Search

In this study, we successfully captured and compared the gas−particle flow field in a real-sized industrial-scale micron air classifier and in a prototype. All simulation calculations were performed using high-performance computing (HPC) systems and 3D transient simulations with the TWC-RSM–DPM (Two-Way Coupling–Reynolds Stress Model–Discrete Phase Model) in ANSYS Fluent (version 2022 R2). The following objectives were achieved: (i) a comparison of the simulation results was made between a real-size industrial-scale micron air classifier and a prototype model (scaled-down model) to show the differences between them and highlight the necessity of a simulation study on a real-size industrial-scale model for optimization purposes; (ii) a detailed analysis of the effects of the multiple vortices inside both the main and secondary classification zones provided a deeper understanding of the classification mechanism of the real-sized industrial-scale micron air classifier; and (iii) on the basis of the classifier's key performance indicators (KPIs: d50, K, η) and the constrained condition (i.e., the know-how particle size distribution curve (KHC) of quartz fine powder material of 0–45 µm) applied in manufacturing engineering stone, the relationship between the operating parameters and classification performance was addressed, and the optimal set of operating parameters for the production of quartz fine powder material (0–45 µm) was selected. The simulation results will be validated using experimental results at the Vicostone Plant, Phenikaa Group. [ABSTRACT FROM AUTHOR]

To enhance the classification efficiency of hydrocyclones, this study introduces a novel hydrocyclone design featuring a composite curved-inlet-body structure. Through numerical simulations, the internal flow field characteristics of this structure are thoroughly investigated. The results reveal several key findings: when the diameter of the overflow tube is reduced below a critical threshold, the axial velocity exhibits predominantly downward movement within the outer cyclone, accompanied by substantial recirculation, leading to a loss of effective separation. Moreover, both static pressure and tangential velocity are largely independent of the insertion depth of the overflow tube. In contrast, the diameter of the bottom flow opening plays a crucial role in determining flow dynamics within the hydrocyclone. An excessively large or small bottom opening leads to flow instabilities, causing fluctuations that disrupt the uniformity of the flow field. Additionally, a small height-to-diameter ratio exacerbates flow instability, increasing turbulence intensity and resulting in irregular fluctuations in the LZVV. These findings provide important theoretical insights for the design of more efficient hydrocyclone separation structures. [ABSTRACT FROM AUTHOR]

This is an article in the field of mining engineering. With the increasing application of cyclone in the field of mineral processing, the influence of its operating factors on the separation performance has been paid more and more attention in the production process. At present, various researches on the separation performance of cyclone have been widely reported, but they are generally limited to the structural factors of cyclone, and do not consider the effects of operational factors (inlet pressure P, dispersion concentration Ci) and the interaction between factors on the separation performance. In order to explore the influence of cyclone operating factors on separation performance, response surface analysis was applied to design experiments to optimize the mathematical models of cyclone main diameter Dc and operating factors on cyclone classification efficiency Ef and classifier size Dm, and explore the interaction between them, which can achieve the best separation performance regulation of cyclone. Through the analysis of 17 groups of test results by design expert, the following results are obtained: classification efficiency fitting formulaandgrading granularity fitting formula. The 3D surface response analysis shows that the main diameter, inlet pressure and inlet concentration have a significant effect on the classification efficiency and particle size of Dc> Ci >P, and for the classification efficiency and particle size, the interaction between Dc and Ci and between Dc and P is relatively strong, while the interaction between Ci and P is weak.

In this study, a new flexible screen surface structure with an additional striking beam was proposed for improved efficiency and addressing the clogging issue while processing moist coal material. Vibration tests were used to analyze the time-frequency characteristics of the screen surface and the striking beam under no-load and load conditions. The effects of different process parameters on the elastic screening effect of viscous and wet materials were explored. Further, the industrial application of the equipment was preliminarily carried out. The results showed that the load did not affect the main vibration frequency of the elastic screen surface and the striking beam. When the axis distance was adjusted to 7.5 cm, the displacement from the feeding end to the discharge end gradually decreased, promoting the rapid loosening and stratification of the material group. Semi-industrial tests showed that the optimal screening effects had a screening efficiency of 92.16% and a total misplaced content of 5.95% when exciting force was 18 kN, the exciting frequency was 10.8 Hz, the feeding rate was 9.36 t/h, the distance was 7.5 cm, and the moisture content was 8.21%. The GQPSS1848 vibrating screen applied in the industry achieved a screening efficiency of 86.66% at a processing capacity of 85 t/h with a broad commercial prospect. The study aims to provide theoretical and technical support for the research and development of flexible screening equipment and the efficient classification of −3 mm sticky and wet materials. [ABSTRACT FROM AUTHOR]

Aiming at the problem of fine particles entrainment in the underflow of hydrocyclone during the metal grinding classification process, a water-injection hydrocyclone was proposed to improve the classification efficiency. Through external injection of water on the wall surface of the cone section, the particles settling in the region were loosely graded so that the fine particles settled in the wall surface returned to the inner swirl again, thus reducing their entry into the underflow. Numerical simulation was used to explore the differences in the internal flow field characteristics and separation performance of the hydrocyclone after adding the water-injection structure. Then the industrial tests were conducted on the classification performance of the water-injection hydrocyclone. Numerical results showed that compared with the conventional hydrocyclone, the static pressure, tangential and axial velocity of the fluid inside the water-injection hydrocyclone increased, while the turbulence intensity decreased. The experimental results showed that with the increase of water-injection flow rate, the content of -74 µm particles in the underflow of water-injection hydrocyclone first decreased and then increased, and the comprehensive classification efficiency increased and then decreased accordingly. Compared with the conventional hydrocyclone, the content of -74 µm particles in the underflow of the water-injection hydrocyclone was reduced from 26.34% to 23.95%, and the comprehensive classification efficiency was increased from 69.22% to 73.03%, which mitigated the phenomenon of fine particles entrainment in the underflow. [ABSTRACT FROM AUTHOR]

Fisher discriminant analysis (FDA) is a widely used dimensionality reduction tool in pattern recognition. However, FDA cannot obtain an optimal subspace for classification without sufficient labeled samples. Thus, semi-supervised discriminant analysis has attracted great attention in recent years. In this paper, the proposed method employs the exponential-adjusted geometric distance as the measure of similarity, which modifies the exponential function and the scaling factor. The distance not only satisfies the global and local consistency requirements, but also the similarity matrix obtained is more consistent with the real data distribution, thus improves the dimensionality reduction performance. First, in order to deal with the nonlinear separated data, the kernel function is used to map the original data into the high-dimensional feature space. Then, both labeled and unlabeled data in feature space are used to capture the consistence assumption of geometrical structure based on exponential-adjusted geometric distance, which are incorporated into the objection function of local Fisher discriminant analysis as a regularization term. Eventually, the optimal projection matrix is obtained by maximizing the objective function. Experiments on artificial datasets, UCI benchmark datasets, and high-dimensional recognition problems indicate that the presented technique has a significantly improvement in discriminant performance compared with the-state-of-art dimensionality reduction techniques. [ABSTRACT FROM AUTHOR]

The intricacy and high dimensionality of data pose serious obstacles in the quickly developing area of text categorisation, especially in multi-label scenarios where cases may fall into more than one category. This work presents a new method for feature selection in multi-label text classification that incorporates the Beta Ant Colony Optimisation (BACO) algorithm. Our technology preserves interpretability while improving classification performance by efficiently shrinking the feature space. Taking use of ant colonies' cooperative nature, the suggested approach uses a beta distribution to probabilistically direct the selection of relevant traits. Extensive trials on benchmark datasets show that the accuracy, precision, and recall of the BACO-based feature selection much exceeds those of conventional approaches. Furthermore, we examine how certain characteristics affect the categorisation outcomes, providing information about the significance of variables and the connections between different categories. Our study advances multi-label text classification methods and provides a strong foundation for practitioners and academics seeking to increase the efficacy and efficiency of models in many applications. [ABSTRACT FROM AUTHOR]

This research proposes a novel approach for Tamil Handwritten Character Recognition (THCR) that combines feature selection and ensemble learning techniques. The Tamil script is complex and highly variable, requiring a robust and accurate recognition system. Feature selection is used to reduce dimensionality while preserving discriminative features, improving classification performance and reducing computational complexity. Several feature selection methods are compared, and individual classifiers (support vector machines, neural networks, and decision trees) are evaluated through extensive experiments. Ensemble learning techniques such as bagging, and boosting are employed to leverage the strengths of multiple classifiers and enhance recognition accuracy. The proposed approach is evaluated on the HP Labs Dataset, achieving an impressive 95.56% accuracy using an ensemble learning framework based on support vector machines. The dataset consists of 82,928 samples with 247 distinct classes, contributed by 500 participants from Tamil Nadu. It includes 40,000 characters with 500 user variations. The results surpass or rival existing methods, demonstrating the effectiveness of the approach. The research also offers insights for developing advanced recognition systems for other complex scripts. Future investigations could explore the integration of deep learning techniques and the extension of the proposed approach to other Indic scripts and languages, advancing the field of handwritten character recognition. [ABSTRACT FROM AUTHOR]

BackgroundOptical sensing solutions are being developed and adopted to classify a wide range of biological objects, including crop seeds. Performance assessment of optical classification models remains both a priority and a challenge.MethodsAs training data, we acquired hyperspectral imaging data from 3646 individual tomato seeds (germination yes/no) from two tomato varieties. We performed three experimental data manipulations: (1) Object assignment error: effect of individual object in the training data being assigned to the wrong class. (2) Spectral repeatability: effect of introducing known ranges (0-10%) of stochastic noise to individual reflectance values. (3) Size of training data set: effect of reducing numbers of observations in training data. Effects of each of these experimental data manipulations were characterized and quantified based on classifications with two functions [linear discriminant analysis (LDA) and support vector machine (SVM)].ResultsFor both classification functions, accuracy decreased linearly in response to introduction of object assignment error and to experimental reduction of spectral repeatability. We also demonstrated that experimental reduction of training data by 20% had negligible effect on classification accuracy. LDA and SVM classification algorithms were applied to independent validation seed samples. LDA-based classifications predicted seed germination with RMSE = 10.56 (variety 1) and 26.15 (variety 2), and SVM-based classifications predicted seed germination with RMSE = 10.44 (variety 1) and 12.58 (variety 2).ConclusionWe believe this study represents the first, in which optical seed classification included both a thorough performance evaluation of two separate classification functions based on experimental data manipulations, and application of classification models to validation seed samples not included in training data. Proposed experimental data manipulations are discussed in broader contexts and general relevance, and they are suggested as methods for in-depth performance assessments of optical classification models.

When constructing a new data classification algorithm, relevant quality indices such as classification accuracy (ACC) or the area under the receiver operating characteristic curve (AUC) should be investigated. End-users of these algorithms are interested in high values of the metrics as well as the proposed algorithm’s understandability and transparency. In this paper, a simple evolving vector quantization (SEVQ) algorithm is proposed, which is a novel supervised incremental learning classifier. Algorithms from the family of adaptive resonance theory and learning vector quantization inspired this method. Classifier performance was tested on 36 data sets and compared with 10 traditional and 15 incremental algorithms. SEVQ scored very well, especially among incremental algorithms, and it was found to be the best incremental classifier if the quality criterion is the AUC. The Scott–Knott analysis showed that SEVQ is comparable in performance to traditional algorithms and the leading group of incremental algorithms. The Wilcoxon rank test confirmed the reliability of the obtained results. This article shows that it is possible to obtain outstanding classification quality metrics while keeping the conceptual and computational simplicity of the classification algorithm.

This paper mainly uses ANSYS-FLUENT 19.2 software to simulate the movement of the air-flow in a three-separation combined classifier. The simulation results indicate that the air-flow is uniformly distributed in the V-classifier under the action of dispersion plates and baffles. As the air-flow enters the rotor channel, the tangential velocity of the air-flow increases uniformly and remains stable in the axial direction, providing a stable centrifugal force field for particle classification. From the analysis of the flow field and particle trajectories, the separation interface between the upward path for particles and the downward path for coarse particles is relatively clear. The experimental results with sodium bicarbonate show that the V-classifier has a good pre-classification effect. The rotor cage speeds of 300 rpm and 500 rpm are the best working conditions for the coarse powder and fine powder collection, respectively. This study not only provides a new strategy for the design and development of air classifier, but also provides theoretical guidance for its application in industrial production. [ABSTRACT FROM AUTHOR]

Medical image classification plays a pivotal role in diagnosing various diseases. However, selecting informative features from these images remains a challenging task due to the high dimensionality and complexity of the data. Genetic algorithms (GAs) offer a promising approach for feature selection in medical image classification tasks by mimicking the process of natural selection to evolve optimal solutions. This study proposes a genetic algorithm optimization framework for feature selection in medical image classification. The GA iteratively searches the feature space to find the subset of features that maximizes the classification performance. Fitness evaluation is based on a classifier’s performance using selected features, and genetic operators such as crossover and mutation are applied to produce new generations of feature subsets. The proposed framework contributes to enhancing the efficiency and effectiveness of medical image classification by identifying relevant features. By employing GAs, it overcomes the limitations of traditional feature selection methods and adapts to the complexity of medical image data. Experimental results on benchmark medical image datasets demonstrate the effectiveness of the proposed approach. Significant improvements in classification accuracy and computational efficiency are observed compared to baseline methods. Moreover, the selected features exhibit robustness across different classifiers, highlighting the generalizability of the proposed framework. [ABSTRACT FROM AUTHOR]

This article explores the impact of operating parameters on the classification efficiency of a rotor classifier. Based on the experimental data of calcined petroleum coke classification, a single-factor experimental analysis is conducted to find the relationship between operating parameters and classification performance. The cut size becomes progressively smaller as the rotor speed and feeding speed increase, and progressively larger as the inlet air volume increases. Newton's classification efficiency and classification accuracy decreased with the increase in feeding speed. The range analysis of the orthogonal experiment shows that the rotor speed and inlet air volume have significant effects on the classification performance, but the effect of feed speed is relatively weak. In addition, the optimal combination of operating parameters is obtained by optimizing the operating parameters. Newton's classification efficiency under this combination is estimated, and the estimated value is 82%. The verification experiment reveals that the Newton's classification efficiency is 83.5%, which is close to the estimated value. Meanwhile, the classification accuracy is 0.626. This study provides theoretical guidance for the industrial production of calcined petroleum coke and accumulates basic experimental data for the development of air classifiers. [ABSTRACT FROM AUTHOR]

The confusion matrix is the tool commonly used for the evaluation of the performance of a classification algorithm. While the computation of the confusion matrix for multi-class classification follows a well-developed procedure, the common approach for computing the confusion matrix for multi-label classification suffers from the ambiguity related to one-vs-rest strategy and ignores the possibility that predictions could be partially correct, which also leads to inaccuracies of the derived evaluation metric. Only recently, the two approaches dealing with the computation of multi-label confusion matrix have been proposed which take into account the specifics of multi-label classification. In this work, a new method for calculating evaluation metrics for multi-label classification is proposed. The proposed method is based on the calculation of two confusion matrices combined into the confusion tensor. It builds upon the insights into the shortcomings of the two existing approaches for calculating the multi-label confusion matrix. The main drawback of these techniques is their inability to compute precision and recall precisely. The Multi-Label Confusion Tensor was tested on synthetic and real data and compared with existing methods for calculating the multi-label confusion matrix. The source code and the data used to test the methodology are made publicly available.

Introduction: e-healthcare management services can be significantly enhanced through the implementation of recommender systems, as highlighted in various research papers. These systems, such as Healthcare Recommender Systems (HRS) and Health Care Recommender Systems (HCRS), utilize advanced algorithms and machine learning techniques to provide personalized health recommendations based on user input and medical data. Objective: recommend healthcare services based on patient’s state. Model healthcare information network for efficient service recommendation. Method: recommends healthcare services based on patient’s critical situation and requirements. Offers reconfigurable healthcare workflows to medical staff. Machine learning method classification is applied using decision tree and its result is presented which reflects 70 to 75 % accuracy in predictive models which ensure that health recommender system is a full proof system. Result: hospital recommender systems represent a significant advancement in healthcare, providing personalized and data-driven recommendations to patients. Conclusion: the integration of recommender systems in e-healthcare management services holds great potential in improving personalized patient care, promoting health awareness, and optimizing the quality of healthcare recommendations. In this paper author analyzed and estimated the level of accuracy of recommendation systems in healthcare for personalized medical treatment. It surveys current applications, challenges, and future directions in this field. [ABSTRACT FROM AUTHOR]

The Industrial Internet of Things (IIoT), which integrates sensors into the manufacturing system, provides new paradigms and technologies to industry. The massive acquisition of data, in an industrial context, brings with it a number of challenges to guarantee its quality and reliability, and to ensure that the results of data analysis and modelling are accurate, reliable, and reflect the real phenomena being studied. Common problems encountered with real industrial databases are missing data, outliers, anomalies, unbalanced classes, and non-exhaustive historical data. Unlike papers present in the literature that respond to those problems in a dissociated way, the work performed in this article aims to address all these problems at once. A comprehensive framework for data flow encompassing data acquisition, preprocessing, and machine class classification is proposed. The challenges of missing data, outliers, and anomalies are addressed with critical and novel class outliers distinguished. The study also tackles unbalanced class classification and evaluates the impact of missing data on classification accuracy. Several machine learning models for the operating state classification are implemented. The study also compares the performance of the proposed framework with two existing methods: the Histogram Gradient Boosting Classifier and the Extreme Gradient Boosting classifier. It is shown that using "hard voting" ensemble learning methods to combine several classifiers makes the final classifier more robust to missing data. An application is carried out on data from a real industrial dataset. This research contributes to narrowing the theory–practice gap in leveraging IIoT technologies, offering practical insights into data analytics implementation in real industrial scenarios. [ABSTRACT FROM AUTHOR]

In this study, the influence of air inlet layout on the flow field distribution and particle movement trajectory for the vertical turbo air classifier are analyzed comparatively using the numerical simulation method. The air inlet layout adjustment can increase the axial velocity and turbulent dissipation rate at the feeding inlet and do not generate the axial negative velocity, which improves powder material pneumatic transportation and dispersion capacity; the air inlet layout adjustment can match the airflow rotation direction with the rotation direction of the rotor cage, which can eliminate the vortices in the rotor cage channel effectively. Moreover, the particle movement time is shortened and fast classification is completed, which can decrease the particle agglomeration probability and weaken the 'fish-hook' effect. The optimization scheme of the air inlet layout is Type-BC. In accordance with the numerical simulation results, the calcium carbonate classification experimental results indicate that the classification performance of the classifier is improved using Type-BC. [ABSTRACT FROM AUTHOR]

Copyright of Journal of East China University of Science & Technology is the property of Journal of East China University of Science & Technology Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In the paper “MLCM: Multi-Label Confusion Matrix” a method for computing the confusion matrix for the multi-label classification problem is proposed. Although the authors state that there is no similar work on computing confusion matrix for multi-label classification problems, we point out that the method for computing a multi-label confusion matrix was previously proposed in the paper “Multi-Label Classifier Performance Evaluation with Confusion Matrix” by Krstinić et al. We will show that both methods are based on the same set of assumptions and scenarios for instances of true and predicted labels, while there are differences in computing the contribution of classification errors to the confusion matrix between these two approaches.

Machine Learning (ML) has made its way into a wide variety of advanced applications, where high accuracies can be achieved when these ML models are evaluated in the same context as they were trained and validated on. However, when these high-accuracy models are exposed to out-of-distribution points such as noisy inputs, their performance could potentially degrade significantly. Recommending the most suitable ML model that retains a higher accuracy when exposed to these noisy inputs can overcome this performance degradation. For this, a mapping between the noise distribution at the input and the resulting accuracy needs to be obtained. Though, this relationship is costly to evaluate as this is a computationally intensive task. To minimize this computational cost, we employ metalearning to predict this mapping; that is, the performance of different ML models is predicted given the distribution parameters of the input noise. Although metalearning is an established research field, performance predictions based on noise distribution parameters have not been accomplished before. Hence, this research focuses on predicting the per-class classification performance based on the distribution parameters of the input noise. For this, our approach is twofold. First, in order to gain insights in this noise-to-performance relationship, we analyse the per-class performance of well-established convolutional neural networks through our multi-level Monte Carlo simulation. Second, we employ metalearning to learn this relationship between the input noise distribution and the resulting per-class performance in a sample-efficient way by incorporating Latin Hypercube Sampling. The noise performance analyses present novel insights about the per-class performance degradation when gradually increasing noise is augmented on the input. Additionally, we show that metalearning is capable of accurately predicting the per-class performance based on the noise distribution parameters. We also show the relationship between the number of metasamples and the metaprediction accuracy. Consequently, this research enables future work to make accurate classifier recommendations in noisy environments.

This study aimed to develop a hydrocyclone with a spiral guide feeding body structure to improve classification efficiency and classification accuracy. The spiral guide feeding body structure could pre-arrange the particles before they entered into the separation chamber, thus avoiding the turbulence of the flow field and effectively improving the separation efficiency and separation accuracy. The flow field features and separation performance of a hydrocyclone were studied using the response surface method to obtain the prediction model of separation efficiency and cut-size and the interaction law between influencing factors and the optimal combination of parameters under multiple response indicators. The industrial application showed that the classification efficiency was improved by 4.56% over the conventional feeding body structure. This indicated that the hydrocyclone with a spiral guide feeding body structure had potential industrial application value and certain advantages in improving the separation performance. [ABSTRACT FROM AUTHOR]

Considering the entire BCI system, a big challenge is that information can be extracted from brain signals in a meaningful way. Therefore, most BCI studies are focused on brain signal processing, in which the stages are preprocessing, feature extraction, feature selection, and classification. Since each of the signal processing methods is subject-specific, it is necessary to select a specific subject group, that is, a data set, for an effective signal processing review. In this study, all stages of BCI signal processing studies that used the 2b data set recorded with the EEG method for the BCI Competition IV were compiled and compared comprehensively. To be an effective review, this paper organized into common components and showed how varying the four stages alter classification performance. Classification of performance obtained with the methods in the compiled studies was compared in terms of kappa values. The results demonstrate that combinations of different methods affect and improve the performance. This study presents comprehensive guidance by considering all stages for BCI Competition IV data set 2b. The purpose of the present study was to shed light on research with the aim to enhance BCI performance with signal processing using BCI Competition IV data set 2b. [ABSTRACT FROM AUTHOR]

This paper addresses the fraud detection problem in the context of Big Data used in remote banking systems. The paper aims to propose a new algorithm for automatic detection of fraudulent transactions using machine learning with a performance that allows to apply it in big data systems. The article identifies promising directions for optimizing the operation of methods for fraudulent transactions detection in anti-fraud systems. Architectural approaches to the operation of anti-fraud systems have been studied. Based on this, an architecture for illegal actions prediction in a near real-time mode was proposed. The research task of the article is to find the most suitable machine learning algorithm, with the least training and prediction time, demonstrating high classification performance. To achieve this goal, an analysis of the supervised and ensemble machine learning algorithms was made. The dataset was preprocessed for the experiment with SMOTE resampling and robust scaling techniques. The chosen methods were compared using different metrics: f1 score, AUC and time consumption for training and classification. As a result of a metrics comparison, it was found that multilayer perceptron (MLP) and boosting methods (Adaptive, Gradient, XGBoost) has the highest classification, but MLP outperforms boosting methods in terms of time consumption for classification. Thus, MLP was selected as the most appropriate algorithm for further integration to proposed Big Data architecture. Based on the data obtained during the experiments, the degree of their implementation in fraud detection systems was assessed and architecture for the anti-fraud detection system for big data was proposed.

INTRODUCTION: Skin cancer is an emerging disease all over the world which causes a huge mortality. To detect skin cancer at an early stage, computer aided systems is designed. The most crucial step in it is the feature selection process because of its greater impact on classification performance. Various feature selection algorithms were designed previously to find the relevant features from a set of attributes. Yet, there arise challenges in selecting appropriate features from datasets related to disease prediction. OBJECTIVES: To design a hybrid feature selection algorithm for selecting relevant feature subspace from dermatology datasets. METHODS: The hybrid feature selection algorithm is designed by integrating the Latent Semantic Index (LSI) along with correlation-based Feature Selection (CFS). To achieve an optimal selection of feature subset, beetle swarm optimization is used. RESULTS: Statistical metrics such as accuracy, specificity, recall, F1 score and MCC are calculated. CONCLUSION: The accuracy and sensitivity value obtained is 95% and 92%. [ABSTRACT FROM AUTHOR]

Hepatocellular Carcinoma (HCC) is the most frequent malignant liver tumor and the third cause of cancer-related deaths worldwide. For many years, the golden standard for HCC diagnosis has been the needle biopsy, which is invasive and carries risks. Computerized methods are due to achieve a noninvasive, accurate HCC detection process based on medical images. We developed image analysis and recognition methods to perform automatic and computer-aided diagnosis of HCC. Conventional approaches that combined advanced texture analysis, mainly based on Generalized Co-occurrence Matrices (GCM) with traditional classifiers, as well as deep learning approaches based on Convolutional Neural Networks (CNN) and Stacked Denoising Autoencoders (SAE), were involved in our research. The best accuracy of 91% was achieved for B-mode ultrasound images through CNN by our research group. In this work, we combined the classical approaches with CNN techniques, within B-mode ultrasound images. The combination was performed at the classifier level. The CNN features obtained at the output of various convolution layers were combined with powerful textural features, then supervised classifiers were employed. The experiments were conducted on two datasets, acquired with different ultrasound machines. The best performance, above 98%, overpassed our previous results, as well as representative state-of-the-art results. [ABSTRACT FROM AUTHOR]

Abstract Background Optical sensing solutions are being developed and adopted to classify a wide range of biological objects, including crop seeds. Performance assessment of optical classification models remains both a priority and a challenge. Methods As training data, we acquired hyperspectral imaging data from 3646 individual tomato seeds (germination yes/no) from two tomato varieties. We performed three experimental data manipulations: (1) Object assignment error: effect of individual object in the training data being assigned to the wrong class. (2) Spectral repeatability: effect of introducing known ranges (0–10%) of stochastic noise to individual reflectance values. (3) Size of training data set: effect of reducing numbers of observations in training data. Effects of each of these experimental data manipulations were characterized and quantified based on classifications with two functions [linear discriminant analysis (LDA) and support vector machine (SVM)]. Results For both classification functions, accuracy decreased linearly in response to introduction of object assignment error and to experimental reduction of spectral repeatability. We also demonstrated that experimental reduction of training data by 20% had negligible effect on classification accuracy. LDA and SVM classification algorithms were applied to independent validation seed samples. LDA-based classifications predicted seed germination with RMSE = 10.56 (variety 1) and 26.15 (variety 2), and SVM-based classifications predicted seed germination with RMSE = 10.44 (variety 1) and 12.58 (variety 2). Conclusion We believe this study represents the first, in which optical seed classification included both a thorough performance evaluation of two separate classification functions based on experimental data manipulations, and application of classification models to validation seed samples not included in training data. Proposed experimental data manipulations are discussed in broader contexts and general relevance, and they are suggested as methods for in-depth performance assessments of optical classification models.

• RFM generates multiple feature subsets by random feature mapping for stabilizing. • RFM enhances classification effect by fusing results of multiple feature subsets. • RFM assigns optimal weights to different weak classifiers by AdaBoost algorithm. • RFM processes datasets of different dimensions and distribution adaptively. The feature mapping method can improve data separability, enhance data representation ability, and reduce data processing complexity. However, on the one hand, the existing feature mapping methods have difficulty processing datasets of different dimensions and distributions adaptively, limiting the scope of application; on the other hand, a single feature mapping method has the problem of instability and poor generalization ability, weakening the classification ability of subsequent classifiers. This paper proposes a random feature mapping method based on the AdaBoost algorithm and results fusion to enhance classification performance. The method adopts horizontal expansion, fine-tuning weights through sparse autoencoders, and uses input-mapped features as feature nodes to generate multiple feature subsets for increasing stability. After training weak classifiers on each multiple feature subset, the weights of classifiers are adjusted adaptively by the Adaboost ensemble algorithm. Finally, the method fuses weak classifiers twice to enhance classification performance, which abandons the traditional voting method and uses the weighted probability selection method. Experiments on twenty classic datasets show that the proposed method can effectively mine essential features and enhance classification accuracy compared with original datasets. For instance, the Balance dataset has an average classification accuracy of more than 20% higher than the original dataset on the KNN classifier. The proposed method outperforms alternative feature mapping methods in terms of performance and efficiency on different classifiers in most cases. [ABSTRACT FROM AUTHOR]

Counter-Propagation-Artificial-Neural-Networks (C P-ANNs) have been applied in several domains due to their learning and classification abilities. Regardless of their strength, the CP-ANNs still have some limitations in pattern recognition tasks when they encounter ambiguities during the learning process, which leads to the inaccurate classification of the Kohonen-Self-Organizing-Map (K-SOM). This problem has an impact on the performance of the CP-ANNs. Therefore, this paper proposes a novel strategy to improve the CP-ANNs by the Gram-Schmidt algorithm (GSHM) as a pre-processing step of the original data without changing their architecture. Three datasets examples from various domains, such as correlation, crop, and fertilizer, were employed for experimental validation. To obtain the results, we relied on two simulations. The first simulation uses CP-ANNs, and the datasets are inputted into the network without any prior pre-processing. The second simulation uses MCP-ANNs, and the datasets are pre-processed through the GSHM block. Experiment results show that the proposed MCP-ANNs recognize all patterns with a classification accuracy of 100% versus 62.5% for CP-ANNs in the Correlation Dataset. Furthermore, the proposed MCP-ANNs reduce the execution time and training parameter values in all datasets versus CP-ANNs. Thus, the proposed approach based on the GSHM algorithm significantly improves the performance of the CP-ANNs.

Mammography is the most preferred method for breast cancer screening. In this study, computer-aided diagnosis (CAD) systems were used to improve the image quality of mammography images and to detect suspicious areas. The main contribution of this study is to reveal the optimal combination of various pre-processing algorithms to enable better interpretation and classification of mammography images because pre-processing algorithms significantly affect the accuracy of segmentation and classification methods. In this study, the effect of combinations of different preprocessing methods in differentiating benign and malignant breast lesions was investigated. All image processing algorithms used for lesion detection were used in the mini-MIAS database. In the first step, label information and pectoral muscle resulting from the acquisition of mammography images were removed. In the second step, median filter (MF), contrast limited adaptive histogram equalization (CLAHE), and unsharp masking (USM) algorithms with different combinations of the resolution and visibility of images are increased. In the third step, suspicious regions are extracted from the mammograms using the k-means clustering technique. Then, features were extracted from the obtained ROIs. Finally, feature datasets were classified as normal/abnormal, and benign/malign (two class classification) using Machine Learning algorithms. Test performance measures of the classification methods were examined. In both classifications made in the study, lower classification performance values were obtained when the CLAHE algorithm was used alone as a pre-processing method compared to other pre-processing combinations. When the median filter and unsharp masking algorithms are added to the CLAHE algorithm, the performance of the classification methods has increased. In terms of classification success, Support Vector Machines, Random Forest, and Neural Networks showed the best performance. It was found by comparing the performances of the classification methods that different preprocessing algorithms were effective in detecting the presence of breast lesions and distinguishing benign and malignant. [ABSTRACT FROM AUTHOR]

Quasi-induced exposure theory requires the clear-cut assignment of crash responsibility for individual crash-involved drivers. The assignment method based on the citation by police officers poses a concern that the citation would be issued due to the nonmoving violations rather than the driving actions that directly contribute to the crash. Thus, the objective of the study is to improve the accuracy of citation-based responsibility assignments. Binary logistic regression is employed to identify the factors affecting the citation decision of the police officers. An ensemble machine learning method that combines random forest, neural network, and extreme gradient boosting classifiers is established to allocate the crash responsibility. The findings include that (1) the police citation is closely related to the presence of hazardous driving behavior, but it can also be influenced by several factors such as driver age, drinking status, and the collision impact point of the vehicle; and (2) compared to the conventional models, the ensemble machine learning methods have better performance for crash responsibility assignment in terms of accuracy, Kappa coefficient, and area under the curve. The study serves to provide a reliable crash responsibility assignment approach to improve the accuracy of exposure estimation. [ABSTRACT FROM AUTHOR]

Recent developments in machine learning applications are deeply concerned with the poor interpretability of most of these techniques. To gain some insights in the process of designing data-based models it is common to graphically represent the algorithm’s results, either in their final or intermediate stage. Specially challenging is the task of plotting multiclass classification results as they involve categorical variables (classes) rather than numeric results. Using the well-known MNIST dataset and a simple neural network as an example, this paper reviews the existing techniques to visualize classification results, from those centered on a particular instance or set of instances, to those representing an overall performance metric. As classification results are commonly summarized in the form of a confusion matrix, special attention is paid to its graphical representation. From this analysis, a new visualization tool is derived, which is presented in two forms: confusion star and confusion gear. The confusion star is centered on the classification errors, while the confusion gear focuses on the classification hits. The proposed visualization tools are also evaluated when facing: (i) balanced and imbalanced classifiers issues; (ii) the problem of representing errors with different orders of magnitude. By using shapes instead of colors to represent the value of each matrix cell, the new tools significantly improve the readability of the confusion matrices. Furthermore, we show how the area enclosed by the confusion stars and gears are directly related to standard classification metrics. The new graphic tools can be also usefully employed to visualize the performances of a sequence of classifiers.

Concise and unambiguous assessment of a machine learning algorithm is key to classifier design and performance improvement. In the multi-class classification task, where each instance can only be labeled as one class, the confusion matrix is a powerful tool for performance assessment by quantifying the classification overlap. However, in the multi-label classification task, where each instance can be labeled with more than one class, the confusion matrix is undefined. Performance assessment of the multi-label classifier is currently based on calculating performance averages, such as hamming loss, precision, recall, and F-score. While the current assessment techniques present a reasonable representation of each class and overall performance, their aggregate nature results in ambiguity when identifying false negative (FN) and false positive (FP) results. To address this gap, we define a method of creating the multi-label confusion matrix (MLCM) based on three proposed categories of multi-label problems. After establishing the shortcomings of current methods for identifying FN and FP, we demonstrate the usage of the MLCM with the classification of two publicly available multi-label data sets: i) a 12-lead ECG data set with nine classes, and ii) a movie poster data set with eighteen classes. A comparison of the MLCM results against statistics from the current techniques is presented to show the effectiveness in providing a concise and unambiguous understanding of a multi-label classifier behavior.

The personal thermal comfort model is used to design and control the thermal environment and predict the thermal comfort responses of individuals rather than reflect the average response of the population. Previous individual thermal comfort models were mainly focused on a single material environment. However, the channels for individual thermal comfort were various in real life. Therefore, a new personal thermal comfort evaluation method is constructed by means of a reliable decision-based fuzzy classification model from two views. In this study, a two-view thermal comfort fuzzy classification model was constructed using the interpretable zero-order Takagi–Sugeno–Kang (TSK) fuzzy classifier as the basic training subblock, and it is the first time an optimized machine learning algorithm to study the interpretable thermal comfort model is used. The relevant information (including basic information, sampling conditions, physiological parameters, physical environment, environmental perception, and self-assessment parameters) was obtained from 157 subjects in experimental chambers with two different materials. This proposed method has the following features: (1) The training samples in the input layer contain the feature data under experimental conditions with two different materials. The training models constructed from the training samples under these two conditions complement and restrict each other and improve the accuracy of the whole model training. (2) In the rule layer of the training unit, interpretable fuzzy rules are designed to solve the existing layers with the design of short rules. The output of the intermediate layer of the fuzzy classifier and the fuzzy rules are difficult to explain, which is problematic. (3) Better decision-making knowledge information is obtained in both the rule layer of the single-view training model and in the two-view fusion model. In addition, the feature mapping space is generated according to the degree of contribution of the decision-making information from the two single training views, which not only preserves the feature information of the source training samples to a large extent but also improves the training accuracy of the model and enhances the generalization performance of the training model. Experimental results indicated that TMV-TSK-FC has better classification performance and generalization performance than several related state-of-the-art non-fuzzy classifiers applied in this study. Significantly, compared with the single view fuzzy classifier, the training accuracies and testing accuracies of TMV-TSK-FC are improved by 3–11% and 2–9%, respectively. In addition, the experimental results also showed good semantic interpretability of TMV-TSK-FC. [ABSTRACT FROM AUTHOR]

To improve the classification performance of existing adaptive boosting (AdaBoost) based algorithms effectively, an improved AdaBoost algorithm based on misclassified samples oriented feature selection and weighted non-negative matrix factorization (WNMF) is proposed in this paper. Firstly, in order to consider the effects of sample weights, a misclassified samples oriented feature selection (called MOFS) is proposed to select the most discriminative features which occur in the samples with high weights. Secondly, the explicit features and the part-based features of the training samples are both considered, and the WNMF algorithm is introduced and combined with MOFS to reduce the dimension of the training sample set. Finally, the concept of misclassification degree is introduced and a fine grained sample weight updating method is proposed to distinguish the samples with different misclassification degrees. Numerical experiments show that the proposed MOFS method achieves higher accuracy compared to traditional feature selection methods, and the proposed MOFS and WNMF based AdaBoost method obtains significant improvement on classification accuracy when comparing with typical existing AdaBoost based algorithms using different classifiers. [ABSTRACT FROM AUTHOR]

In the present research, we propose the use of a novel hydraulic classifier equipped with a W-shaped reflector to enhance classification performance. The effects of the structural dimensions of a W-shaped reflector on the flow field of a classifier and its classification performance were investigated using numerical simulations and experiments. The results demonstrate that the reflection of the W-shaped reflector results in the return of the feed material back to the classification cavity. After this, the materials are mixed with a rising water flow in order to avoid the settlement of particles. Thus, the particles can stay longer in the classification cavity, facilitating the generation of a suspension bed and effectively improving the classification efficiency and accuracy. Our data indicates that the overall classification efficiency of the classifier embedded with the W-shaped reflector was 11.19% higher than that of a traditional classifier. Our results provide a reference for classifier optimization. [ABSTRACT FROM AUTHOR]

This study proposed a classification model for timber defect classification based on an artificial neural network (ANN). Besides that, the research also focuses on determining the appropriate parameters for the neural network model in optimizing the defect identification performance, such as the number of hidden layers nodes and the number of epochs in the neural network. The neural network's performance is compared with other standard classifiers such as Naïve Bayes, K-Nearest Neighbours, and J48 Decision Tree in finding their significant differences across the multiple timber species. The classifier's performance is measured based on the F-measure due to the imbalanced dataset of the timber species. The experimental results show that the proposed classification model based on the neural network outperforms the other standard classifiers in detecting many types of defects across multiple timber species with an F-measure of 84.01%. This research demonstrates that ANN can accurately classify the defects across multiple species while defining appropriate parameters (hidden layers and epochs) for the neural network model in optimizing defect identification performance.

A new dynamic air classifier was designed to address the problems of uneven material dispersion and high dust concentration in industrial applications of turbo air classifiers. This paper presents a study on the use of guide cones in the new dynamic air classifier. The ANSYS-Fluent 19.2 software was implemented to simulate the airflow in the dynamic air classifier, and the impact of the guide cone size on the flow field and classification performance of the dynamic air classifier was investigated. The simulation results indicated that with the increase in the guide cone height, the flow field distribution becomes reasonable and the velocity distributions become uniform. When the guide cone height is greater than twice the distance between the guide cone and the bottom of the rotor cage, there is no discernible change in the flow field distribution and classification efficiency. When the guide cone diameter is approximately 0.9 times the diameter of the rotor cage, the airflow pathline is more reasonable, and the flow field and velocity distributions are more uniform. An improper guide cone diameter and height will worsen the classification environment, resulting in a significant decline in classification performance. The material experimental and discrete phase simulation (DPM) showed that DPM can anticipate the changing trends of the cut size and classification accuracy. This study provides theoretical assistance for the structural design and optimization of an air classifier. [ABSTRACT FROM AUTHOR]