Search Results (564)

Integrating advanced computing techniques into food safety management has attracted significant attention recently. Machine learning (ML) algorithms offer innovative solutions for Hazard Analysis Critical Control Point (HACCP) monitoring by providing advanced data analysis capabilities and have proven to be powerful tools for assessing the safety of Animal-Source Foods (ASFs). Studies that link ML with HACCP monitoring in ASFs are limited. The present review provides an overview of ML, feature extraction, and selection algorithms employed for food safety. Several non-destructive techniques are presented, including spectroscopic methods, smartphone-based sensors, paper chromogenic arrays, machine vision, and hyperspectral imaging combined with ML algorithms. Prospects include enhancing predictive models for food safety with the development of hybrid Artificial Intelligence (AI) models and the automation of quality control processes using AI-driven computer vision, which could revolutionize food safety inspections. However, handling conceivable inclinations in AI models is vital to guaranteeing reasonable and exact hazard assessments in an assortment of nourishment generation settings. Moreover, moving forward, the interpretability of ML models will make them more straightforward and dependable. Conclusively, applying ML algorithms allows real-time monitoring and predictive analytics and can significantly reduce the risks associated with ASF consumption. Full article

Rolling Bearings are important supporting components of rotating machines in industrial processes; the faults of rolling bearings will cause the deterioration of the operation conditions of rotating machines. How to effectively extract the fault features and identify the fault modes of rolling bearings quickly and accurately has become a key issue for the safe operation of rotating machines. In this paper, a novel hybrid fault diagnosis method of an optimized random forests classifier for rolling bearings is proposed. Firstly, the original vibration signals are decomposed by recursive variational mode decomposition (RVMD), and the typical time–frequency domain features are extracted from decomposed signals at different scales. The Relief-F ranking method is utilized to assess the quality of time–frequency domain features, and the top-ranked features with high weight gain are selected for identifying the fault modes. Then, an improved bee colony algorithm is proposed based on the simulated binary crossover criterion, which is used to optimize the key parameters of random forests (RF). Finally, several experiments are conducted on the Case Western Reserve University bearing dataset and the dataset collected from our rolling bearing fault testbed. The experimental results show that the proposed method can accurately identify bearing faults and outperform other state-of-the-art methods. Full article

Fault diagnostics in internal combustion engines (ICEs) is vital for optimal operation and avoiding costly breakdowns. This paper reviews methodologies for ICE fault detection, including model-based and data-driven approaches. The former uses physical models of engine components to diagnose defects, while the latter employs statistical analysis of sensor data to identify patterns indicating faults. Various methods for ICE fault identification, such as vibration analysis, thermography, acoustic analysis, and optical approaches, are reviewed. This paper also explores the latest approaches for detecting ICE faults. It highlights the challenges in the diagnostic process and ways to enhance result accuracy and reliability. This paper concludes with a review of the progress in fault identification in ICE components and prospects, highlighted by an experimental investigation using 16 machine learning algorithms with seven feature selection techniques under three load conditions to detect faults in a four-cylinder ICE. Additionally, this study incorporates advanced deep learning techniques, including a deep neural network (DNN), a one-dimensional convolutional neural network (1D-CNN), Transformer and a hybrid Transformer and DNN model which demonstrate superior performance in fault detection compared to traditional machine learning methods. Full article

Background/Objectives: New indole/1,2,4-triazole hybrids were synthesized and tested for antiproliferative activity against the NCI 60 cell line as tubulin polymerization inhibitors. Methods: All final compounds, 6a–j and 7a–j were evaluated at a single concentration of 10 µM against a panel of sixty cancer cell lines. Results: Compounds 7a–j, featuring the NO-releasing oxime moiety, exhibited superior anticancer activity to their precursor ketones 6a–j across all tested cancer cell lines. Compounds 6h, 7h, 7i, and 7j were chosen for five-dose evaluations against a comprehensive array of 60 human tumor cell lines. The data showed that all tested compounds had significant anticancer activity throughout the nine tumor subpanels studied, with selectivity ratios ranging from 0.52 to 2.29 at the GI₅₀ level. Compounds 7h and 7j showed substantial anticancer effectiveness against most cell lines across nine subpanels, with GI₅₀ values ranging from 1.85 to 5.76 µM and 2.45 to 5.23 µM. Compounds 6h, 7h, 7i, and 7j were assessed for their inhibitory effects on tubulin polymerization. Conclusions: The results showed that compound 7i, an oxime-based derivative, was the most effective at blocking tubulin, with an IC₅₀ value of 3.03 ± 0.11 µM. This was compared to the standard drug CA-4, which had an IC₅₀ value of 8.33 ± 0.29 µM. Additionally, cell cycle analysis and apoptosis assays were performed for compound 7i. Molecular computational investigations have been performed to examine the binding mode of the most effective compounds to the target enzyme. Full article

The salient object detection (SOD) of forward-looking sonar images plays a crucial role in underwater detection and rescue tasks. However, the existing SOD algorithms find it difficult to effectively extract salient features and spatial structure information from images with scarce semantic information, uneven intensity distribution, and high noise. Convolutional neural networks (CNNs) have strong local feature extraction capabilities, but they are easily constrained by the receptive field and lack the ability to model long-range dependencies. Transformers, with their powerful self-attention mechanism, are capable of modeling the global features of a target, but they tend to lose a significant amount of local detail. Mamba effectively models long-range dependencies in long sequence inputs through a selection mechanism, offering a novel approach to capturing long-range correlations between pixels. However, since the saliency of image pixels does not exhibit sequential dependencies, this somewhat limits Mamba’s ability to fully capture global contextual information during the forward pass. Inspired by multimodal feature fusion learning, we propose a hybrid CNN–Transformer–Mamba architecture, termed FLSSNet. FLSSNet is built upon a CNN and Transformer backbone network, integrating four core submodules to address various technical challenges: (1) The asymmetric dual encoder–decoder (ADED) is capable of simultaneously extracting features from different modalities and systematically modeling both local contextual information and global spatial structure. (2) The Transformer feature converter (TFC) module optimizes the multimodal feature fusion process through feature transformation and channel compression. (3) The long-range correlation attention (LRCA) module enhances CNN’s ability to model long-range dependencies through the collaborative use of convolutional kernels, selective sequential scanning, and attention mechanisms, while effectively suppressing noise interference. (4) The recursive contour refinement (RCR) model refines edge contour information through a layer-by-layer recursive mechanism, achieving greater precision in boundary details. The experimental results show that FLSSNet exhibits outstanding competitiveness among 25 state-of-the-art SOD methods, achieving MAE and ${\mathit{E}}_{\mathit{\xi }}$ values of 0.04 and 0.973, respectively. Full article

This paper proposes a novel two-stage ensemble framework combining Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) with randomized feature selection to enhance diabetes prediction accuracy and calibration. The method first trains multiple LSTM/BiLSTM base models on dynamically sampled feature subsets to promote diversity, followed by a meta-learner that integrates predictions into a final robust output. A systematic simulation study conducted reveals that feature selection proportion critically impacts generalization: mid-range values (0.5–0.8 for LSTM; 0.6–0.8 for BiLSTM) optimize performance, while values close to 1 induce overfitting. Furthermore, real-life data evaluation on three benchmark datasets—Pima Indian Diabetes, Diabetic Retinopathy Debrecen, and Early Stage Diabetes Risk Prediction—revealed that the framework achieves state-of-the-art results, surpassing conventional (random forest, support vector machine) and recent hybrid frameworks with an accuracy of up to 100%, AUC of 99.1–100%, and superior calibration (Brier score: 0.006–0.023). Notably, the BiLSTM variant consistently outperforms unidirectional LSTM in the proposed framework, particularly in sensitivity (98.4% vs. 97.0% on retinopathy data), highlighting its strength in capturing temporal dependencies. Full article

Autism Spectrum Disorder (ASD) presents challenges in early screening due to its varied nature and sophisticated early signs. From a machine-learning (ML) perspective, the primary challenges include the need for large, diverse datasets, managing the variability in ASD symptoms, providing easy-to-understand models, and ensuring ASD predictive models that can be employed across different populations. Interpretable or explainable classification algorithms, like rule-based or decision tree, play a crucial role in dealing with some of these issues by offering classification models that can be exploited by clinicians. These models offer transparency in decision-making, allowing clinicians to understand reasons behind diagnostic decisions, which is critical for trust and adoption in medical settings. In addition, interpretable classification algorithms facilitate the identification of important behavioural features and patterns associated with ASD, enabling more accurate and explainable diagnoses. However, there is a scarcity of review papers focusing on interpretable classifiers for ASD detection from a behavioural perspective. Thereby this research aimed to conduct a recent review on rule-based classification research works in order to provide added value by consolidating current research, identifying gaps, and guiding future studies. Our research would enhance the understanding of these techniques, based on data used to generate models and obtain performance by trying to highlight early detection and intervention ways for ASD. Integrating advanced AI methods like deep learning with rule-based classifiers can improve model interpretability, exploration, and accuracy in ASD-detection applications. While this hybrid approach has feature selection relevant features that can be detected in an efficient manner, rule-based classifiers can provide clinicians with transparent explanations for model decisions. This hybrid approach is critical in clinical applications like ASD, where model content is as crucial as achieving high classification accuracy. Full article

Metal additive manufacturing (MAM) has advanced significantly, yet accurately predicting clad characteristics from processing parameters remains challenging due to process complexity and data scarcity. This study introduces a novel hybrid machine learning (ML) framework that integrates validated multi-physics computational fluid dynamics simulations with experimental data, enabling prediction of clad characteristics unattainable through conventional methods alone. Our approach uniquely incorporates physics-aware features, such as volumetric energy density and linear mass density, enhancing process understanding and model transferability. We comprehensively benchmark ML models across traditional, ensemble, and neural network categories, analyzing their computational complexity through Big O notation and evaluating both classification and regression performance in predicting clad geometries and process maps. The framework demonstrates superior prediction accuracy with sub-second inference latency, overcoming limitations of purely experimental or simulation-based methods. The trained models generate processing maps with 0.95 AUC (Area Under Curve) accuracy that directly guide MAM parameter selection, bridging the gap between theoretical modeling and practical process control. By integrating physics-based simulations with ML techniques and physics-aware features, our approach achieves an R² of 0.985 for clad geometry prediction and improved generalization over traditional methods, establishing a new standard for MAM process modeling. This research advances both theoretical understanding and practical implementation of MAM processes through a comprehensive, physics-aware machine learning approach. Full article

Background: Accurate reconstruction and quantification in the post-therapy SPECT/CT imaging of ¹⁶⁶Ho microspheres for hepatic malignancies is crucial for treatment evaluation. This present study aimed to explore the impact of the OSEM reconstruction parameters on SPECT/CT image features for dose distribution determination, using Hybrid Recon™ (Hermes Medical Solutions AB) and full Monte Carlo (MC) collimator modeling. Methods: Image quality and activity quantification were assessed through two acquisitions of the Jaszczak phantom using a Siemens Symbia Intevo Bold SPECT/CT system. The datasets were reconstructed using the OSEM method, with variations in the number of iterations for 15 and 8 subsets, both with and without full MC collimator modeling. Contrast recovery coefficient (${\mathrm{Q}}_{\mathrm{H}}$), coefficient of variation (CV), contrast-to-noise ratio (CNR), calibration factor (CF), and activity recovery coefficient (ARC) were calculated and used to evaluate image quality and activity quantification. Results: Reconstructions with 5 iterations and 15 subsets, as well as 10 iterations and 8 subsets, were selected as the most suitable for ¹⁶⁶Ho imaging, as they provided higher ${\mathrm{Q}}_{\mathrm{H}}$ and ARCs. Incorporating full MC collimator modeling in both reconstructions led to significant improvements in image quality and activity recovery. The CFs remained consistent for a fixed value of 15 and 8 subsets, with values of (14.9 ± 0.5) cps/MBq and (14.6 ± 0.5) cps/MBq, respectively. However, when applying full collimator modeling, the CF values decreased to a range between 10.9 and 12.1 cps/MBq. Conclusions: For ¹⁶⁶Ho SPECT/CT imaging, OSEM (with either 5 iterations and 15 subsets or 10 iterations and 8 subsets) combined with full MC collimator modeling yielded superior image quality and quantification results. Full article

Accurately forecasting sales for fast-moving consumer goods (FMCG) remains a significant challenge due to the volatile and multi-faceted nature of sales data. Existing methods often struggle to capture intricate patterns driven by seasonal trends, external factors, and consumer behavior, hindering effective inventory management and strategic decision-making. To overcome these challenges, we propose STL-DCSInformer-ETS, a hybrid model that integrates three complementary components: STL decomposition, an enhanced DCSInformer model, and the ETS model. The model uses monthly sales data from a FMCG company, with key features including sales volume, product prices, promotional activities, and regulatory factors such as holidays, geographical information, consumer behavior, product factors, etc. STL decomposition partitions time-series data into trend, seasonal, and residual components, reducing data complexity and enabling more targeted forecasting. The enhanced DCSInformer employs dilated causal convolution and a multi-scale feature extraction mechanism to capture long-term dependencies and short-term variations effectively. Meanwhile, the ETS model specializes in modeling seasonal patterns, further refining forecasting precision. To further improve predictive performance, the Random Forest-based Recursive Feature Elimination (RF-RFE) method is applied to optimize feature selection. RF-RFE identifies key predictive factors from multiple dimensions, such as time, geography, and economy, which significantly influence forecasting accuracy. Through numerical experiments, the method demonstrates excellent performance by achieving a 35.9% reduction in Mean Squared Error and a 21.4% decrease in Mean Absolute Percentage Error, significantly outperforming traditional methods. Furthermore, the model effectively captures both medium- and long-term sales trends while addressing short-term fluctuations, leading to more accurate forecasting and improved decision-making for fast-moving consumer goods. This research provides new theoretical insights into hybrid forecasting models and practical solutions for optimizing inventory management and strategic planning in the FMCG industry. Full article

Seed quality is a critical factor in crop production. Therefore, seed classification is required to obtain high-quality seeds and to enhance agricultural sustainability and productivity. This study focuses on the varietal classification of chickpeas, an important source of protein and fiber. Chickpea seed varieties can currently be identified by domain experts; their reliability and efficiency depend on the experience and skills of experts and are prone to human error. The design of classification models with high accuracy to assist in selection mechanisms is required for chickpea varieties. In this study, a novel hybrid methodology is proposed for the chickpea classification problem. This methodology combines three well-suited and robust components: feature extraction using three pre-trained models, feature selection with the ReliefF algorithm, and classification employing classical machine learning methods to enhance classification accuracy and efficiency. Various experiments have been conducted using the four hybrid models developed. Their performance has been compared in terms of accuracy, recall, F1-score, precision, and AUC. TL+SVM and TL+LDA outperformed the other models, with test accuracies of 94.4% and 94%, respectively. These results demonstrate the potential of a powerful model that will be beneficial as a component of computer vision systems in smart agriculture applications. Full article

Camellia japonica is a plant species with high cultural and biological relevance. Besides being used as an ornamental plant species, C. japonica has relevant biological properties. Due to hybridization, thousands of cultivars are known, and their accurate identification is mandatory. Infrared spectroscopy is currently recognized as an accurate and rapid technique for species and/or subspecies identifications, including in plants. However, selecting proper analysis tools (spectra pre-processing, feature selection, and chemometric models) highly impacts the accuracy of such identifications. This study tests the impact of two distinct machine learning-based approaches for discriminating C. japonica cultivars using near-infrared (NIR) and Fourier transform infrared (FTIR) spectroscopies. Leaves infrared spectra (NIR—obtained in a previous study; FTIR—obtained herein) of 15 different C. japonica cultivars (38 plants) were modeled and analyzed via different machine learning-based approaches (Approach 1 and Approach 2), each combining a feature selection method plus a classifier application. Regarding Approach 1, NIR spectroscopy emerged as the most effective technique for predicting C. japonica cultivars, achieving 81.3% correct cultivar assignments. However, Approach 2 obtained the best results with FTIR spectroscopy data, achieving a perfect 100.0% accuracy in cultivar assignments. When comparing both approaches, Approach 2 also improved the results for NIR data, increasing the correct cultivar predictions by nearly 13%. The results obtained in this study highlight the importance of chemometric tools in analyzing infrared data. The choice of a specific data analysis approach significantly affects the accuracy of the technique. Moreover, the same approach can have varying impacts on different techniques. Therefore, it is not feasible to establish a universal data analysis approach, even for very similar datasets from comparable analytical techniques. Full article

Forecasting particulate matter with a diameter of 2.5 μm (PM_2.5) is critical due to its significant effects on both human health and the environment. While ground-based pollution measurement stations provide highly accurate PM_2.5 data, their limited number and geographic coverage present significant challenges. Recently, the use of aerosol optical depth (AOD) has emerged as a viable alternative for estimating PM_2.5 levels, offering a broader spatial coverage and higher resolution. Concurrently, long short-term memory (LSTM) models have shown considerable promise in enhancing air quality predictions, often outperforming other prediction techniques. To address these challenges, this study leverages geographic information systems (GIS), remote sensing (RS), and a hybrid LSTM architecture to predict PM_2.5 concentrations. Training LSTM models, however, is an NP-hard problem, with gradient-based methods facing limitations such as getting trapped in local minima, high computational costs, and the need for continuous objective functions. To overcome these issues, we propose integrating the novel orchard algorithm (OA) with LSTM to optimize air pollution forecasting. This paper utilizes meteorological data, topographical features, PM_2.5 pollution levels, and satellite imagery from the city of Tehran. Data preparation processes include noise reduction, spatial interpolation, and addressing missing data. The performance of the proposed OA-LSTM model is compared to five advanced machine learning (ML) algorithms. The proposed OA-LSTM model achieved the lowest root mean square error (RMSE) value of 3.01 µg/m³ and the highest coefficient of determination (R²) value of 0.88, underscoring its effectiveness compared to other models. This paper employs a binary OA method for sensitivity analysis, optimizing feature selection by minimizing prediction error while retaining critical predictors through a penalty-based objective function. The generated maps reveal higher PM_2.5 concentrations in autumn and winter compared to spring and summer, with northern and central areas showing the highest pollution levels. Full article

Asymmetric cove-edged graphene nano-ribbons were employed as metallic electrodes in a hybrid gap device structure with zig-zag graphene nano-ribbons terminations for amino acid recognition and peptide sequencing. On a theoretical basis, amino acid recognition is attained by calculating, using the non equilibrium Green function scheme based on density functional theory, the transversal tunnelling current flowing across the gap device during the peptide translocation through the device. The reliability and robustness of this sequencing method versus relevant operations parameters, such as the bias, the gap size, and small perturbations of the atomistic structures, are studied for the paradigmatic case of Pro-Ser model peptide. I evidence that the main features of the tunnelling signal, that allow the recognition, survive for all of the operational conditions explored. I also evidence a sort of geometrical selective sensitivity of the hybrid cove-edged graphene nano-ribbons versus the bias that should be carefully considered for recognition. Full article

This paper presents an approach for event recognition in sequential images using human body part features and their surrounding context. Key body points were approximated to track and monitor their presence in complex scenarios. Various feature descriptors, including MSER (Maximally Stable Extremal Regions), SURF (Speeded-Up Robust Features), distance transform, and DOF (Degrees of Freedom), were applied to skeleton points, while BRIEF (Binary Robust Independent Elementary Features), HOG (Histogram of Oriented Gradients), FAST (Features from Accelerated Segment Test), and Optical Flow were used on silhouettes or full-body points to capture both geometric and motion-based features. Feature fusion was employed to enhance the discriminative power of the extracted data and the physical parameters calculated by different feature extraction techniques. The system utilized a hybrid CNN (Convolutional Neural Network) + RNN (Recurrent Neural Network) classifier for event recognition, with Grey Wolf Optimization (GWO) for feature selection. Experimental results showed significant accuracy, achieving 98.5% on the UCF-101 dataset and 99.2% on the YouTube dataset. Compared to state-of-the-art methods, our approach achieved better performance in event recognition. Full article