Multi-domain Feature Extraction Research Articles

An ensemble deep learning model for classification of students as weak and strong learners via multiparametric analysis

Academic data predictions are significantly important for improving the overall education system's effectiveness by providing early identification of weak students and personalized learning strategies. This paper proposes a deep learning model for the identification of weak and strong students using ensemble learning and multiparametric analysis. It combines several machine learning algorithms, including Naive Bayes, Support Vector Machines, Multi-Layer Perceptron, and Logistic Regression using an ensemble learning approach to enhance the model’s performance. Additionally, a custom 1D Convolutional Neural Network (CNN) is designed for classification. It utilizes multiparametric analysis to identify weak and strong students considering various parameters such as age, academic performance, location, and online learning behavior. The evaluation results indicate the performance of the proposed model has been improved in comparison to MLA FIS, SHMM, and DRL by 16.5%, 5.5%, and 2.4%, in terms of precision, 16.4%, 6.5%, and 3.5 % in terms of accuracy and 10.4%, 2.5% and 6.5% in terms of recall. These improvisations described that the model is efficient for multidomain feature extraction, ensemble classification, and high-variance feature selection, which result in a deeper understanding of student performance.

Intelligent prediction of rail corrugation evolution trend based on self-attention bidirectional TCN and GRU

Analyzing the evolution trend of rail corrugation using signal processing and deep learning is critical for railway safety, as current traditional methods struggle to capture the complex evolution of corrugation. This present study addresses the challenge of accurately capturing this trend, which relies significantly on expert judgment, by proposing an intelligent prediction method based on self-attention (SA), a bidirectional temporal convolutional network (TCN), and a bidirectional gated recurrent unit (GRU). First, multidomain feature extraction and adaptive feature screening were used to obtain the optimal feature set. These features were then combined with principal component analysis (PCA) and the Mahalanobis distance (MD) method to construct a comprehensive health indicator (CHI) that reflects the evolution of rail corrugation. A bidirectional fusion model architecture was employed to capture the temporal correlations between forward and backward information during corrugation evolution, with SA embedded in the model to enhance the focus on key information. The outcome was a rail corrugation trend prediction network that combined a bidirectional TCN, bidirectional GRU, and SA. Subsequently, a multi-strategy improved crested porcupine optimizer (CPO) algorithm was constructed to automatically obtain the optimal network hyperparameters. The proposed method was validated with on-site rail corrugation data, demonstrating superior predictive performance compared to other advanced methods. In summary, the proposed method can accurately predict the evolution trend of rail corrugation, offering a valuable tool for on-site railway maintenance.

Tool Wear Prediction Combining Global Feature Attention and Long Short-Term Memory Network

This study aims to accurately predict tool flank wear in milling and simplify the complexity of feature selection. A hybrid approach is proposed to eclectically integrate the advantages between the long short-term memory (LSTM) network and the global feature attention (GFA) module. First, the feature matrix is calculated using the multi-domain feature extraction method. Subsequently, a parallel network is employed to achieve feature fusion. The stacked LSTM network extracts the temporal dependencies between features and the GFA module is used to adaptively complement key features representing global information of samples. Finally, the output features are concatenated, and tool wear prediction is achieved through a fully connected layer. Experiments demonstrate the optimal performance in predicting tool flank wear. Specifically, using the proposed GFA-LSTM framework reduces the mean absolute error (MAE) by 36.9%, 17.7%, and 25.2% in three experiments compared to the simple LSTM, validating the effectiveness of the proposed method.

A Fault Diagnosis Method for Key Components of the CNC Machine Feed System Based on the DoubleEnsemble–LightGBM Model

To solve the problem of fault diagnosis for the key components of the CNC machine feed system under the condition of variable speed conditions, an intelligent fault diagnosis method based on multi-domain feature extraction and an ensemble learning model is proposed in this study. First, various monitoring signals including vibration signals, noise signals, and current signals are collected. Then, the monitoring signals are preprocessed and the time domain, frequency domain, and time–frequency domain feature indices are extracted to construct a multi-dimensional mixed-domain feature set. Finally, the feature set is entered into the constructed DoubleEnsemble–LightGBM model to realize the fault diagnosis of the key components of the feed system. The experimental results show that the model can achieve good diagnosis results under different working conditions for both the widely used dataset and the feed system test bench dataset, and the average overall accuracy is 91.07% and 98.06%, respectively. Compared with XGBoost and other advanced ensemble learning models, this method demonstrates better accuracy. Therefore, the proposed method provides technical support for the stable operation and intelligence of CNC machines.

A hybrid method combining Lévy process and neural network for predicting remaining useful life of rotating machinery

The accurate prediction of remaining useful life (RUL) for rotating machinery with gears and bearings at its core plays a crucial role in ensuring equipment's safe operation and preventing catastrophic accidents. Therefore, this paper focuses on the RUL issue of rotating machinery, proposing a novel RUL prediction framework. Initially, leveraging multi-domain feature extraction and self-organizing map (SOM) networks, this paper constructs the dynamic entropy weighted minimum quantization error (DEWMQE) as the initial health indicator (HI). Subsequently, employing windowed inertia smoothing and scale correction mechanisms, this paper filters anomalies in HI, ensuring trend smoothing and scale consistency. Considering the degradation process and characteristics of rotating machinery systems, this study utilizes linear multi-fractional Lévy stable motion (LMLSM) to build a degradation model. Building upon this, to effectively utilize historical data and overcome the shortcomings of random process theory requiring predefined degradation paths, this study further integrates bidirectional gated recurrent units (BGRU) and similarity transfer learning methods to devise the BGTLLM hybrid model, enabling adaptive fitting of the degradation trend. Finally, the Monte Carlo (MC) method is employed to evaluate the uncertainty in RUL prediction. The effectiveness and accuracy of the proposed hybrid prediction model in RUL forecasting are validated using heavy-duty truck axle data and the PHM2012 dataset.

Construction of multi-features comprehensive indicator for machinery health state assessment

Health state assessment is critical for mechanical equipment’s smooth and healthy operation. This paper proposes a novel approach for health state assessment based on acoustic signals during the process of machinery running. It consists of multi-domain feature (MF) extraction and comprehensive health indicator (CHI) construction. MF is extracted from various acoustic features, including time and frequency (TF) features, mel-frequency cepstral coefficients, and gammatone frequency cepstral coefficients. The stacked long short-term memory (LSTM) is used to extract the high-level features of the MF, which are then input to the downstream PCA to obtain the LSTM-PCA health indicator (LP-HI). Parallelly, the MF is fed into the self-organizing mapping (SOM) model to calculate the minimum quantization error (MQE) as SOM-MQE health indicator (SM-HI). These two indicators are fused using weighted fusion and nonlinear mapping to calculate CHI. The experimental results on air compressor dataset show a 25.8% reduction in evaluation error compared with SOTA results in this paper. The proposed nonlinear mapping function furthermore reduces fitting error on HI by 38.9%. These demonstrate the effectiveness and superiority of the proposed method in machinery health state assessment.

A bearing remaining life prediction method under variable operating conditions based on cross-transformer fusioning segmented data cleaning

Bearing remaining useful life (RUL) prediction research based on deep learning mostly emphasizes model performance and effective feature vectors, overlooking different densities of outlier distributions in vibration signals at varying degradation stages. Moreover, forecasting models focus on capturing cross-time dependencies, ignoring the dependencies between different variables. To solve these problems, this paper proposes an unsupervised segmented data cleaning algorithm and a RUL prediction framework adaptable to variable operating conditions. The method consists of four steps: (1) Multi-domain feature extraction and selection establish a feature vector space reflecting degradation trends. (2) Segmented data cleaning divides degradation stages, using different penalty factors for outlier cleaning. (3) Cleaned vibration signals undergo a second round of multidomain feature engineering and degradation-stage division. (4) A two-stage Cross-Transformer model is used for RUL prediction. The method proposed has been validated on the prognostics and health management (PHM) bearing degradation dataset. In the constant condition prediction task, the root mean square error (RMSE) and mean absolute error (MAE) were improved to 1.88 and 5.78, respectively. In the variable condition prediction task, the proposed method outperformed existing methods, with an improvement of 59.10 % in RMSE, demonstrating strong generalization performance and practical application value.

Intelligent real-time perception method for rock strength based on vibration and power fusion characteristics

To improve the sensing accuracy of rock strength parameters, a method based on the fusion of vibration response and electrical response is proposed. Firstly, multi-layer noise reduction based on local mean decomposition is carried out for vibration response, and the product function containing exceeds 85% of the original information is selected by the multi-index comprehensive decision method, and the wavelet threshold denoising is used as the pre-processing unit of singular value decomposition. Next, the multi-domain feature extraction and KPCA are carried out for vibration signal after noise reduction to construct fusion feature vectors. Also, the statistical characteristics of the electrical parameters are extracted and fused with the fusion features. Finally, multi-source information is input into the least squares support vector machine model optimized by the improved whale algorithm to perceive the rock strength parameters. The experimental results show that the sensing accuracy of the proposed method is the highest (96.875%) compared with the single information while drilling and the traditional prediction method, which verifies the superiority of the proposed method.

A cross-domain state monitoring method for high-speed train brake pads based on data generation under small sample conditions

The proper functioning of the high-speed train brake system is of utmost importance to ensure the train comes to a halt at the intended distance. Among the key components, brake pads are susceptible to abnormal wear over prolonged usage, posing a safety risk to train operations. However, limited real-world data on brake pad damage poses challenges for effective monitoring. This paper proposes an effective transfer learning model that utilizes a small amount of data at different speeds, which can solve the small sample data problem on brake pad damage. Initially, a deep convolutional generative adversarial network (DCGAN) is employed to simulate and expand a limited dataset of vibration signals. The generated signals accurately preserve both the high-frequency and low-frequency characteristics observed in the actual signals. Subsequently, a domain-adversarial neural network (DANN) is utilized to map features from the source and target domains onto a shared feature space, enabling multi-domain feature extraction and adaptation. Finally, an intelligent monitoring network is employed to identify the health status of unlabeled brake pads under fluctuating speeds. Experimental results demonstrate the proposed approach outperforms other commonly used algorithms, providing evidence of the feasibility to monitor the small sample conditions of high-speed train brake pads.

KFREAIN: Design of A Kernel-Level Forensic Layer for Improving Real-Time Evidence Analysis Performance in IoT Networks

An exponential increase in number of attacks in IoT Networks makes it essential to formulate attack-level mitigation strategies. This paper proposes design of a scalable Kernel-level Forensic layer that assists in improving real-time evidence analysis performance to assist in efficient pattern analysis of the collected data samples. It has an inbuilt Temporal Blockchain Cache (TBC), which is refreshed after analysis of every set of evidences. The model uses a multidomain feature extraction engine that combines lightweight Fourier, Wavelet, Convolutional, Gabor, and Cosine feature sets that are selected by a stochastic Bacterial Foraging Optimizer (BFO) for identification of high variance features. The selected features are processed by an ensemble learning (EL) classifier that use low complexity classifiers reducing the energy consumption during analysis by 8.3% when compared with application-level forensic models. The model also showcased 3.5% higher accuracy, 4.9% higher precision, and 4.3% higher recall of attack-event identification when compared with standard forensic techniques. Due to kernel-level integration, the model is also able to reduce the delay needed for forensic analysis on different network types by 9.5%, thus making it useful for real-time & heterogenous network scenarios.

Engagement Recognition Using a Multi-Domain Feature Extraction Method Based on Correlation-Based Common Spatial Patterns

Engagement ability plays a fundamental role in allocating attentional resources and helps us perform daily tasks efficiently. Therefore, it is of great importance to recognize engagement level. Electroencephalography is frequently employed to recognize engagement for its objective and harmless nature. To fully exploit the information contained in EEG signals, an engagement recognition method integrating multi-domain information is proposed. The proposed method extracts frequency information by a filter bank. In order to utilize spatial information, the correlation-based common spatial patterns method is introduced and extended into three versions by replacing different correlation coefficients. In addition, the Hilbert transform helps to obtain both amplitude and phase information. Finally, features in three domains are combined and fed into a support vector machine to realize engagement recognition. The proposed method is experimentally validated on an open dataset composed of 29 subjects. In the comparison with six existing methods, it achieves the best accuracy of 87.74±5.98% in binary engagement recognition with an improvement of 4.03%, which proves its efficiency in the engagement recognition field.

A multidomain bio-inspired feature extraction and selection model for diabetic retinopathy severity classification: an ensemble learning approach

Diabetes retinopathy (DR) is one of the leading causes of blindness globally. Early detection of this condition is essential for preventing patients' loss of eyesight caused by diabetes mellitus being untreated for an extended period. This paper proposes the design of an augmented bioinspired multidomain feature extraction and selection model for diabetic retinopathy severity estimation using an ensemble learning process. The proposed approach initiates by identifying DR severity levels from retinal images that segment the optical disc, macula, blood vessels, exudates, and hemorrhages using an adaptive thresholding process. Once the images are segmented, multidomain features are extracted from the retinal images, including frequency, entropy, cosine, gabor, and wavelet components. These data were fed into a novel Modified Moth Flame Optimization-based feature selection method that assisted in optimal feature selection. Finally, an ensemble model using various ML (machine learning) algorithms, which included Naive Bayes, K-Nearest Neighbours, Support Vector Machine, Multilayer Perceptron, Random Forests, and Logistic Regression were used to identify the various severity complications of DR. The experiments on different openly accessible data sources have shown that the proposed method outperformed conventional methods and achieved an Accuracy of 96.5% in identifying DR severity levels.

Elevator Fault Diagnosis Method Based on IAO-XGBoost under Unbalanced Samples

Elevators are essential tools in daily life; timely and accurate fault diagnosis plays a crucial role in ensuring their safe operation. However, the existing elevator fault diagnosis methods often neglect the imbalance between the actual collected normal samples and the fault samples, resulting in low diagnostic accuracy. In this study, we propose an improved Aquila optimizer (IAO) extreme gradient boosting tree (XGBoost)-based elevator fault diagnosis method under unbalanced samples. The proposed method includes three main components: multi-domain feature extraction, sample balancing, and fault diagnosis. In the feature extraction phase, the time domain, frequency domain and entropy features of the vibration signal are extracted. In the sample balance phase, aiming at the problem of unbalanced fault samples, after feature selection using recursive feature elimination (RFE), the minority class samples are oversampled by applying SMOTE-Tomek. In the fault diagnosis phase, IAO is used to optimize the hyperparameters in the XGBoost, and the optimized hyperparameters are brought into XGBoost for fault diagnosis. The fault diagnosis accuracy of the method proposed in this study can reach 99.06%, and the method can accurately identify the fault state of the elevator.

A study on feature selection using multi-domain feature extraction for automated k-complex detection.

K-complex detection plays a significant role in the field of sleep research. However, manual annotation for electroencephalography (EEG) recordings by visual inspection from experts is time-consuming and subjective. Therefore, there is a necessity to implement automatic detection methods based on classical machine learning algorithms. However, due to the complexity of EEG signal, current feature extraction methods always produce low relevance to k-complex detection, which leads to a great performance loss for the detection. Hence, finding compact yet effective integrated feature vectors becomes a crucially core task in k-complex detection. In this paper, we first extract multi-domain features based on time, spectral analysis, and chaotic theory. Those features are extracted from a 0.5-s EEG segment, which is obtained using the sliding window technique. As a result, a vector containing twenty-two features is obtained to represent each segment. Next, we explore several feature selection methods and compare their performance in detecting k-complex. Based on the analysis of the selected features, we identify compact features which are fewer than twenty-two features and deemed as relevant and proceeded to the next step. Additionally, three classical classifiers are employed to evaluate the performance of the feature selection models. The results demonstrate that combining different features significantly improved the k-complex detection performance. The best performance is achieved by applying the feature selection method, which results in an accuracy of 93.03%7.34, sensitivity of 93.81%5.62%, and specificity 94.135.81, respectively, using a smaller number of the combined feature sets. The proposed method in this study can serve as an efficient tool for the automatic detection of k-complex, which is useful for neurologists or doctors in the diagnosis of sleep research.

Interpretable Emotion Classification Using Multidomain Feature of EEG Signals

Research on affective computing by physiological signals has become a new and important research branch of artificial intelligence. However, most of the research focuses only on the improvement of emotion classification models while neglecting the correlation between subjective emotional labels and implicit EEG feature information. In this study, we proposed an interpretable emotion classification framework based on food images as visual stimuli and a proposed EEG dataset acquired by a portable single-electrode EEG device. Then, we construct a multi-domain feature extraction method based on sliding time windows, which can effectively extract single-channel EEG features. Finally, we chose the extreme gradient boosting (XGBoost) model as our classifier in the proposed classification framework. The optimal accuracy of our single-channel EEG classification model was 94.76%, which reached the classification performance of multi-channel EEG classification. Based on XGBoost, we interpret the global and the local emotion feature selection method by calculating Shapley additive explanation (SHAP) values. Through the analysis of interpretable methods, we can not only obtain the contribution degree of features from three types of emotional labels (positive, negative, and neutral), but also know whether features have a positive or negative effect on classification results. The source codes of this work are publicly available at:https://github.com/VCMHE/MDF-EEG.

Development of Structural Damage Detection Method Working with Contaminated Vibration Data via Autoencoder and Gradient Boosting

Vibration-based structural damage detection is one of the most promising venues for building smart and automated structural health monitoring applications; however, its applicability is impeded by a large amount of collected vibration data, and the performance could be undermined by degraded data. Therefore, this study develops a robust framework, dubbed AutoBoost-SDD, that can effectively handle contaminated vibration data and provide reliable monitoring results within reasonable computational resources. The proposed method consists of three key components. Firstly, multi-domain feature extraction techniques are utilized to convert high-dimensional raw data into informative feature vectors. Secondly, the auto-encoder deep learning architecture is leveraged to refine feature vectors of contaminated data. Finally, a tree-based boosting machine learning algorithm, namely LightGBM, is employed to assess the structures’ operational states using learned output from the second step. The viability and performance of the proposed framework are illustrated via three case studies involving numerical data of a 5-degree of freedom system, a 2D frame structure, and experimental data of a large-scale 18-story frame structure from the literature. The results show that the AutoBoost-SDD framework is able to provide reasonable detection results despite the presence of various contaminations, including noisy, missing, and anomalous data.

An imbalanced sample intelligent fault diagnosis method using data enhancement and improved broad learning system

Broad learning system (BLS) has been widely applied in the field of fault diagnosis because of its high computational efficiency, simple structure, and strong interpretability. However, traditional BLS cannot extract deep level fault features. Meanwhile, some fault samples are difficult to obtain, which leads to the imbalance of samples and further affects the diagnostic results of BLS. To solve these problems, an improved BLS fault diagnosis method based on data enhancement and multi-domain feature fusion is proposed. First, to solve the problem of sample imbalance, some false samples are generated through deep convolutional generative adversarial networks. Second, to solve the problem of poor feature extraction ability of BLS, the multi-domain feature extraction and feature optimization based on ReliefF algorithm are carried out for the enhanced samples. Compared with traditional BLS, the improved BLS effectively solves the problem of sample imbalance and greatly improves the diagnostic accuracy. The proposed method is then testified on the rolling bearing fault simulation test bench. The results show that, samples generated by the proposed method are highly similar to the real samples. In addition, the diagnostic accuracy of the BLS after multi-domain feature extraction and optimization is improved by about 19.67%, which proves the effectiveness of the method. This method provides a new perspective in fault diagnosis and could further expand the application of BLS in fault diagnosis.

Design of subject independent 3D VAD emotion detection system using EEG signals and machine learning algorithms

This work aims to develop a subject-independent Emotion Detection System (EDS) based on EEG signals and the 3D Valence-Arousal-Dominance (VAD) model. The DEAP database physiological signals are considered for the system design. A multi-domain feature extraction is performed using the Wavelet-based Atomic Function time–frequency domain technique; and various time and frequency domain feature extraction techniques. Further, principal component analysis reduces the data dimensionality and redundancy in the obtained feature set. The minimal feature set is analysed using machine learning classifiers, i.e., gradient boosting, decision tree, and random forest. Additionally, the hyperparameters of machine learning algorithms are tuned using Optuna to improve the performance of the proposed model. Three EDS are designed in this work; EDS1 considers the 3D VAD model for three class classifications. 2D VA model is used in EDS2 to determine 9 discrete emotions, and EDS3 detects 12 discrete emotions using the 3D VAD model. Results reveal the highest classification accuracy of about 99% with EDS1, whereas an average accuracy of 99.82% and 98.44% is obtained with EDS2 and EDS3, respectively. The results reveal that both EDS1 and EDS2 show improvement as compared to the existing literature. Also, the proposed EDS3 provides the classification of the highest number of discrete emotions, which may facilitate the design of an efficient human–computer interface system.

Simulation of big data mixed attribute feature detection for power system intelligent operation and maintenance based on improved random forest algorithm

The diversity of attribute categories brings certain difficulties to data feature detection. In order to improve the accuracy and efficiency of feature detection, a hybrid attribute feature detection method for power system intelligent operation and maintenance big data based on improved random forest algorithm is proposed. Clustering processing power system intelligent operation and maintenance big data, based on data clustering results to reduce the characteristics of data mixed attributes, reduce the scale of data processing, and discretize the data mixed attributes; BP neural network is used to preprocess the results. Make corrections to improve the accuracy of feature detection, use the improved random forest algorithm to classify the data, and improve the convergence speed of the method. Finally, the support vector machine method is used to realize the feature detection of data mixed attributes. The experimental results show that the feature detection accuracy and efficiency of the method designed in this paper are high, and more features can be detected, which verifies its effectiveness. The method designed in this paper has the minimum RMSE value and the maximum value is only 0.12, which is far lower than the RMSE value of the improved spectral clustering algorithm and multi-domain feature extraction method, and has high detection accuracy.

A RUSBoosted tree method for k-complex detection using tunable Q-factor wavelet transform and multi-domain feature extraction.

K-complex detection traditionally relied on expert clinicians, which is time-consuming and onerous. Various automatic k-complex detection-based machine learning methods are presented. However, these methods always suffered from imbalanced datasets, which impede the subsequent processing steps. In this study, an efficient method for k-complex detection using electroencephalogram (EEG)-based multi-domain features extraction and selection method coupled with a RUSBoosted tree model is presented. EEG signals are first decomposed using a tunable Q-factor wavelet transform (TQWT). Then, multi-domain features based on TQWT are pulled out from TQWT sub-bands, and a self-adaptive feature set is obtained from a feature selection based on the consistency-based filter for the detection of k-complexes. Finally, the RUSBoosted tree model is used to perform k-complex detection. Experimental outcomes manifest the efficacy of our proposed scheme in terms of the average performance of recall measure, AUC, and F10-score. The proposed method yields 92.41 ± 7.47%, 95.4 ± 4.32%, and 83.13 ± 8.59% for k-complex detection in Scenario 1 and also achieves similar results in Scenario 2. The RUSBoosted tree model was compared with three other machine learning classifiers [i.e., linear discriminant analysis (LDA), logistic regression, and linear support vector machine (SVM)]. The performance based on the kappa coefficient, recall measure, and F10-score provided evidence that the proposed model surpassed other algorithms in the detection of the k-complexes, especially for the recall measure. In summary, the RUSBoosted tree model presents a promising performance in dealing with highly imbalanced data. It can be an effective tool for doctors and neurologists to diagnose and treat sleep disorders.