predicting energy consumption, these findings are expected to promote the efficient operation and management of energy in the food industry. [ABSTRACT FROM AUTHOR]

This paper presents a data-driven framework for evaluating collaboration efficiency within scientific research teams. The framework introduces a team efficiency evaluation system consisting of 40 specific indicators, which are analyzed and modeled using statistical methods. The adaptive enhancement algorithm model achieves the highest accuracy, recall, and F1 values, with scores of 0.852, 0.530, and 0.620, respectively. These findings demonstrate the feasibility of the proposed data-driven research team collaboration model, offering theoretical support for enhancing the effectiveness of group collaboration. Moreover, the study is significant for further research on group collaboration in diverse fields. [ABSTRACT FROM AUTHOR]

Solar energy is one of the most promising renewable energy sources that can reduce greenhouse gas emissions and fossil fuel dependence. However, solar energy production is highly variable and uncertain due to the influence of weather conditions and environmental factors. Accurate forecasting of photovoltaic (PV) power output is essential for optimal planning and operation of PV systems, as well as for integrating them into the power grid. This study develops a deep learning model based on Bidirectional Long Short-Term Memory (Bi-LSTM) to predict short-term PV power output. The main objective is to examine the effect of hyperparameter tuning on the forecasting accuracy and the actual PV output power. The main contribution is identifying the optimal combination of hyperparameters, namely the optimizer, the learning rate, and the activation function, for the PV output. The dataset consists of 143786 observations from sensors measuring solar irradiation, PV surface temperature, ambient temperature, ambient humidity, wind speed, and PV power output for 50 days in Bandung, Indonesia. The data is preprocessed by smoothing and splitting into training (70%, 35 days), validation (15%, 7.5 days), and testing (15%, 7.5 days) sets. The Bi-LSTM model is trained and tested with two optimizers: Adam and RMSprop, and three activation functions: Tanh, ReLU, and Swish, with different learning rates. The results indicate that the optimal performance is obtained by the Bi-LSTM model with Adam optimizer, learning rate of 1e-4, and Tanh activation function. This model has the lowest MAE of 0.002931070979684591, the lowest RMSE of 0.008483537231080387, and the highest Rsquared of 0.9988813964105624 when tested with the validation dataset and requires 93 epochs to build. The model also performs well on the test dataset, with the lowest MAE of 0.002717077964916825, the lowest RMSE of 0.007629486798682186, and the highest R-squared of 0.9992563395109665. This study concludes that hyperparameter tuning is a vital step in developing the Bi-LSTM model to improve the accuracy of PV output power prediction. [ABSTRACT FROM AUTHOR]

Topic modeling has emerged as a successful approach to uncovering topics from textual data. Various topic modeling techniques have been introduced, ranging from traditional algorithms to those based on neural networks. In this research, we explore advanced topic modeling techniques, including BERT-based approaches, to enhance the analysis of scientific articles. We first investigate a widely used Latent Dirichlet Allocation (LDA) model and then explore the capabilities of BERT, to automatically uncover latent topics within scientific papers. The goal of this study is to identify the optimal hyperparameter setting for BERT-based topic modeling of scientific articles. We conduct experiments across several scenarios involving combinations of word embedding, dimension reduction, and clustering methods. The results were analyzed based on the coherence values, average execution time, number of topics generated, visualization through the inter-topic distance map, and the top-N-words of each topic. Our findings suggest that combination of RoBERTa for word embedding, PCA for dimension reduction, and K-Means for clustering yields superior results among the tested scenarios. Further evaluation of BERT-based topic modeling is necessary to validate these findings and explore its applications in various academic and industrial contexts. The implications of these advanced techniques could significantly streamline the process of staying updated with scientific literature, potentially revolutionizing research methodologies across disciplines. [ABSTRACT FROM AUTHOR]

Symmetric data play an effective role in the risk assessment process, and, therefore, integrating symmetrical information using Failure Mode and Effects Analysis (FMEA) is essential in implementing projects with big data. This proactive approach helps to quickly identify risks and take measures to address them. However, this task is always time-consuming and costly. On the other hand, there is an essential need for an expert in this field to carry out this process manually. Therefore, in the present study, the authors propose a new methodology to automatically manage this task through a deep-learning technique. Moreover, due to the different nature of the risk data, it is not possible to consider a single neural network architecture for all of them. To overcome this problem, a Genetic Algorithm (GA) was employed to find the best architecture and hyperparameters. Finally, the risks were processed and predicted using the new proposed methodology without sending data to other servers, i.e., external servers. The results of the analysis for the first risk, i.e., latency and real-time processing, showed that using the proposed methodology can improve the detection accuracy of the failure mode by 71.52%, 54.72%, 72.47%, and 75.73% compared to the unique algorithm with the activation function of Relu and number of neurons 32, respectively, related to the one, two, three, and four hidden layers. [ABSTRACT FROM AUTHOR]

The novelty of this paper revolves around monitoring the condition of nitrogen-filled tires through the fusion of features and utilization of weightless neural networks. A tire pressure monitoring system (TPMS) plays a crucial role in ensuring vehicle safety and comfort. Reckless driving, poor road conditions, continual operation, higher road friction and excessive load are certain factors that can degrade the longevity of tires. Such conditions can result in instantaneous fault attacks in tires raising a concern for safety and comfort. To apply instantaneous and accurate fault diagnosis, the present study leverages machine learning techniques through the integration of an adaptive and robust algorithm, namely, the Wilkes, Stonham and Aleksander Recognition Device (WiSARD) classifier. The experiment uses three types of features namely, statistical, histogram and autoregressive moving average (ARMA) features. The J48 decision tree algorithm was used to pinpoint the key attributes crucial for classification. Following this, the identified attributes were segregated into training and testing datasets, facilitating the evaluation of the WiSARD classifier. Hyperparameter tuning was carried out to achieve optimal value for maximizing classification accuracy and minimizing computational time. Among the features considered, ARMA features delivered the best test set accuracy of about 96.18%. [ABSTRACT FROM AUTHOR]

This paper introduces an efficient real-coded genetic algorithm (RCGA) evolved for constrained real-parameter optimization. This novel RCGA incorporates three specially crafted evolutionary operators: Tournament Selection (RS) with elitism, Simulated Binary Crossover (SBX), and Polynomial Mutation (PM). The application of this RCGA is directed toward optimizing the MLPRGA+5 model. This model is designed to configure Multilayer Perceptron neural networks by optimizing both their architecture and associated hyperparameters, including learning rates, activation functions, and regularization hyperparameters. The objective function employed is the widely recognized learning loss function, commonly used for training neural networks. The integration of this objective function is supported by the introduction of new variables representing MLP hyperparameter values. Additionally, a set of constraints is thoughtfully designed to align with the structure of the Multilayer Perceptron (MLP) and its corresponding hyperparameters. The practicality and effectiveness of the MLPRGA+5 approach are demonstrated through extensive experimentation applied to four datasets from the UCI machine learning repository. The results highlight the remarkable performance of MLPRGA+5, characterized by both complexity reduction and accuracy improvement. [ABSTRACT FROM AUTHOR]

Coupled data assimilation (CDA) has been attracting researchers' interests to improve Earth system modeling. The CDA methods are classified into two: weakly coupled data assimilation (wCDA), which considers cross‐compartment interaction only in a forecast phase, and strongly coupled data assimilation (sCDA), which additionally uses other compartment's information in an analysis phase. Although sCDA can theoretically provide better estimates than wCDA since sCDA fully uses inter‐compartment covariances, the potential of sCDA in practice is still in debate. We investigate conditions under which sCDA is effective by applying Local Ensemble Transform Kalman Filter (LETKF) to a coupled Lorenz96 model. Especially, we aim to generalize CDA's behavior on coupling strength basis. By continuously changing a cross‐compartment interaction in the coupled Lorenz96 model, we find that sCDA's superiority over wCDA is particularly evident with large cross‐compartment interactions which attenuates dynamics of each compartment (the increase of variables in one compartment induces the reduction of those in the other compartment) and does not intensify the chaoticity of the whole system. The performance of sCDA is more sensitive to the LETKF's hyperparameters (i.e., localization and inflation) than wCDA. Furthermore, through two imperfect model experiments in which model parameters are uncertain, we quantify sCDA's vulnerability to model inaccuracies. The findings of the relationship between the skill of sCDA and inter‐compartment interactions give implications for future CDA strategies. Plain Language Summary: Data assimilation (DA) integrates computer simulation and observation to improve the estimate of system states. Since Earth is a coupled system in which different compartments such as atmosphere and ocean interact with each other, DA methods for coupled systems, namely coupled data assimilation (CDA), have been attracting researchers' interests in Earth system sciences. The CDA methods are classified into two: weakly coupled DA (wCDA) and strongly coupled DA (sCDA). Although sCDA has a higher potential than wCDA as it explicitly uses the information of cross‐compartment interactions, previous studies have mixed results regarding its accuracy and efficiency. Thus, we aimed to reveal conditions under which sCDA is effective. Especially, we focused on the intensity of compartment interactions. We found that the superiority of sCDA against wCDA depends heavily on the cross‐compartment interaction. When the cross‐compartment interaction is large, while it attenuates dynamics of each compartment and does not increase the chaoticity of the system, sCDA works more efficiently than wCDA. Furthermore, sCDA is more vulnerable to model inaccuracy than wCDA. The findings can give the insights to design a DA method for coupled systems. Key Points: Weakly and strongly coupled data assimilation (sCDA) are compared by applying Local Ensemble Transform Kalman Filter to the coupled‐Lorenz96 modelsCDA performs better when cross‐compartment interaction is large, but it does not increase system's chaoticitysCDA is vulnerable to inaccuracy of modeled processes and modeled cross‐compartment interactions [ABSTRACT FROM AUTHOR]

Pick-and-place operations are an integral part of robotic automation and smart manufacturing. By utilizing deep learning techniques on resource-constraint embedded devices, the pick-and-place operations can be made more accurate, efficient, and sustainable, compared to the high-powered computer solution. In this study, we propose a new technique for object detection on an embedded system using SSD Mobilenet V2 FPN Lite with the optimisation of the hyperparameter and image enhancement. By increasing the Red Green Blue (RGB) saturation level of the images, we gain a 7% increase in mean Average Precision (mAP) when compared to the control group and a 20% increase in mAP when compared to the COCO 2017 validation dataset. Using a Learning Rate of 0.08 with an Edge Tensor Processing Unit (TPU), we obtain high real-time detection scores of 97%. The high detection scores are important to the control algorithm, which uses the bounding box to send a signal to the collaborative robot for pick-and-place operation.

Alzheimer’s disease is a neurodegenerative disorder prevalent in older adults, and early diagnosis is crucial for effective treatment. A deep learning model can automatically classify Alzheimer’s disease from magnetic resonance imaging to aid clinicians in diagnosis. Convolutional Neural Networks (CNNs) are commonly used for disease detection in medical images, but their performance is limited due to inadequate labeled data, high inter-class similarity, and overfitting problems. Key hyperparameters influencing CNN performance include the number of convolution layers and filters assigned to each convolution layer. About other hyperparameters, numerous combinations exist. Since CNN models take a long time to train, it is quite costly to try all combinations to find the optimal model. Existing studies have optimized only a few hyperparameters, such as learning rate, batch size, and optimizer in custom and transfer learning models. In this study, we propose an algorithm based on particle swarm optimization to fine-tune the hyperparameters, including the number of convolution layers, filters, and other hyperparameters, in CNN architectures designed to classify Alzheimer’s disease severity. Using the proposed lightweight model, Alzheimer’s disease was accurately classified with an accuracy of 99.53% and an F1-score of 99.63% on a public dataset. Our model surpasses the performance of previous studies, offering the potential to alleviate the burden on doctors and expedite their decision-making processes. The developed framework can be accessed via the link: “https://ai.gop.edu.tr/alzheimer”.

This study proposes a hybrid network model based on data enhancement to address the problem of low accuracy in photovoltaic (PV) power prediction that arises due to insufficient data samples for new PV plants. First, a time-series generative adversarial network (TimeGAN) is used to learn the distribution law of the original PV data samples and the temporal correlations between their features, and these are then used to generate new samples to enhance the training set. Subsequently, a hybrid network model that fuses bi-directional long-short term memory (BiLSTM) network with attention mechanism (AM) in the framework of deep & cross network (DCN) is constructed to effectively extract deep information from the original features while enhancing the impact of important information on the prediction results. Finally, the hyperparameters in the hybrid network model are optimized using the whale optimization algorithm (WOA), which prevents the network model from falling into a local optimum and gives the best prediction results. The simulation results show that after data enhancement by TimeGAN, the hybrid prediction model proposed in this paper can effectively improve the accuracy of short-term PV power prediction and has wide applicability.

Lithium-ion batteries are widely used as power sources for various devices, so rapid and accurate estimation of the health status of lithium-ion batteries is an important means to reduce battery failures. This article conducts charging and discharging experiments on NCA batteries and LFP battery modules. A probability density function based method for predicting the health status of lithium-ion batteries has been proposed. The kurtosis at the peak of the probability density function (PDF) curve of the battery charging voltage was used as input for the model to achieve accurate prediction of battery SOH. The experimental results show that there is a good correlation between this health indicator and battery SOH, with Pearson correlation coefficients greater than 0.96. Therefore, it can be concluded that it can indirectly reflect the current situation of battery SOH and serve as input for the model to further predict SOH. Long short-term memory networks (LSTM) have become a popular deep learning network method for predicting the health status (SOH) of lithium-ion batteries. The LSTM method without optimizing hyperparameters can easily lead to low accuracy in battery SOH prediction models. A modified LSTM method based on Sparrow Search Algorithm (SSA) is proposed for the prediction of State of Health (SOH) in lithium-ion batteries. When the training set only accounts for 20 % of the total data, the root mean square error (RMSE) of LFP battery prediction results is within 0.85 %, and the maximum absolute error (AE) is less than 2.5 %, while the RMSE of NCA battery SOH prediction results is within 0.7 %, and the maximum AE is less than 2.0 %. SSA-LSTM can accurately predict battery SOH under limited training data and has good robustness.

This paper proposes a long short-term memory (LSTM) network to predict the power degradation of proton exchange membrane fuel cells (PEMFCs), and in order to promote the performance of the LSTM network, the ant colony algorithm (ACO) is introduced to optimize the hyperparameters of the LSTM network. First, the degradation mechanism of PEMFCs is analyzed. Second, the ACO algorithm is used to set the learning rate and dropout probability of the LSTM network combined with partial aging data, which can show the characteristics of the dataset. After that, the aging prediction model is built by using the LSTM and ACO (ACO-LSTM) method. Moreover, the convergence of the method is verified with previous studies. Finally, the fuel cell aging data provided by the Xiangyang Da'an Automotive Testing Center are used for verification. The results show that, compared with the traditional LSTM network, ACO-LSTM can predict the aging process of PEMFCs more accurately, and its prediction accuracy is improved by about 35%, especially when the training data are less. At the same time, the performance of the model trained by ACO-LSTM is also excellent under other operating conditions of the same fuel cell, and it has strong versatility. [ABSTRACT FROM AUTHOR]