Support Vector Machine Network Research Articles

An Efficient Machine Learning-Based Model to Effectively Classify the Type of Noises in QR Code: A Hybrid Approach

Granting smart device consumers with information, simply and quickly, is what drives quick response (QR) codes and mobile marketing to go hand in hand. It boosts marketing campaigns and objectives and allows one to approach, engage, influence, and transform a wider target audience by connecting from offline to online platforms. However, restricted printing technology and flexibility in surfaces introduce noise while printing QR code images. Moreover, noise is often unavoidable during the gathering and transmission of digital images. Therefore, this paper proposed an automatic and accurate noise detector to identify the type of noise present in QR code images. For this, the paper first generates a new dataset comprising 10,000 original QR code images of varying sizes and later introduces several noises, including salt and pepper, pepper, speckle, Poisson, salt, local var, and Gaussian to form a dataset of 80,000 images. We perform extensive experiments by reshaping the generated images to uniform size for exploiting Convolutional Neural Network (CNN), Support Vector Machine (SVM), and Logistic Regression (LG) to classify the original and noisy images. Later, the analysis is further widened by incorporating histogram density analysis to trace and target highly important features by transforming images of varying sizes to obtain 256 features, followed by SVM, LG, and Artificial Neural Network (ANN) to identify the noise type. Moreover, to understand the impact of symmetry of noises in QR code images, we trained the models with combinations of 3-, 5-, and 7-noise types and analyzed the classification performance. From comparative analyses, it is noted that the Gaussian and Localvar noises possess symmetrical characteristics, as all the classifiers did not perform well to segregate these two noises. The results prove that histogram analysis significantly improves classification accuracy with all exploited models, especially when combined with SVM, it achieved maximum accuracy for 4- and 6-class classification problems.

Analysis of Art Classroom Teaching Behavior Based on Intelligent Image Recognition

To solve the problem of intelligent image recognition in classroom behavior, this paper proposes a fast target detection based on FFmpeg CODEC, extracts MHI-HOG joint features according to the located foreground target area, and finally completes the behavior recognition model through a BP neural network support vector machine joint classifier based on the look-up table. The results are as follows: the motion detection method based on H.264 FFmpeg CODEC video has the highest detection accuracy, which can reach 95%. The foreground detection method takes about 10 ms and saves 90% of the time. The behavior classification and recognition effect of MHI-HOG joint features based on the model has been significantly improved, and the comprehensive recognition rate has reached 95%. The built-in BP neural network support vector machine has 97% accuracy in extracting, classifying, and recognizing the characteristics of a single sample. This study attempts to identify and analyze the class behavior and validate the effectiveness of the collaborative classifiers proposed in this paper to build an intellectual classroom.

Machine Learning Techniques Applied to Identify the Two-Phase Flow Pattern in Porous Media Based on Signal Analysis

The development of flow pattern identification technology for gas–liquid two-phase flow in porous media is of great significance to engineering research and production. In this paper, a high accuracy identification method for two-phase flow pattern in porous media is proposed with Machine learning techniques. The gas–liquid two-phase flow patterns and corresponding differential pressure signals in porous beds with particle diameters of 1.5 mm, 3 mm, and 6 mm are obtained through visual experiments. Three time domain characteristic parameters (Mean, Standard deviation, and Range) are calculated by a statistical method, while the EMD energy spectrum of the signal is obtained by empirical mode decomposition. Based on these parameters, machine learning technology, including support vector machine (SVM) and BP neural network, are employed to identify the flow pattern. Four flow pattern identification models are trained based on SVM and BP neural network, with accuracies of 94.77%, 93.4%, 96.08%, and 91.5%. Furthermore, the three models with good performance are integrated by integrated learning technology. An integrated identification model of gas–liquid two-phase flow pattern in porous media with an overall accuracy of 98.04% is finally obtained.

SGLT1/2 as the potential biomarkers of renal damage under Apoe-/- and chronic stress via the BP neural network model and support vector machine.

BackgroundChronic stress (CS) could produce negative emotions. The molecular mechanism of SGLT1 and SGLT2 in kidney injury caused by chronic stress combined with atherosclerosis remains unclear.MethodsIn total, 60 C57BL/6J mice were randomly divided into four groups, namely, control (CON, n = 15), control diet + chronic stress (CON+CS, n = 15), high-fat diet + Apoe−/− (HF + Apoe−/−, n = 15), and high-fat diet + Apoe−/− + chronic stress (HF+Apoe−/− + CS, n = 15) groups. The elevated plus maze and open field tests were performed to examine the effect of chronic stress. The expression of SGLT1 and SGLT2 in the kidney was detected. The support vector machine (SVM) and back propagation (BP) neural network model were constructed to explore the predictive value of the expression of SGLT1/2 on the renal pathological changes. The receiver operating characteristic (ROC) curve analysis was used.ResultsA chronic stress model and atherosclerosis model were constructed successfully. Edema, broken reticular fiber, and increased glycogen in the kidney would be obvious in the HF + Apoe−/− + CS group. Compared with the CON group, the expression of SGLT1/2 in the kidney was upregulated in the HF + Apoe−/− + CS group (P < 0.05). There existed positive correlations among edema, glycogen, reticular fiber, expression of SGLT1/2 in the kidney. There were higher sensitivity and specificity of diagnosis of SGLT1/2 for edema, reticular fiber, and glycogen in the kidney. The result of the SVM and BP neural network model showed better predictive values of SGLT1 and SGLT2 for edema and glycogen in the kidney.ConclusionIn conclusion, SGLT1/2 might be potential biomarkers of renal damage under Apoe−/− and chronic stress, which provided a potential research direction for future related explorations into this mechanism.

On English Courses and Teaching Strategies in Compound Application-Oriented Talents Training Based on the Education Ideas of CDIO in the Construction of Smart Cities

To improve the level of English education in China and promote the cultivation of compound application-oriented talents in colleges and universities, this research integrates the concept of conceive design implement operate (CDIO) into college English education. First, the CDIO educational concept is introduced in detail. Second, the moth-flame optimization (MFO) algorithm and the support vector machine (SVM) model by the MFO algorithm are presented, and they are applied to teaching evaluation, and then the teaching plan is proposed by the CDIO concept. The teaching evaluation model is applied to the English teaching of students in class A with ordinary learning levels in a university in Hengyang City. The implementation method of the teaching plan is explained in detail by taking “Unit 2 Charlie Chaplin” in the textbook “New Horizons College English” as an example. Finally, the accuracy of the MFO-SVM model for teaching evaluation is verified by simulation analysis, and the corresponding teaching strategies are summarized through teaching evaluation and customization of teaching plans. The experimental results reveal that the MFO-SVM model has an accuracy of 96.43% in teaching evaluation, while the extreme learning machine (ELM), SVM, and back propagation neural network (BPNN) models have an accuracy of 92.27%, 90.45%, and 86.36%, respectively in the evaluation results of the teaching quality. The summarized teaching strategies are mainly divided into four parts, namely, autonomous learning, project-based teaching, dynamic generation and information feedback, and production teaching. This research has a certain reference for the current reform of English education in China and the cultivation of compound application-oriented talents.

Intelligent Feature Selection Using GA and Neural Network Optimization for Real-Time Driving Pattern Recognition

Driving cycles have a great influence on vehicles’ fuel economy, control performance and drivability. In this paper, vehicle speed is considered and twelve statistical features for driving pattern recognition are selected to obtain the characteristics of driving cycles. To extract the statistical features online from the speed distribution information, the sampling and updating windows are set and the number of features is reduced. Moreover, since the structure and parameters of neural network are crucial to the classifier, the structure and parameters of neural network, the sampling and updating window size, the feature subset selection are simultaneously optimized by genetic algorithm (GA) to improve the classifying accuracy and simplify neural network structure. The hybrid encoding/decoding and the structure operator are designed to optimize the whole neural network classifier. Four typical driving patterns, i.e., congested urban road, flowing urban road, suburban and highway, are selected based on multiple driving cycles. Simulation results show that the classifiers with GA optimized features have more powerful classification capability than principle component analysis (PCA) and kernel PCA (KPCA) based on k-nearest neighbor, support vector machine neural network classifiers and KPCA Convolutional neural network classifier. The proposed classifier obtains the satisfying classification accuracy with faster real-time classifying speed.

A Novel SVM Network Using HOG Feature for Prohibition Traffic Sign Recognition

To recognize prohibition traffic sign, this paper proposes a novel method that is trained by a small number of samples and uses the feature of histogram of oriented gradient (HOG) and support vector machine (SVM) network. The recognition method is mainly divided into three stages. The first stage is image preprocessing, which includes image interception based on ellipse detection, image resizing, and Gamma correction. In the part of image interception, a new ellipse detection method called RHT_MCN is proposed based on RHT, which uses the maximum coincidence number (MCN) of image edge points and detected ellipse edge to choose the final ellipse for image interception. The second stage is the feature extraction of HOG. The third stage is the prohibition traffic sign recognition (PTSR) based on SVM network. In the design and implementation of the PTSR model, a new single-layer SVM network is proposed. The ascending spiral training method of the recognition model is introduced in detail. Finally, the data from GTSRB is used to test and analyze the prohibition traffic sign recognition method. The method is proven to have good applicability.

Deep-Sea Seabed Sediment Classification Using Finely Processed Multibeam Backscatter Intensity Data in the Southwest Indian Ridge

In 2007, China discovered a hydrothermal anomaly in the Longqi hydrothermal area of the Southwest Indian Ridge. It was the first seabed hydrothermal area discovered in the ultraslow spreading ocean ridge in the world. Understanding the types of seabed sediments in this area is critical for studying the typical topography and geological characteristics of deep-sea seabed hydrothermal areas. The traditional classification of deep-seabed sediments adopts box sampling or gravity column sampling and identifies the types of seabed sediments through laboratory analysis. However, this classification method has some shortcomings, such as the presence of discrete sampling data points and the failure of full-coverage detection. The geological sampling in deep-sea areas is particularly inefficient. Hence, in this study, the EM122 multibeam sonar data collected in the Longqi hydrothermal area, Southwest Indian Ridge, in April 2019 are used to analyze multibeam backscatter intensity. Considering various errors in the complex deep-sea environment, obtaining backscatter intensity data can truly reflect seabed sediment types. Through unsupervised and supervised classification, the seabed sediment classification in the Longqi hydrothermal area was studied. The results showed that the accuracy of supervised classification is higher than that of unsupervised classification. Thus, unsupervised classification is primarily used for roughly classifying sediment types without on-site geological sampling. Combining the genetic algorithm (GA) and the support vector machine (SVM) neural network, deep-sea sediment types, such as deep-sea clay and calcareous ooze, can be identified rapidly and efficiently. Based on comparative analysis results, the classification accuracy of the GA-SVM neural network is higher than that of the SVM neural network, and it can be effectively applied to the high-precision classification and recognition of deep-sea sediments. In this paper, we demonstrate the fine-scale morphology and surface sediment structure characteristics of the deep-sea seafloor by finely processing high-precision deep-sea multibeam backscatter intensity data. This research can provide accurate seabed topography and sediment data for the exploration of deep-sea hydrothermal resources and the assessment of benthic habitats in deep-sea hydrothermal areas.

Data Fusion Approach for Managing Clinical Data in an Industrial Environment using IoT

As health issues continue to become more prevalent as the population grows, building a public health network is critical for enhancing the overall health quality of the community. This study offers an Internet of Things (IoT) based health care system that can be employed in the context of community medical care industrial areas. The main focus of this research is to develop a disease prediction strategy that could be applied to community health services using theoretical modelling. Using principal component analysis (PCA) and cluster analysis, an artificial bee colony (ABC) creates a nonlinear support vector machine (SVM) classifier pair. Feature-level fusion analysis was performed to detect probable abnormalities. The results of the experiments reveal that the SVM model offers significant benefits in disease prediction. In the SVM illness prediction model, the ABC algorithm has the best parameter optimization effect in terms of accuracy, time, and other factors. The suggested method outperformed the traditional SVM and BP neural network methods by 17.24 percent and 72.41 percent, respectively. It can lower the RMSE and improve assessment indicators like the precision recall rate and the F-measure, demonstrating the method’s validity and accuracy. As a result, it is frequently used in community health management, geriatric community monitoring, and clinical medical therapy in an industrial environment.

Research on the lying pattern of grouped pigs using unsupervised clustering and deep learning

Changes in lying pattern and lying center position can reflect information about the production efficiency, health, and welfare of pigs. This study investigates the lying pattern and lying center position of grouped pigs by using unsupervised clustering, deep learning, and image processing technology under commercial farming conditions. Images of the lying pattern of grouped pigs under commercial conditions were collected over 110 days, and then ellipse fitting was used to calculate the center of the target pigs. Unsupervised clustering was used to calculate the cluster center of pigs (lying pattern central location). Variance and standard deviation of the distance from the cluster center to each pig were calculated, Finally, the lying patterns of pigs were classified as “Crowded,” “Close,” “Normal,” “Dispersed,” and “Far.” These patterns were based on the standard deviation. In this research, a Convolutional Neural Network-Support Vector Machine (CNN-SVM) classification model was used to classify five lying patterns of grouped pigs; classification accuracy was up to 97.1%. We studied the change in the lying pattern center of grouped pigs over time, and found that the lying pattern center of grouped pigs showed positional variations over time. We collected experimental site data during the period from July 15, 2021 to October 31, 2021 and analyzed the lying patterns of the grouped pigs. Our results showed that pigs mainly had three lying patterns: “Crowded,” “Close,” and “Normal.” Their proportions were 35%, 33.75%, and 18%, respectively. In addition, the study also found that there were significant changes among the five lying patterns in a day. Standard deviation of grouped pigs showed positive correlation with pigpen temperature, and lying pattern variation of grouped pigs was affected by pigpen temperature. Based on these conclusions, suitable environmental conditions could be created for grouped pigs. In the future, more experiments and research will be extended to other types of livestock, as this technology is helpful for large-scale automatic welfare breeding of livestock.

Classification of macular abnormalities using a lightweight CNN-SVM framework

Macular abnormalities are the main reason for central vision loss, especially in elderly people. Due to global population aging, a heavy burden will be placed on the health care system. Therefore, it is urgent and necessary to develop an automatic and intelligent tool to identify macular abnormalities. Optical coherence tomography is a non-invasive rapid imaging technique to diagnose macular abnormalities. We propose a lightweight convolutional neural network–support vector machine (CNN-SVM) framework consisting of a novel lightweight CNN backbone and an SVM classifier for the accurate detection of macular abnormalities. The CNN-SVM framework achieves excellent performance based on various metrics (precision, recall, F1-score, and accuracy) with an accuracy of 99.8% and demonstrates strong interpretability using heatmap visualization, outperforming several state-of-the-art models (Joint-Attention Network, OpticNet, MobileNet-V3, DenseNet-169, ResNet-50, lesion-aware CNN, Atten-ResNet, least-squares generative adversarial network and others). The proposed CNN-SVM framework is a feasible and reliable tool for the classification of macular abnormalities and shows potential for diagnostic ophthalmology in clinical practice.

A New Hybrid Model of Deep Learning ResNeXt-SVM for Weed Detection

A set of experiments has shown that deep learning as well as traditional learning can be used in the weed detection process and perform well, although sometimes these models cannot fully exploit and utilize the long-term dependency relationship between some key features of images and image labels. To remedy this known problem in the field of image classification, the authors have introduced a classifier known as the linear support vector machine (SVM). Specifically, they have combined a ResNeXt and SVM network to provide the ResNeXt-SVM framework that can deepen the exploitation of the structured features of images and the understanding of their content. The experimental results show that compared to other algorithm models such as ResNet, ResNeXt, and VGG, the proposed solution is more precise and efficient in classifying weeds.

Automatic scoring of postures in grouped pigs using depth image and CNN-SVM

Animal posture is a manifestation of animal behavior, and an animal’s behavior provides information about their health, welfare, and living environment. In recent years, machine vision and machine learning technologies have been widely used to detect individual or group behavior of pigs. The purpose of this study is to use machine vision and deep learning technologies to recognize and score multiple postures (standing, sitting, sternal recumbency, ventral recumbency and lateral recumbency) of pigs under commercial conditions based on depth images. In this study, the Azure Kinect DK depth camera with a top view was used to obtain the depth image of pigs, and the target pig image was obtained by GrabCut image segmentation and watershed segmentation of target object calibration. Then, based on the characteristics of the image, the convex hull, boundary, and the depth distance of the shoulder and the hip were obtained. The ratio of the convex hull perimeter to the boundary and the ratio of the convex hull area to the boundary, as well as the depth distance of the shoulder and the hip, and the depth distance ratio of the shoulder to the hip were obtained as the input of the Convolutional Neural Network-Support Vector Machine (CNN-SVM) classification model, and the model was trained and tested. In various classifier detection experiments, the performance of our pig posture classifier for standing posture and lateral recumbency posture was better, with the area under the receiver operating characteristic (AUC) values being 0.9969 and 0.9967, respectively. However, the performance of sitting posture, sternal recumbency posture and ventral recumbency posture classifier was slightly worse but still had good performance: AUC values were 0.9790, 0.9355 and 0.9795, respectively. The model in this article was used to detect the average postures of pigs in one day (taking the average for eight consecutive days), and it was found that the proportion of lying postures was higher than other postures (lying postures were 72%, standing postures were 20%, and sitting postures were 8%). The proportion of standing postures in the daytime was higher than that in the evening, and lying posture was the opposite. The proportion of the three lying postures also changes over time. This study compared the difference of posture recognition accuracy between the model in this paper (CNN-SVM), SVM and CNN; using the same training data and experimental data, the accuracy of posture recognition of the three models was 94.6368%, 92.2175% and 90.5396%, respectively. Therefore, the recognition accuracy of the model in this paper was improved greatly compared with CNN and SVM.

Joint efficacy of the three biomarkers SNCA, GYPB and HBG1 for atrial fibrillation and stroke: Analysis via the support vector machine neural network.

Atrial fibrillation (AF) is the most common type of persistent arrhythmia. Although its incidence has been increasing, the pathogenesis of AF in stroke remains unclear. In this study, a total of 30 participants were recruited, including 10 controls, 10 patients with AF and 10 patients with AF and stroke (AF + STROKE). Differentially expressed genes (DEGs) were identified, and functional annotation of DEGs, comparative toxicogenomic database analysis associated with cardiovascular diseases, and predictions of miRNAs of hub genes were performed. Using RT‐qPCR, biological process and support vector machine neural networks, numerous DEGs were found to be related to AF. HBG1, SNCA and GYPB were found to be upregulated in the AF group. Higher expression of hub genes in AF and AF + STROKE groups was detected via RT‐PCR. Upon training the biological process neural network of SNCA and GYPB for HBG1, only small differences were detected. Based on the support vector machine, the predicted value of SNCA and GYPB for HBG1 was 0.9893. Expression of the hub genes of HBG1, SNCA and GYPB might therefore be significantly correlated to AF. These genes are involved in the incidence of AF complicated by stroke, and may serve as targets for early diagnosis and treatment.

The Path of Film and Television Animation Creation Using Virtual Reality Technology under the Artificial Intelligence

The purpose is to provide a more reliable human-computer interaction (HCI) guarantee for animation works under virtual reality (VR) technology. Inspired by artificial intelligence (AI) technology and based on the convolutional neural network—support vector machine (CNN-SVM), the differences between animation works under VR technology and traditional animation works are analyzed through a comprehensive analysis of VR technology. The CNN-SVM gesture recognition algorithm using the error correction strategy is designed based on HCI recognition. To have better recognition performance, the advantages of depth image and color image are combined, and the collected information is preprocessed including the relations between the times of image training iterations and the accuracy of different methods in the direction of the test set. After experiments, the maximum accuracy of the preprocessed image can reach 0.86 showing the necessity of image preprocessing. The recognition accuracy of the optimized CNN-SVM is compared with other algorithm models. Experiments show that the accuracy of the optimized CNN-SVM has an upward trend compared with the previous CNN-SVM, and the accuracy reaches 0.97. It proves that the designed algorithm can provide good technical support for VR animation, so that VR animation works can interact well with the audience. It is of great significance for the development of VR animation and the improvement of people’s artistic life quality.

State identification of large diameter wet steam pipeline in nuclear power conventional island

The safety of nuclear power plant is an inevitable condition for maintaining the long-term and stable development of nuclear power. Pipeline vibration is one of the common causes of power plant operation faults. Therefore, the state identification of pipeline vibration is very important. Taking a nuclear power heating pipeline as an example, the original vibration signal is decomposed and reconstructed by Wavelet Packet Transform, and the energy eigenvectors in X, Y and Z directions are established. In order to analyze the state of pipeline more accurately, two recognition algorithms of support vector machine (SVM) and SOM neural network are proposed. By comparing the classification effect and classification accuracy of the SVM before and after optimization, the optimized SVM and SOM neural network, it is concluded that the optimized SVM has better effect on pipeline state recognition, and the effectiveness of this method on pipeline state recognition is verified.

Application of a newly developed naive Bayes algorithm in fire alarm

To address the problems of low recognition accuracy of traditional early fire warning systems in actual scenarios, a newly developed naive Bayes (NB) algorithm, namely, improved naive Bayes (INB), was proposed. An optimization method based on attribute weighting and an orthogonal matrix was used to improve the NB algorithm. Attribute weighting considers the influence of different values of each attribute on classification performance under every decision category; the orthogonal matrix weakens the linear relationship between the attributes reducing their correlations, which is more closely related to the conditional independence assumption. Data from the technology report of the National Institute of Standards and Technology (NIST) regarding fire research were used for the simulation, and eight datasets of different sizes were constructed for INB training and testing after filtering and normalization. A ten-fold cross-validation suggests that INB has been effectively trained and demonstrates the stable ability in fire alarms when the dataset contains 190 sets of samples; namely, the INB can be fully trained by using small datasets. A support vector machine (SVM), a back propagation (BP) neural network, and NB were selected for comparison. The results showed that the recognition accuracy, average precision, average recall, and average &lt;inline-formula&gt;&lt;tex-math id="M1"&gt;\begin{document}$\rm{F}_{1}$\end{document}&lt;/tex-math&gt;&lt;alternatives&gt;&lt;graphic xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="JUST-2021-0258_M1.jpg"/&gt;&lt;graphic xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="JUST-2021-0258_M1.png"/&gt;&lt;/alternatives&gt;&lt;/inline-formula&gt; measure of INB were 96.1%, 97.3%, 97.2%, and 97.3%, respectively, which is the highest among the four different algorithms. Additionally, INB has a better performance compared to NB, SVM, and BP neural networks when the training time is short . In conclusion, INB can be used as a core algorithm for fire alarm systems with excellent and stable fire alarm capabilities.

PNN-SVM Approach of Ti-Based Powder’s Properties Evaluation for Biomedical Implants Production

The advent of additive technologies has provided a significant breakthrough in the production of medical implants. It has reduced costs, increased productivity and accuracy of the implant manufacturing process. However, there are problems associated with assessing defects in the microstructure, mechanical and technological properties of alloys, both during their production by powder metallurgy and in the process of 3D printing. Thus traditional research methods of alloys properties demand considerable human, material, and time resources. At the same time, artificial intelligence tools create opportunities for intelligent evaluation of the conformity for the microstructure, phase composition, and properties of titanium powder’s alloys. It provides new possibilities for the efficient production of biocompatible implants for various functional purposes. However, the accuracy of the methods and models used should be as high as possible. In this paper we designed a hybrid PNN-SVM (Probabilistic Neural Network-Support Vector Machine) high-precision approach for the intelligent evaluation of alloy properties for additive manufacturing of biomedical implants. We have proposed a new approach for extending the dimensionality of input data space by the outputs of the summation layer of the modified PNN topology. Subsequent classification based on the expanded dataset is performed using SVM. We conducted experimental modeling of the proposed approach using a data set on the properties of titanium alloys Ti-6Al-4V and Ti-Al-V-Zr. We have demonstrated a significant increase in the accuracy of the PNN-SVM scheme compared to the single classifiers that form it and other machine learning methods.

The research of emotion recognition based on multi-source physiological signals with data fusion

In the context of the increasing application value of emotion recognition and the continuous development of data fusion technology, it is of great significance to study the emotion recognition model based on multi-source physiological signals with data fusion. In this paper, the onedimensional-convolutional neural network-support vector machine (1D-CNN-SVM) emotion recognition model is constructed to extract the emotional features of multi-source physiological signal data, realize data fusion and complete emotion recognition. Firstly, based on the data level fusion method, dimension splicing for data of each channel is used to compare and analyze different data splicing combinations to explore the best one. Secondly, based on the feature level fusion method, the depth features of each part are fused and extracted by convolutional neural network models. Finally, feature stitching and support vector machine algorithm are used to classify and recognize emotion categories. The experimental results verify the effectiveness of the proposed model in the valence-arousal of the four-class task on DEAP dataset, and the recognition accuracy of the optimal combination can reach 93.10%.

A CNN-SVM hybrid model for the classification of thyroid nodules in medical ultrasound images

The thyroid nodule is one of the endocrine issues which is caused by the formation of irregular cells in the thyroid region. The recent success of machine and deep learning techniques in image recognition task leads to solve challenges in diagnostic imaging. An effective Convolutional Neural Network-Support Vector Machine (CNN-SVM) hybrid model is proposed using hinge loss function to achieve better results and stable convergence. The efficiency of the proposed model has been evaluated on public and collected data sets having 1180 and 2616 thyroid USG images after data augmentation. The proposed model has achieved an accuracy of 94.57%, specificity of 91.89%, sensitivity of 96.70% and f-measure of 95.64% on dataset-1 and an accuracy of 96%, specificity of 93.93%, sensitivity of 97.80% and f-measure of 98.33% on dataset-2. It has shown an improvement of 3% to 5% on dataset-1 and 4% to 6% on dataset-2 in comparison with state-of-the-art models.