•Hybrid solar forecasting method exhibits generalized and improved performance.•Novel data preprocessing method constructs more distinguishing input data.•Deep feature extraction network obtains ...high-level features.•Experiments are verified in different weather conditions.•Integrated algorithms are used for feature selection and forecasting.
In this study, a novel deep solar forecasting approach is proposed based on the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), continuous wavelet transform (CWT), feature extraction networks, RReliefF feature selection, and extreme learning machine (ELM). The global solar radiation is decomposed into mode functions with the CEEMDAN method. The CWT reconstructs one-dimensional data into two-dimensional scalogram images to include both frequency and the time of the daily and hourly correlations. For the feature extraction process, a cascade convolutional neural network architecture, which consists of AlexNet and GoogLeNet, was designed to extract distinctive deep features. As the high-performance features provide a high level of forecasting accuracy, these are concatenated as the subset feature vector and RReliefF utilized to rank and select the most distinctive features from the subset. The designed ELM is then trained with the selected features and the fully-trained ELM model is then used to evaluate the forecast performance. In the experiments, root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) of the proposed method were observed as 0.0642, 0.0241, and 0.1201 for one-step ahead, 0.0686, 0.0285, and 0.1279 for two-step ahead, and 0.0724, 0.0315, and 0.1317 for three-step ahead, respectively. The obtained results show that the proposed method exhibits accurate and robust forecasting performance and outperforms conventional regression models.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
With deepened interactions between human and computer, the need for a reliable and practical system for emotion recognition has become significant. The aim of this study is to propose a practical ...system for estimation of a continuous measure of valence based on a few number of EEG channels.
A vast spectrum of time, frequency and coherence features were implemented with linear Regression (LR), Support Vector Regression (SVR) and Multi-Layer Perceptron (MLP) models and then ranked for the performance on DEAP database using a regression-based Relief filter. Regression outcomes were also classified to compare the performance of the proposed method with the literature. Finally, a video-based emotion recognition experiment was designed and conducted on 12 subjects using F7, F8, FC2 and T7 electrodes.
Magnitude Squared Coherence Estimate(MSCE) on F7–F8 with SVR model provided the highest performance on DEAP dataset. Classification of the output led to an average accuracy of 67.5%. For the gathered data, combination of MSCE and Hilbert–Huang Spectrum provided the best performance with 0.22 root mean square error and 0.67 correlation with self-reported valence in the scale of 1–9.
MSCE could provide a good accuracy in estimation of Valence using 2 EEG channels on Deep dataset, and with addition of Hilbert–Huang Spectrum, it also demonstrated good accuracy and correlation with self-reported valence, in a completely different experiment.
Continuous-value estimation of the valence can be achieved with only 2 EEG channels for practical applications out of the lab.
Display omitted
•Frequency and coherence features outperform time features for estimation of valence.•Changing in valence is more correlated with prefrontal and temporal regions.•Number of channels was reduced to only two while achieving 0.22 regression error.•Combination of frequency features improves the emotional recognition error.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
In this research, machine learning (ML) based frameworks are developed to predict the in-plane mixed-mode fracture load of asphalt mixtures. In an initial stage, the RReliefF technique guides the ...selection of four parameters out of eight, namely fracture toughness, T-stress term, and modes I and II stress intensity factors (SIFs), to serve as inputs for predictive models. Subsequently, three ML models, namely support vector machine regressor (SVR), extra tree regressor (ETR), and gradient boosting regressor (GBR), are trained and tested using 675 experimental data points. Optimal hyper-parameter values for each model are determined through the particle swarm optimization (PSO) technique. To leverage the strengths of each individual model, two techniques, ensemble voting and stacking, are employed to combine the individual models. The performance of the presented models is assessed using 88 previously unseen datasets. The results underscore the significant promise of ML approaches for predicting the fracture load of asphalt mixture components, achieving accuracies of 90.48%, 91.11%, and 90.38% for SVR, ETR, and GBR, respectively. Notably, ensemble voting and stacking techniques further enhance predictive accuracy, achieving impressive accuracies of 91.25% and 91.57% respectively. Ultimately, model interpretation is accomplished via individual conditional expectation (ICE) plots, and the correlation of each predictor with the output is determined, which closely aligned with previous research findings and experimental observations. Our findings underscore the substantial potential of ML approaches in studying the fracture behavior of asphalt mixture components, with implications for enhancing infrastructure durability and safety. Compared to traditional analytical methods, ML-based frameworks offer improved accuracy and robustness in modeling the complex behavior of asphalt mixtures under varying conditions, thereby facilitating more precise assessments of infrastructure performance and durability.
•Three ML models were built to predict the fracture load of asphalt mixtures.•Models were trained using 675 empirical data and assessed using 88 unseen data.•Using PSO enabled accurate choice of optimal hyper-parameters for each ML model.•Using RReliefF, fracture toughness emerged as the key feature for failure load.•ML models were interpreted to identify the correlation between predictors and target.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
Objective: A new approach, named bilateral motion data fusion, was proposed for the analysis of movement symmetry, which takes advantage of cross-information between both sides of the body and ...processes the unilateral motion data at the same time. Methods: This was accomplished using canonical correlation analysis and joint independent component analysis. It should be noted that human movements include many categories, which cannot be enumerated one by one. Therefore, the gait rhythm fluctuations of the healthy subjects and patients with neurodegenerative diseases were employed as an example for method illustration. In addition, our model explains the movement data by latent parameters in the time and frequency domains, respectively, which were both based on bilateral motion data fusion. Results: They show that our method not only reflects the physiological correlates of movement but also obtains the differential signatures of movement asymmetry in diverse neurodegenerative diseases. Furthermore, the latent variables also exhibit the potentials for sharper disease distinctions. Conclusion: We have provided a new perspective on movement analysis, which may prove to be a promising approach. Significance: This method exhibits the potentials for effective movement feature extractions, which might contribute to many research fields such as rehabilitation, neuroscience, biomechanics, and kinesiology.
A complete fault diagnosis for the rolling bearing is proposed in this paper. Variable predictive model class discrimination (VPMCD) is a conventional pattern recognition method; however, in ...practice, when the fault diagnosis method is applied to small samples or in multi-correlative feature space, the stability of the VPM constructed based on the least squares (LS) method is not sufficient. Based on affinity propagation (AP) clustering, RReliefF, and sequential forward search, the ARSFS is proposed to select the significant subset of original feature set and to reduce the dimension and multiple correlations of the feature space. Further, this paper uses two kinds of Gaussian Neural Network, namely the Radial Basis Function Neural Network (RBF) and the Generalized Regression Neural Network (GRNN), instead of the LS method to construct predictive models of VPMCD, called AOR-VPMCD. Compared with the conventional VPMCD and its improvements, based on sufficient experiments, the entire process presented in this paper can effectively identify the fault of the rolling bearing.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UL, UM, UPCLJ, UPUK, ZRSKP
•In this paper an accurate method for fault location in series-compensated transmission lines (SCTLs) is presented.•Adaptive network-based fuzzy inference system (ANFIS) is used as a intelligent ...network.•The desired features are obtained using the Least-Squares Estimation (LSE).•The subtractive clustering technique has been utilized to design the basic ANFIS.•This study has scrutinized different cases, including two different ANFIS arrangements, implementing a single network to perform fault location on the entire line and a separate network for each line segment.•Moreover, the efficiency of feature selection has been analyzed using the RReliefF algorithm.•The results validate that the proposed approach has accurately located the fault under a wide range of system variations.
In this paper a new and accurate method for fault location in series-compensated transmission lines (SCTLs) is presented using adaptive network-based fuzzy inference system (ANFIS). In the proposed approach, the desired features are obtained using the Least-Squares Estimation (LSE) from two cycles of voltage and current data in only one line terminal. After that, extracted features, including the decaying DC offset component of the current, the fundamental components of current and voltage, and the phase difference between current and voltage for three-phase current and voltage signals, are normalized and applied as inputs to the fuzzy neural network for fault location. The subtractive clustering technique has been utilized to design the basic ANFIS. This study has scrutinized different cases, including two different ANFIS arrangements, implementing a single network to perform fault location on the entire line and a separate network for each line segment. Moreover, the efficiency of feature selection has been analyzed using the RReliefF algorithm. Various training patterns and tests have been provided in a test transmission system for different system conditions, such as different fault inception angles, fault locations, fault resistances, and structural conditions. The results validate that the proposed approach has accurately located the fault under a wide range of system variations.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
The assessment of structural damage is of great significance for ensuring the service safety of carbon fiber reinforced plastics (CFRP) structures. In this paper, the damage degree prediction method ...of CFRP structure based on fiber Bragg grating and epsilon-support vector regression was studied. The structural dynamic response signals were detected by fiber Bragg grating sensors. Then, the Fourier transform was used to extract the dynamic characteristics of the structure as the damage feature, and the damage feature dimensionality was reduced by using the RReliefF algorithm. On this basis, the damage degree prediction model of CFRP structure based on epsilon-support vector regression was established. Finally, the method proposed in this paper was experimentally verified. The results showed that the epsilon-support vector regression model can accurately predict the damage degree of unknown samples, and the absolute relative error of 27 experiments was less than 10% for 30 testing experiments. This paper provided a feasible method for predicting the damage degree of CFRP structures.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
Finding the reduced and the relevant subsets of the predictors is inevitable when it comes to predictive modelling. If the datasets involved are heterogeneous and heteroscedastic in nature such as ...the soil samples, the task turns out to be trickier and such scenarios demand ensemble-based feature selection approach. The proposed algorithm uses an ensemble of filter (RReliefF), wrapper (Adaptive Plus –l and Minus –r), and embedded (Neighbourhood Component Analysis) approaches and applies it to the datasets in a heterogeneous and homogeneous manner. The Adaptive Plus –l and Minus –r is an experiment done on the Plus –l and Minus –r wrapper method to enhance the performance of the algorithm. The proposed combination rule of the ensemble filters out the irrelevant predictors for each response variable. Further, this ensemble is recursively implemented using the floating set of predictors to estimate the optimal subsets for multiple response variables at one go. Each step in recursion ensures that only the best subset of features (in terms of length and weights) among the current and previous iterations is retained. The Akaikes Information Criteria and the predictors weights calculated, assessed the efficiency of the resultant predictor sets of the recursive ensemble. The adjusted R2, Median Absolute Deviation and Root Mean Square Error on the unseen datasets confirmed the suitability of the same for predictive modelling in ecological domains.
•Finding reduced subsets of features is difficult in the context of heterogeneous and heteroscedastic datasets.•A recursive heterogeneous ensemble of feature selection algorithms is implemented to tackle the stochasticity.•The Plus -l Minus -r wrapper method is improvised to reduce the computational cost by learning from the data.•A unique combination rule to extract the relevant predictors for the response variable.•Multiple response variables are selected in one go using the recursive rule of the algorithm.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
Display omitted
•We evaluated how auto-correlation affects machine learning sugarcane yield models.•We adapted the feature selection RReliefF algorithm for use with auto-correlated data.•Naive ...assumption of data-independence leads to underestimated generalization error.•Proposed protocol improves estimates of generalization error.•Model performance slightly improved without changing the machine learning techniques.
With the increased application of information technology in agriculture, data is being produced and used in an unprecedented scale. While these advances, combined with machine learning techniques, benefited yield modeling, most of the current literature about data-driven yield modeling has not yet accounted for potential sources of correlation in data, assuming independence between samples. In this scenario, random sampling can lead to correlated samples across sets being used for model evaluation. We implemented a spatially-aware protocol and compared it with the naive approach of assuming independence between samples. The protocols were applied through all the model development pipeline: data splitting for hold-out sets, feature selection, cross-validation for model adjustment and model evaluation. Three different machine learning techniques were used to create models in each protocol. The resulting models were evaluated both in the validation set created by each protocol and in a manually created independent set. This independent set ensured there was no auto-correlation between the samples used for modeling. We showed that assuming independence when modeling yield leads to underestimating model errors and overfit during model adjustment. Despite better error tracking, the model with the smallest error in the test set was not the model with the smallest validation error, suggesting overfit for the model selection. While this effect was small for the spatially-aware protocol, the effect was a lot stronger in the naive protocol. Future efforts in yield modeling should address the effect of spatial auto-correlation and other potential sources of correlation to improve correctness and robustness of the results.
Full text
Available for:
GEOZS, IJS, IMTLJ, KILJ, KISLJ, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, UILJ, UL, UM, UPCLJ, UPUK, ZAGLJ, ZRSKP
The abrupt changes in tool-workpiece interaction during machining process induce variation in the surface quality of work material. These interactions include built-up edge formation and their ...break-off, environmental conditions (use of coolant, rise of temperature etc.), material imperfections, improper structural fitness of machine & tool components, etc. This study presents prediction of surface roughness in turning of EN353 steel implementing the variational mode decomposition (VMD) for processing the vibration data, followed by estimation of the surface roughness using the relevance vector regression (RVR) optimized by particle swarm optimization (PSO). The raw vibration data has been decomposed in five discrete sets of frequency components known as variational mode functions (VMFs). A set of twenty-one statistical features in each three axes have been extracted for raw data and each VMF. The RVR has been trained using these 21×3 = 63 features and 3 cutting parameters — cutting speed, feed depth of cut. The RVR has also been trained separately using top 5 features selected through RreliefF algorithm. The optimal decomposition level has been determined to minimize the noise and predict the surface finish accurately. The results obtained in 1st VMF (high frequency, low amplitude) using its top 5 features for prediction have been found to be reliable with higher prediction accuracy.
Full text
Available for:
EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, KILJ, KISLJ, MFDPS, NLZOH, NUK, OILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ