Recurrent Neural Network Method Research Articles

Fault Diagnosis of Three-Phase Induction Motor (IM) Using a Hybrid ELSE-RNN Technique

This manuscript proposes a hybrid technique for fault detection and classification in a three-phase induction motor (IM). The proposed hybrid technique combines enhanced ladder spherical evaluation (ELSE) and recurrent neural networks (RNN); hence, it is called the ELSE–RNN method. The main goal of the manuscript is to detect and categorize the faults that occur in the IM. The key objective of the proposed method is to enhance the performance of supercapacitor (SC) storage technology and fuzzy-tuned proportional–integral (PI) supervision over conventional control. The main contribution of this paper is developing an effective fault diagnosis method for three-phase IMs. The presence of stator, rotor, winding, and bearing faults is employed using signal processing techniques, such as the Hilbert transform and SIFT. The proposed ELSE–RNN technique is utilized with the end goal of detecting and classifying faults. Here, the proposed ELSE–RNN technique recognizes motors’ healthy or unhealthy conditions in many situations to distinguish faults for protection. The proposed ELSE–RNN technique reduces the complexity of detecting and classifying faults with a validated system and increases the system's accuracy. The ELSE–RNN technique is implemented in MATLAB, and its performance is compared to existing techniques.

Diabetic Retinopathy Blood Vessel Detection Using CNN and RNN Techniques

This research aims to detect diabetic retinopathy using Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN). The main objective is to compare these two methods in detecting the condition. Based on the study’s result after training 10 times on each method, the accuracy results were 92% for the CNN method and 50% for the RNN method. These results show, this study with the dataset used, the CNN method is much more effective in detecting diabetic retinopathy than the RNN method. The CNN method is better due to its ability to extract spatial features from images, which is important in image classification tasks.

Multivariat Predict Sales Data Using the Recurrent Neural Network (RNN) Method

Multivariat Predict Sales Data Using the Recurrent Neural Network (RNN) Method

Automated lung disease detection, classification and prediction using RNN framework

Lung diseases are widespread throughout the globe. This group of illnesses includes chronic obstructive pulmonary disease, pneumonia, asthma, TB, fibrosis, and others. The earliest possible diagnosis of lung illness is crucial. Numerous image processing and artificial intelligence models have been created with this goal in mind. Several types of research have been initiated around the world since the arrival of the novel Covid-19 for its reliable estimation. The earlier respiratory disease pneumonia is linked to Covid-19 because several patients died as a result of severe chest congestion (pneumonic condition). Medical experts find it pulmonary illnesses caused by pneumonia and Covid-19 are difficult to differentiate. Chest CT-Scan imaging is the most accurate approach for predicting lung disease. Recently, a number of academics reported using AI-based methods to classify medical images using training data from CT scans. Deep learning is a very effective technique for understanding difficult cognitive difficulties, and more and more challenges are using and evaluating it. Recurrent neural network method, a deep learning system that can accurately detect COVID from CT-scan pictures, was employed in this study. Detect various lung illnesses like pneumonia and TB by using Multi-class RNN. The experimental findings demonstrate the suggested approach increases the precision of disease prediction and also gives information on the diagnoses of the illnesses under study.

Enhanced Brain Tumor Diagnosis Through Differential and Canonical Quadri –Partitioned Neutrosophic Set Classification Methods:A Comparative Study

An early cancer diagnosis is carried out for adequate management of diseases. Magnetic resonance imaging (MRI) is most commonly preferred method for cancer diagnosis. Due to the uncontrolled and rapid growth of cells, brain tumor is occurred. If not treated at a preliminary phase, it may lead to death. Thus, a noteworthy prerequisite for a successful treatment outcome is an early and precise diagnosis.Many conventional methods are discussed for performing efficient tumor detection. But, conventional classification methods not distinguish MRI as primary and metastases tumors in an accurate manner. Therefore, the performance comparison of deep learning-based classification (i.e., Differential Quadri-Partitioned Neutrosophic Interval-valued Polynomial Attention-based Deep CNN (DQNI-PADCNN) method and Canonical Quadri-Partitioned Neutrosophic Set based Otsuka–Ochiai Deep Recurrent Neural Network (CQNS-ODRNN) method) is introduced to provide exact image classification results. The brain MRI images are considered as an input. MRI image classification is carried out through CNN and RNN to find the brain tumor disease. Before the classification process, input images are de-noised. The noise-removed images are get segmented to identify the region of interested regions. Later, the images are classified into four classes such as glioma, meningioma, no tumor, and pituitary classes to detect the brain tumor. Both classification methods use Quadri-Partitioned Neutrosophic set for categorizing the images. Depending on CNNs and RNNs achievement in handling intricate tasks, an optimal multi-class brain tumor diagnosis is carried out. Experimental evaluation is implemented using MATLAB 2017 for brain tumor detection with the Brain Tumor MRI dataset. To the total number of MRI images, the various performance metrics are calculated in terms of sensitivity, specificity, accuracy, and time for the detection of brain tumors.

A Systematic Review of Echo State Networks From Design to Application

Recurrent Neural Network (RNN) has demonstrated its outstanding ability in sequence tasks and has achieved state-of-the-art in many applications, such as industrial and medical. Echo State Network (ESN) is a simple type of RNN and has emerged in the last decade as an alternative to gradient descent training-based RNN. ESN is practical, conceptually simple, and easy to implement with a strong theoretical ground. It can avoid non-converging and computationally expensive issues in gradient descent RNN methods. Since ESN was put forward in 2002, abundant existing works have promoted the progress of ESN, and the recently introduced deep ESN opened the way to uniting the merits of deep learning and reservoir computing. Besides, the combinations of ESNs with other machine learning models have also overperformed baselines in some applications. However, the apparent simplicity of ESNs can sometimes be deceptive. Successfully applying ESNs needs some experience. Thus, we reviewed over 300 related papers and provided a systematic overview for the first time. In this paper, we categorize the related methods into classical ESN, DeepESN, and combination. Then, we analyze them from the perspective of network designs and specific applications. Finally, we discuss the challenges and opportunities by proposing open problems and future work.

Penerapan Metode Recurrent Neural Network dengan Pendekatan Long Short-Term Memory (LSTM) Untuk Prediksi Harga Saham

Investing in shares is now increasingly popular and growing rapidly in Indonesia. By investing in shares, investors will get quite fast and large profits in a fairly short time. Investors need to analyze previous stock movements as a form of investment strategy to get maximum investment results. Many techniques have been applied to predict stock prices, one of which uses techniques in Deep Learning such as Recurrent Neural Network. In this research, research was conducted on stock price predictions using the Recurrent Neural Network method with the Long-Short Term Memory (LSTM) approach on BBNI stock data. The research only uses close data or data about daily closing stock prices. In designing LSTM there are several things that are configured such as dropout size, density, activation function, and number of neurons used, and in the training process one of the optimizers provided by the Keras framework is used, namely the ADAM (Adaptive Moment Estimation) optimizer. The test scenario is carried out using a number of epochs of 10 and 20 with a batch size of 32. The test results will produce MAE and MAPE values, where the lower the MAE and MAPE values, the better the model's performance in making accurate predictions. The test results using epoch 10 got an MAE value of 0.0286 and a MAPE value of 0.0488, while the test results using epoch 20 got an MAE value of 0.0150 and a MAPE value of 0.0257.

Research on Malware Detection Technology for Mobile Terminals Based on API Call Sequence

With the development of the Internet, the types and quantities of malware have grown rapidly, and how to identify unknown malware is becoming a new challenge. The traditional malware detection method based on fixed features is becoming more and more difficult. In order to improve detection accuracy and efficiency for mobile terminals, this paper proposed a malware detection method for mobile terminals based on application programming interface (API) call sequence, which was characterized by the API call sequence and used a series of feature preprocessing techniques to remove redundant processing of the API call sequence. Finally, the recurrent neural network method (RNN) was used to build the model and perform detection and verification. Furthermore, this paper constructed a malware detection model based on a two-way recurrent neural network and used the two-way long short-term memory network model (LSTM) to train the data set containing 5986 malware samples and 5065 benign software samples to obtain the final detection model and its parameters. Finally, the feature vector of the APK file to be detected was passed into the model and obtained the detection results. The experimental results indicated that the detection accuracy of this method can reach 93.68%.

The Integration of Instrumental Elements and Vocal Music Teaching in Colleges and Universities Based on the Core Literacy View of Music

Abstract In this paper, Fourier transforms in big data technology is used to realize the data preprocessing process of vocal signals from analog signals to digital signals and to explore the identification law of various music styles in vocal teaching in colleges and universities. According to the application of a convolutional neural network in vocal music teaching, the evaluation model of vocal music teaching based on a convolutional neural network is constructed by utilizing a convolutional neural network with accurate sensitivity and high-speed eigenvalue training speed. Starting from the design of relevant variables and the selection of evaluation tools, the experimental study of emotional therapy for the integration of percussion instruments and vocal music teaching is designed, and statistical analysis and simulation analysis are used to analyze the integration of musical instruments and vocal music teaching in the context of music core literacy. The results show that the evaluation error of the recurrent neural network method is 10.07%, while the minimum value is 0.24%. The maximum evaluation error of the convolutional neural network method is 8.22%, and the minimum evaluation error is 0.22%. The evaluation results of the quality of the integrated teaching of percussion instruments and vocal music of the model in this paper are better than the evaluation method of recurrent neural networks. During the maintenance period, when subject three was given weekly emotional therapy for teaching percussion instruments and vocal fusion, it was found that the subject’s emotional behavioral problems also completely subsided. This study realizes that the elimination of negative emotions in the teaching of musical instrument elements and college vocal music has a facilitating effect on the formation of students’ core musical literacy perspective.

Modeling market trading strategies of the intermediary entity for microgrids: A reinforcement learning-based approach

Participation of a large number of microgrids (MGs) in the energy markets faces several challenges. To address these challenges, MGs can be aggregated by an Intermediary entity (IE). In this paper, a new decision-making framework is proposed to solve the energy management problem of multiple MGs and their participation in the energy market through the IE. This framework has three main functions. First, energy price forecasting is conducted using the long short-term memory (LSTM) recurrent neural networks method. Then, the power exchange of the MGs with the main grid is optimized by solving their energy management problem. In this stage, the integrated power exchanges of the MGs with the grid are determined based on the predicted prices. In the final function, the Monte Carlo reinforcement learning technique is employed to optimize real-time pricing decisions and identify potential obstacles that may affect proposed bid prices. This approach addresses energy management challenges and enables profitable energy trading in multiple MGs within energy markets, with confirmation supported by simulation results. The results show a maximum increment of 4.55% in the profit of the IE when purchasing energy and a 3.79% maximum increment when selling energy in the real-time market compared to day-ahead decisions, respectively.

Recurrent Neural Network Methods for Extracting Dynamic Balance Variables during Gait from a Single Inertial Measurement Unit.

Monitoring dynamic balance during gait is critical for fall prevention in the elderly. The current study aimed to develop recurrent neural network models for extracting balance variables from a single inertial measurement unit (IMU) placed on the sacrum during walking. Thirteen healthy young and thirteen healthy older adults wore the IMU during walking and the ground truth of the inclination angles (IA) of the center of pressure to the center of mass vector and their rates of changes (RCIA) were measured simultaneously. The IA, RCIA, and IMU data were used to train four models (uni-LSTM, bi-LSTM, uni-GRU, and bi-GRU), with 10% of the data reserved to evaluate the model errors in terms of the root-mean-squared errors (RMSEs) and percentage relative RMSEs (rRMSEs). Independent t-tests were used for between-group comparisons. The sensitivity, specificity, and Pearson's r for the effect sizes between the model-predicted data and experimental ground truth were also obtained. The bi-GRU with the weighted MSE model was found to have the highest prediction accuracy, computational efficiency, and the best ability in identifying statistical between-group differences when compared with the ground truth, which would be the best choice for the prolonged real-life monitoring of gait balance for fall risk management in the elderly.

Named entity recognition of rice genes and phenotypes based on BiGRU neural networks

Named Entity Recognition (NER) is a fundamental but crucial task in natural language processing (NLP) and big data analysis, with wide application range. NER for rice genes and phenotypes is a technique to identify genes and phenotypes from a large amount of text. NER for rice genes and phenotypes can facilitate the acquisition of information in the field of crops and provide references for our research on higher quality crops. At the same time, named entity recognition still faces many challenges. In this paper, we propose an improved bidirectional gated recurrent unit neural network (BI-GRU) method, which is used to automatically identify the required entities (i.e. gene names, rice phenotypes) from relevant rice literature and patents. The neural network model is combined with the Softmax function to directly output the probabilities of labels, forming the BI-GRU-SF model. With the ability of deep learning methods, the semantic information in the context can be learned without the need for feature engineering. Finally, we conducted experiments, and the results showed that our proposed model provided better performance compared to other models. All datasets and resource codes of BI-GRU-SF are available at https://github.com/qqeeqq/NER for academic use.

Analisis Sentimen Terhadap Pengaruh Minat Belanja Berdasarkan Komentar di Marketplace Menggunakan Metode Recurrent Neural Network (RNN)

Product reviews on the marketplace can provide useful information if they are properly processed. Product review analysis can be performed by merchants to obtain information that can be used to evaluate products and services. It is not enough to look at the number of stars in product review analysis activities; it is also necessary to look at the entire contents of the review comments to determine the intent of the review. This can be done manually in small quantities, but in large quantities, the system is more efficient. In order to understand the intent of the reviews, a system that can effectively analyze many reviews is required. Using the Recurrent Neural Network (RNN) method, this study aims to analyze sentiment on the influence of shopping interest based on comments in the marketplace. The RNN model is trained to recognize positive and negative sentiments using data from the marketplace. The sentiment analysis results will be used to assess the impact on user shopping interest in the marketplace. Sentiment analysis was performed in this study using the RNN method in the GRU/LSTM training model with epochs. The researcher determined the epoch to achieve high accuracy. The data used for model training and testing is separated into training and testing data before it is used. A comparison of 80% of training data and 20% of test data is used to split data. This study uses a training model with 77 epochs and a batch size of 128 to create a system that automatically calculates comment sentiment in the marketplace with a 100% accuracy value and determines positive and negative sentiments.

Prediction of Wind Power with Machine Learning Models

Wind power is a vital power grid component, and wind power forecasting represents a challenging task. In this study, a series of multiobjective predictive models were created utilising a range of cutting-edge machine learning (ML) methodologies, namely, artificial neural networks (ANNs), recurrent neural networks (RNNs), convolutional neural networks, and long short-term memory (LSTM) networks. In this study, two independent data sets were combined and used to predict wind power. The first data set contained internal values such as wind speed (m/s), wind direction (°), theoretical power (kW), and active power (kW). The second data set was external values that contained the meteorological data set, which can affect the wind power forecast. The k-nearest neighbours (kNN) algorithm completed the missing data in the data set. The results showed that the LSTM, RNN, CNN, and ANN algorithms were powerful in forecasting wind power. Furthermore, the performance of these models was evaluated by incorporating statistical indicators of performance deviation to demonstrate the efficacy of the employed methodology effectively. Moreover, the performance of these models was evaluated by incorporating statistical indicators of performance deviation, including the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and mean square error (MSE) metrics to effectively demonstrate the efficacy of the employed methodology. When the metrics are examined, it can be said that ANN, RNN, CNN, and LSTM methods effectively forecast wind power. However, it can be said that the LSTM model is more successful in estimating the wind power with an R2 value of 0.9574, MAE of 0.0209, MSE of 0.0038, and RMSE of 0.0614.

SGCRNN: A ChebNet-GRU fusion model for eeg emotion recognition

The paper proposes a deep learning model based on Chebyshev Network Gated Recurrent Units, which is called Spectral Graph Convolution Recurrent Neural Network, for multichannel electroencephalogram emotion recognition. First, in this paper, an adjacency matrix capturing the local relationships among electroencephalogram channels is established based on the cosine similarity of the spatial locations of electroencephalogram electrodes. The training efficiency is improved by utilizing the computational speed of the cosine distance. This advantage enables our method to have the potential for real-time emotion recognition, allowing for fast and accurate emotion classification in real-time application scenarios. Secondly, the spatial and temporal dependence of the Spectral Graph Convolution Recurrent Neural Network for capturing electroencephalogram sequences is established based on the characteristics of the Chebyshev network and Gated Recurrent Units to extract the spatial and temporal features of electroencephalogram sequences. The proposed model was tested on the publicly accessible dataset DEAP. Its average recognition accuracy is 88%, 89.5%, and 89.7% for valence, arousal, and dominance, respectively. The experiment results demonstrated that the Spectral Graph Convolution Recurrent Neural Network method performed better than current models for electroencephalogram emotion identification. This model has broad applicability and holds potential for use in real-time emotion recognition scenarios.

Comparison of the use of a chemistry transport model with a recurrent neural network method for modeling ambient ozone concentrations

Comparison of the use of a chemistry transport model with a recurrent neural network method for modeling ambient ozone concentrations

RNNtcs: A test case selection method for Recurrent Neural Networks

Recurrent Neural Network (RNN) is a typical feedback neural network, which is particularly effective in processing time-series data tasks such as image description, text generation and classification, etc. However, its safety and quality issues have aroused strong concern in academia and industry. One of the most commonly used methods to ensure RNN-driven software quality is testing and optimizing RNN. Its core idea is to introduce human efforts to label test cases so that many labeled test cases can be used to trigger faults in RNNs. But this brings a huge overhead. Therefore, it is of great importance to study the test case selection method for RNN. Current research on test case selection for RNN has achieved some results from the stateful view of RNN, but there are still some problems, such as a low bug detection rate, poor inclusiveness, and lack of diversity.In this paper, according to the structural characteristics of RNN, we propose a method of RNN test case selection, RNNtcs, which combines clustering and uncertainty, in order to select test cases that can touch RNN bugs as much as possible, thus reducing the cost of labeling. In order to evaluate RNNtcs, we conduct experiments on popular image datasets, text datasets and RNN models. The experimental results show that RNNtcs is superior to the state-of-the-art approach DeepState in bug detection effectiveness, bug detection inclusiveness and bug detection diversity. The test cases selected by RNNtcs can effectively improve the robustness of the RNN model.

Weighted automata extraction and explanation of recurrent neural networks for natural language tasks

Recurrent Neural Networks (RNNs) have achieved tremendous success in processing sequential data, yet understanding and analyzing their behaviours remains a significant challenge. To this end, many efforts have been made to extract finite automata from RNNs, which are more amenable for analysis and explanation. However, existing approaches like exact learning and compositional approaches for model extraction have limitations in either scalability or precision. In this paper, we propose a novel framework of Weighted Finite Automata (WFA) extraction and explanation to tackle the limitations for natural language tasks. First, to address the transition sparsity and context loss problems we identified in WFA extraction for natural language tasks, we propose an empirical method to complement missing rules in the transition diagram, and adjust transition matrices to enhance the context-awareness of the WFA. We also propose two data augmentation tactics to track more dynamic behaviours of RNN, which further allows us to improve the extraction precision. Based on the extracted model, we propose an explanation method for RNNs including a word embedding method – Transition Matrix Embeddings (TME) and TME-based task oriented explanation for the target RNN. Our evaluation demonstrates the advantage of our method in extraction precision than existing approaches, and the effectiveness of TME-based explanation method in applications to pretraining and adversarial example generation.

Research on Propagation Characteristics Based on Channel Measurements and Simulations in a Typical Open Indoor Environment

At present, it is difficult to obtain the indoor propagation loss quickly and accurately by directly using measurements in the millimeter wave band. To solve this problem, in this paper, a ray tracing method suitable for indoor scenes based on geometric optics theory, the uniform theory of diffraction and image theory is presented; the space-alternation generalized expectation-maximization (SAGE) algorithm is used to analyze the measured data and the multipath information of the wireless channel is analyzed; three deep learning models are used to predict the path loss at different receiving distances based on 1600 sets of path loss data. The results show that the comparison between the ray tracing and experimental results shows a good agreement. Moreover, the root-mean-square error (RMSE) and mean absolute error (MAE) of the long short-term memory (LSTM) network are the smallest, and the LSTM has a better fitting effect on the propagation loss sequences predicted at more distant locations when compared with the recurrent neural network (RNN) and gate recurrent unit (GRU) methods, which can better reflect the propagation trend. This provides theoretical support for the layout of base stations and network optimization in typical open indoor environments.

Human Activity Recognition Method Based on Edge Computing-Assisted and GRU Deep Learning Network

Human Activity Recognition (HAR) has been proven to be effective in various healthcare and telemonitoring applications. Current HAR methods, especially deep learning, are extensively employed owing to their exceptional recognition capabilities. However, in pursuit of enhancing feature expression abilities, deep learning often introduces a trade-off by increasing Time complexity. Moreover, the intricate nature of human activity data poses a challenge as it can lead to a notable decrease in recognition accuracy when affected by additional noise. These aspects will significantly impair recognition performance. To advance this field further, we present a HAR method based on an edge-computing-assisted and GRU deep-learning network. We initially proposed a model for edge computing to optimize the energy consumption and processing time of wearable devices. This model transmits HAR data to edge-computable nodes, deploys analytical models on edge servers for remote training, and returns results to wearable devices for processing. Then, we introduced an initial convolution method to preprocess large amounts of training data more effectively. To this end, an attention mechanism was integrated into the network structure to enhance the analysis of confusing data and improve the accuracy of action classification. Our results demonstrated that the proposed approach achieved an average accuracy of 85.4% on the 200 difficult-to-identify HAR data, which outperforms the Recurrent Neural Network (RNN) method’s accuracy of 77.1%. The experimental results showcase the efficacy of the proposed method and offer valuable insights for the future application of HAR.