جستجوی مقالات مرتبط با کلیدواژه « support vector machine » در نشریات گروه « برق »

This paper presents a new method for predicting fault location on distribution networks based on the support vector machine (SVM) technique. The proposed in an online non model based scheme which works based on the real data provided by wide area signals, performs as an intelligent indicator for online estimation of fault locations in distribution systems. In this case, for training intelligent SVM based indicator, a feature selection technique is used to find the best combination of the system phasor variables as input signal to the relay. For this purpose, several stable/unstable scenarios with the potential of oscillating dynamic behaviors are created by time domain transient stability simulation. The main merit of the proposed protection scheme is its ability for predicting instead of detection which can reasonably increase relay speed. The proposed approach is applied on IEEE 33-bus test system and the simulation results show promising performance for the SVM based relay.

One method for determining motor commands to control hand prostheses is to use surface electr omy ogr am (sEMG) signal patterns. Due to the random and non-stationary nature of the signal, the idea of using signal information in small time intervals was investigated. In this study, with the aim of more accurate and faster detection of hand movements, two signal decomposition methods, namely discrete wavelet transform (DWT) and empirical mode decomposition (EMD) were evaluated. The sEMG sign als of the Ninapro-DB1 dataset, which were extracted from 27 healthy subjects while performing hand and finger movements, were used to design the system. Simple time domain features with fast calculation capability were extracted for each subband of the decomposed signals. Also, support vector machine (SVM) using different kernel functions was applied as a classifier. The results show that the use of DWT and EMD methods with the ability to access the information of time and frequency sub-intervals of the signals, provides better results in identifying hand movements compared to previous studies. With the EMD method and eight intrinsic mode functions (IMF), the highest recognition accuracy of 83.3% was obtained for six movements. Also, the DWT with the Bior5.5 mother wavelet and five levels of decomposition, achieved 80% recognition accuracy for ten movements and with the Coif2 mother wavelet and six levels of decomposition, the accuracy was 83.33% for eight movements. The results show the better performance of the DWT decomposition method compared to EMD for the design of the hand movement recognition system using sEMG signal patterns.

The diagnosis of neurodevelopmental diseases, such as attention deficit/hyperactivity disorder (ADHD), has gained great attention in clinical studies due to its effect on the quality of human life. This disorder is caused by genetic factors, and anatomical and functional brain abnormalities, which can lead to timing deficits, working memory impairments, and inattention. Since the investigation of symmetry between activities of different brain regions may play an important role in the early diagnosis of this disorder, similarity quantification between brain signals is one of the existing challenges in the field of ADHD detection. The goal of this study is to compute symmetry between certain cortical areas from inter-hemispheric or intra-hemispheric channel pairs. For this purpose, a new algorithm based on dynamic time warping as a bivariate feature extraction step and support vector machine (SVM) classifier has been proposed. The proposed method's ability in distinct brain regions has also been explored.The proposed methods have been evaluated on electroencephalogram (EEG) recordings of 14 ADHD children and 19 age-matched healthy controls performing a time-reproduction task. It has achieved high average accuracy rates of 94.38±0.007 in discriminating between healthy controls and patients with ADHD. The experimental results have also demonstrated the superior performance of the proposed method in comparison with previous ADHD detection methods using EEG signals.

The purpose of this paper's prediction of survival or death within 30 days is based on a cerebral hemorrhage. Timely and correct diagnosis and treatment of cerebral hemorrhage are essential. If the patient's death is predicted during these thirty days, the treating physician should use intensive care and more treatment for the patient. Cerebral hemorrhages require immediate treatment and rapid and accurate diagnosis. In this article, using the volume of cerebral hemorrhage and the patient's age and using the neural network of support vector machine (SVM), it is predicted what percentage of people with cerebral hemorrhage survive and what percentage die. Parameters of cerebral hemorrhage volume and, age of patients, neural network input are considered. The network's output is the survival or death of patients with cerebral hemorrhage over the next thirty days. The data we used included the bleeding volume and age of 66 patients with lobar hemorrhage, 76 patients with deep bleeding, nine patients with Pontine hemorrhage and 11 patients with cerebellar hemorrhage. All bleeding models are considered as input to the support vector machine neural network. The overall accuracy of the designed support vector machine neural network is 93%. Regardless of the type of cerebral hemorrhage, the survival or death of people with cerebral hemorrhage within 30 days is predicted.

Due to the progress of the digital image world and increasing numbers, preparing a system for image retrieval is essential. A content-based image retrieval system should find similar images to the image search by a user. In this paper, a novel content-based image retrieval system is proposed. Considering the importance of texture in an image, we introduce a new feature as the histogram of the texture difference in the equal edge orientation. Then, the expressed features are extracted from training images in the proposed system. Then these features are learned using a support vector machine. The proposed system is examined using the standard WANG database. The results show the efficiency of the proposed system in retrieving images compared to similar methods.

In recent years, the issue of credit rating and identification of good and bad customers have received a lot of attention from banks. Granting facilities to well-accounted customers and avoiding granting facilities to badly accounted customers, which leads to reducing bank arrears, are always the major concerns of bank managers, which are not out of reach with the help of an efficient and good credit rating system. This paper presents a new ensemble model based on support vector machine algorithm for the credit rating of bank customers. First, the data set is divided into several subsets by a bootstrap approach. The support vector machine algorithm is then applied to each subset and several models are formed. At the end, voting is done between the models, and the final model is obtained. In order to show the accuracy of the ensemble model, the credit data of Pasargad Bank’s costumers including 2218 instances of credit applicants, each instance contains 14 explanatory attribute, are evaluated using the proposed method. Based on different criteria, the results obtained on the data of the Pasargad Bank’s costumers confirm the superiority of the ensemble support vector machine method over the usual support vector machine method and the random forest method. The type II error (indicates the proportion of bad applicants who are wrongly predicted to be good applicants) of the proposed ensemble method with linear core is 17% less than the usual support vector machine method and 15% less than the random forest method.

This study is aimed to apply three machine learning techniques including Random Forest (RF), Support Vector Machine (SVM), and Multivariate Adaptively Regression Spline (MARS) in both short-term and long-term electric load forecasting problems and compare their performance. Last hour load, temperature, and solar altitude angle of the present hour with holidays were considered as inputs. Three different criteria including root mean square error, mean absolute error, and coefficient of determination were used to evaluate the performance of prediction methods. These methods are all applied to practical electric load demand data obtained from a sub-transmission substation in Hamedan using R programming language. The temperature data are collected from the nearest meteorological weather station and the hourly solar altitude angle for the whole year is accurately calculated with astronomical equations for the studied location. The results show that the implemented methods provide acceptable forecasts and RF and SVM models exhibit superb results and provide more accurate forecasts in short-term load and long-term load forecasting respectively.

Computer networks play an important and practical role in communication and data exchange, and they also share resources with complete ease. Today, various types of computer networks have emerged, one of which is the Internet of Things. In the Internet of Things, network nodes can be smart objects, and in this sense, this network has many nodes and there is a lot of traffic in this network. Like any computer network, it faces its own challenges and problems, one of which is the issue of network intrusion and disruption. This dissertation focuses on detecting anomaly-based intrusion into the Internet of Things using data mining. In this study, after collecting and preparing data, the improved support vector machine with grasshopper optimization algorithm is used as a proposed method to detect anomaly-based intrusion in the Internet of Things. The bagging and k-nearest neighbor classifiers and Basic SVM are compared based on error types and standard performance criteria. The simulation results show 97.2% accuracy in the proposed method and better performance compared to other methods.

Using the unlabeled data in the semi-supervised learning can significantly improve the accuracy of supervised classification. But in some cases, it may dramatically reduce the accuracy of the classification. The reason of such degradation is incorrect labeling of unlabeled data. In this article, we propose the method for high confidence labeling of unlabeled data. The base classifier in the proposed algorithm is the support vector machine. In this method, the labeling is performed only on the set of the unlabeled data that is closer to the decision boundary from the threshold. This data is called informative data. the adding informative data to the training set has a great effect to achieve the optimal decision boundary if the predicted label is correctly. The Epsilon- neighborhood Algorithm (DBSCAN) is used to discover the labeling structure in the data space. The comparative experiments on the UCI dataset show that the proposed method outperforms than some of the previous work to achieve greater accuracy of the self-training semi-supervised classification.

With the development of technology, the use of social networks has become more popular. One of the most popular social networks is the Twitter platform. Sentiment analysis of users' tweets plays an important role in showing users' feelings about the existing conditions of society. In recent years, due to the fact that the text of users' tweets have become more conversational, sentiment analysis has become problematic and reduced its accuracy; It also makes natural language processing difficult. In this research, a method for sentiment analysis of Persian tweets based on the combination of semantic relations and support vector machine classification has been presented. FastText semantic relation is used to extract features. Considering that a large number of features have been extracted; They should be reduced, which is done by using long-short-term memory (LSTM) neural network. In the last part of the proposed method for the classification of sentiments in tweets, the support vector machine model is used. The evaluation criteria used in this research were precision, accuracy, recall and F criterion, and the evaluation results were 83.9, 84.3, 83.9 and 84, respectively. The results of the experiments show the applicability of the proposed method in analyzing the emotions of Persian tweets into six classes of anger, sad, joy, disgust, surprise and fear.

In recent years, machine learning approaches play an important role in quality identification of  manufactured products including tiles and ceramics. Deviation of tile dimensions is one the main challenge in ceramic and tile industry. Prediction of this deformation will be beneficial if it can be predicted before producing the tile. In this paper, an automatic system has been proposed to predict the deviation of the ceramic tiles. Besides, a machine learning approach is utilized to identify the most effective parameters that leads to tiles’ defect. In this way, three different classification approaches including logistic regression, random forest, and support vector machine have been studied and the best solution is determined for this purpose. Moreover, several feature sets and forward feature selection method have been employed to select more effective variables on our decision making. The experimental results conducted on real-world dataset show that, random forest approach achieves better performance than others, and the results illustrate that improper temperature parameters has more effect on tile deviation.