Multivariate Feature Extraction Research Articles

How many specimens make a sufficient training set for automated three-dimensional feature extraction?

Deep learning has emerged as a robust tool for automating feature extraction from three-dimensional images, offering an efficient alternative to labour-intensive and potentially biased manual image segmentation methods. However, there has been limited exploration into the optimal training set sizes, including assessing whether artficial expansion by data augmentation can achieve consistent results in less time and how consistent these benefits are across different types of traits. In this study, we manually segmented 50 planktonic foraminifera specimens from the genus Menardella to determine the minimum number of training images required to produce accurate volumetric and shape data from internal and external structures. The results reveal unsurprisingly that deep learning models improve with a larger number of training images with eight specimens being required to achieve 95% accuracy. Furthermore, data augmentation can enhance network accuracy by up to 8.0%. Notably, predicting both volumetric and shape measurements for the internal structure poses a greater challenge compared with the external structure, owing to low contrast differences between different materials and increased geometric complexity. These results provide novel insight into optimal training set sizes for precise image segmentation of diverse traits and highlight the potential of data augmentation for enhancing multivariate feature extraction from three-dimensional images.

Spatio-temporal multivariable time vario-zoom network for water level forecasting based on high-resolution hydrological dataset

Water level forecasting is a crucial element in river-basin water resource management, particularly for shipping and extreme events, such as floods and droughts. However, accurately forecasting water levels is challenging due to the influence of non-stationarity hydrological processes, complex characteristic variables, and noisy data. Data-driven approaches in water level forecasting generally lack annotated high-resolution data. To address these problems, this paper presents the first hydrological dataset on the Xijiang River basin-Xijiang Water Dataset (XJWD) and a Spatio-Temporal Multivariable-based Time Vario-Zoom Network (STM- TVZN) model. XJWD combines multi-source raw hydrological data and has real-time and higher resolution data than comparable datasets. STM-TVZN is specifically tailored to enhance the discernment of salient features within key time windows for each variable by fusing spatio-temporal multivariate features through vario-zoom time windows. Finally, a novel metric, Extremum Time Offsets (ETO), is proposed to evaluate the accuracy of the occurrence of local extremum in time series, and it is significant for predicting and evaluating extremum events such as floods and droughts. Experimental results show that the proposed method outperformed existing models on XJWD dataset regarding error rate, stationarity, similarity, and ETO. Overall, the proposed method of combining time-variable focusing network design and spatio-temporal multivariate feature extraction is appealing for short-term water level forecasting in inland rivers, exhibiting potential capacity in strategic water-saving planning and extremum event forecasting caused by climate change.

AC Series Arc Fault Diagnosis Method Based on AO-VMD Multidimensional Feature Extraction

In view of the difficulty in determining the number of mode components K and penalty factor α in VMD, which leads to poor signal decomposition effect and low diagnosis and recognition rate due to insufficient feature extraction in AC arc fault, an arc fault diagnosis method based on the Aquila algorithm was proposed to optimize the multivariate feature extraction of variational mode decomposition. First of all, the arc fault test platform was set up to consider the resistive, inductive, capacitive, and other household loads were considered to obtain the normal and fault current data. Second, the optimal K and α parameters were obtained by AO-VMD optimization and then decomposed into intrinsic mode functions (IMFs) by substituting them into VMD. Then, the time domain characteristics of the IMF2 component, the Energy Entropy of IMFs, and the Fuzzy Entropy of the kurtosis maximum IMFk component were extracted respectively, and the multidimensional fault characteristic matrix was constructed. Finally, the random forest (RF) model was used to accurately identify arc faults. The experiment shows that the average fault recognition rate of each type of load is above 99%, which has an excellent diagnostic effect.

One-Class Machine Learning Classifiers-Based Multivariate Feature Extraction for Grid-Connected PV Systems Monitoring under Irradiance Variations

In recent years, photovoltaic (PV) energy production has witnessed overwhelming growth, which has inspired the search for more effective operations. Nevertheless, different PV faults may appear, which leads to various degradation stages. Furthermore, under different irradiance levels, these faults may be misclassified as a healthy mode owing to the high resemblances between them, thus provoking serious challenges in terms of power losses and maintenance costs. Hence, interposing the irradiance variation in grid-connected PV (GCPV) systems modeling is important for monitoring tasks to ensure the effective operation of these systems, to increase their reliability and to prevent false alarms. Therefore, in this paper, a fault detection and diagnosis (FDD) method for the GCPV systems using machine learning (ML) based on principal component analysis (PCA) is proposed in order to ensure the reliability and security of the whole system under irradiance variations. The proposed strategy consists of three main steps: (i) introduce the irradiance variations in PV system modeling because of its great impact on power production; (ii) feature extraction and selection through PCA; and (iii) fault classification using ML techniques. In this study, we generate a database that is used to compare the proposed strategy with the standard strategy (considering a fixed irradiance during FDD), to make, at first, a complete and significant comparative assessment of fault diagnosis and to demonstrate the efficiency of the proposed strategy. The achieved results show the high effectiveness of the proposed one-class classification-based approach to detect and diagnose PV array anomalies, reaching an accuracy up to 99.68%.

Spatiotemporal wind power forecasting approach based on multi-factor extraction method and an indirect strategy

Accurate ultra-short-term wind power forecasting is a prerequisite for decision making related to the management of power systems. Existing approaches used to forecast wind power ignored the correlation between wind power outputs under similar wind power, which provides important information for wind power forecasting. A spatiotemporal approach based on an indirect strategy and a multi-factor extraction model is proposed in this study to achieve more accurate power prediction. Specifically, wind power and other variables are classified into four categories according to wind direction, and multivariate feature extraction and wind power forecasting are performed separately for each category. In the experimental study, taking two real-world datasets of wind farms in Northeast China as examples, the proposed approach is compared with eight benchmarking approaches, and the results demonstrate the effectiveness and robustness of our approach. In addition, we further conduct the DM test, and the sensitivity analysis to discuss the effect of hyperparameters on forecasting performance. The results of DM verify the superiority of the proposed approach in statistic. This study provides a new high-precision approach and a new forecasting strategy for future wind power forecasting.

Multi Variate Feature Extraction and Feature Selection using LGKFS Algorithm for Detecting Alzheimer’s Disease

Multi Variate Feature Extraction and Feature Selection using LGKFS Algorithm for Detecting Alzheimer’s Disease

Remote sensing multi-scale object detection based on multivariate feature extraction and characterization optimization

Remote sensing multi-scale object detection based on multivariate feature extraction and characterization optimization

A novel artificial neural network improves multivariate feature extraction in predicting correlated multivariate time series

The existing multivariate time series prediction schemes are inefficient in extracting intermediate features. This paper proposes an artificial neural network called Feature Path Efficient Multivariate Time Series Prediction (FPEMTSP) to predict the next element of the main time series in the presence of several secondary time series. We propose to generate all the possible combinations of the secondary time series and extract multivariate features by doing the Cartesian product of the main and the secondary time series features. Our calculations prove that the FPEMTSP’s complexity and network size are acceptable. We have considered a few internal parameters in FPEMTSP that can be configured to improve the prediction accuracy and adjust the network size. We trained and evaluated FPEMTSP using two public datasets. Our evaluation revealed the optimal values for the internal parameters and showed that FPEMTSP surpasses the existing schemes in terms of prediction accuracy and the number of correctly predicted steps.

Multivariate EEG Signal Processing Techniques for the Aid of Severely Disabled People

Electroencephalography (EEG) has been used for several years as a trace of signals for facilitating subjects with serious infirmities to communicate with computers and other devices. Many studies have revealed the correlation of mental tasks with the EEG signals for actual or fictional movements. However, the performance of Brain Computer Interface (BCI) using EEG signal is still below enough to assist any disabled people. One reason could be that the researchers in this field (motor imagery based BCI) normally use two to three channels of EEG signal. This might limit the performance of BCI, as an extra source of information generally helps in detecting a person's motor movement intentions. Therefore, the proposed research work is involved with three or more channels of EEG signal for online BCI. Two fundamental objectives for BCI based on motor movement imagery from multichannel signals are aimed at in this research work: i) to develop a technique of multivariate feature extraction for motor imagery related to multichannel EEG signals; and ii) to develop an appropriate machine learning based feature classification algorithm for Brain Computer Interface. Nevertheless, all other problems like interfacing and real-time operations with current BCIs are also addressed and attempts are made to reduce the problems. The methodology can be described by following steps as follows: i) at least 3 channels of EEG signal are recorded; ii) a few features are extracted from preprocessed EEG signal; iii) all extracted features are classified to generate commands for BCI; iv) finally evaluate the performance of the proposed algorithm for BCI. The challenge of this research work is to investigate and find an appropriate model for online (real-time) BCI with a realistic performance to be made in achieving better lives for people with severe disabilities in Malaysia and abroad.

Study on microstructure damage characteristics of asphalt mixture based on multivariate feature extraction

Study on microstructure damage characteristics of asphalt mixture based on multivariate feature extraction

Study on microstructure damage characteristics of asphalt mixture based on multivariate feature extraction

Study on microstructure damage characteristics of asphalt mixture based on multivariate feature extraction

3D Geophysical Post-Inversion Feature Extraction for Mineral Exploration through Fast-ICA

A major problem in the post-inversion geophysical interpretation is the extraction of geological information from inverted physical property models, which do not necessarily represent all underlying geological features. No matter how accurate the inversions are, each inverted physical property model is sensitive to limited aspects of subsurface geology and is insensitive to other geological features that are otherwise detectable with complementary physical property models. Therefore, specific parts of the geological model can be reconstructed from different physical property models. To show how this reconstruction works, we simulated a complex geological system that comprised an original layered Earth model that has passed several geological deformations and alteration overprints. Linear combination of complex geological features comprised three physical property distributions: electrical resistivity, induced polarization chargeability, and magnetic susceptibility models. This study proposes a multivariate feature extraction approach to extract information about the underlying geological features comprising the bulk physical properties. We evaluated our method in numerical simulations and compared three feature extraction algorithms to see the tolerance of each method to the geological artifacts and noises. We show that the fast-independent component analysis (Fast-ICA) algorithm by negentropy maximization is a robust method in the geological feature extraction that can handle the added unknown geological noises. The post-inversion physical properties were also used to reconstruct the underlying geological sources. We show that the sharpness of the inverted images is an important constraint on the feature extraction process. Our method successfully separates geological features in multiple 3D physical property models. This methodology is reproducible for any number of lithologies and physical property combinations and can recover the latent geological features, including the background geological patterns from overprints of chemical alteration.

Multivariate feature extraction based supervised machine learning for fault detection and diagnosis in photovoltaic systems

Fault detection and diagnosis (FDD) in the photovoltaic (PV) array has become a challenge due to the magnitudes of the faults, the presence of maximum power point trackers, non-linear PV characteristics, and the dependence on isolation efficiency. Thus, the aim of this paper is to develop an improved FDD technique of PV systems faults. The common FDD technique generally has two main steps: feature extraction and selection, and fault classification. Multivariate feature extraction and selection is very important for multivariate statistical systems monitoring. It can reduce the dimension of modeling data and improve the monitoring accuracy. Therefore, in the proposed FDD approach, the principal component analysis (PCA) technique is used for extracting and selecting the most relevant multivariate features and the supervised machine learning (SML) classifiers are applied for faults diagnosis. The FDD performance is established via different metrics using data extracted from different operating conditions of the grid-connected photovoltaic (GCPV) system. The obtained results confirm the feasibility and effectiveness of the proposed approaches for fault detection and diagnosis.

RETRACTED: Fault diagnosis of heating systems using multivariate feature extraction based machine learning classifiers

Abstract An essential tool for an effective monitoring of heating systems is the accurate Fault Detection and Diagnosis (FDD). Moreover, the implementation of high-efficiency FDD techniques leads to the reduction of building's energy consumption and the improvement of the comfort level. Thus, this paper proposes an enhanced FDD approach based on multiscale representation, principal component analysis (PCA), and machine learning (ML) classifiers. The main goal of this paper is to merge the benefits of the aforementioned techniques to improve the efficiency of the detection and diagnosis of failures occurring in heating systems. First, a multiscale decomposition (MSD) is used to extract the dynamics of the systems at different scales. The multiscale representation is very suitable for the monitoring of heating systems, which are generally characterized by different dynamics in time and frequency frameworks. Then, the multiscaled data-set are introduced into the PCA to extract more efficient features. Finally, the ML classifiers are applied to the extracted and selected features to address the problem of fault diagnosis. The effectiveness and higher classification accuracy of the developed approach is demonstrated using two examples: synthetic data and simulated data extracted from heating systems.

Identification of NR and EPDM Samples by Means of Thermogravimetric Analysis and Multivariate Methods

Products based on ethylene-propylene-diene m-class (EPDM) and natural rubber (NR) are widely used in a different applications, including the automotive industry, heating, ventilation, and air conditioning applications, roofing systems, or the construction sector among others. The growing demand of these types of polymeric products has forced rubber industry to implement strict control schemes to ensure the specifications of the final products. The focus of this paper is the identification of different treatment types of NR and EPDM samples without any preceding analytical treatment to carry out a fast and reliable supervision of the vulcanization processes to improve the quality of the final rubber products. To this end, the thermogravimetric analysis (TGA) technique is applied in combination with the principal component analysis (PCA) and canonical variate analysis (CVA) multivariate feature extraction methods and the k-nearest neighbor (k-NN) classifier. Experimental results prove the suitability of the proposed approach and the potential of the TGA method for a fast supervision of the vulcanization processes. Using the information provided by the TGA technique in association with the PCA + CVA + k-NN approach, the system achieved 100% identification accuracy.

Recovered Paperboard Samples Identification by Means of Mid-Infrared Sensors

Paperboard is widely and increasingly applied as a packaging material, and in many applications is in direct contact with foodstuff. The increasing use of recovered paperboard has led to the production of paperboard containing several types of contaminants. In the case of using recovered paperboard, some of these contaminants may migrate into the food in concentrations considered harmful to human health. To prevent this problem, a very fast and nondestructive method to identify recovered paperboard samples from those produced mainly from virgin fibers is developed in this paper. Therefore, recovered samples may be identified, so a special consideration may be given to these samples. To this end, Fourier transform mid-infrared spectroscopy is applied to acquire the mid-infrared spectra of the paperboard samples. Next, statistical multivariate feature extraction and classification methods are applied to identify incoming samples produced from recovered fibers. Experimental results presented here prove that the proposed scheme allows obtaining high classification accuracy with a very fast response.

Surface texture characterization of an Italian pasta by means of univariate and multivariate feature extraction from their texture images

Surface texture is an important characteristic of foods, as well as color, shape, consistency and taste. It plays an important role in consumers' decision and it can affect the properties of a product during its preparation. This work shows the ability of three different image analysis techniques to characterize the surface texture of three Italian pasta samples. The first method is based on the evaluation of Heterogeneity (HTG); the second on the gray level co-occurrence matrix (GLCM) and Haralic statistics; the third, the angle measure technique (AMT), is based on image multivariate feature extraction. The results obtained showed that it is possible to highlight differences in the surface aspect of pasta samples both before and after cooking, and that it is also possible to correlate them to some of their chemical–physical characteristics (e.g., total starch and protein contents, solids lost in the cooking water, pasta adhesiveness; r>0.6, p<0.05). A partial least square discriminant analysis (PLS-DA) applied on GLCM and AMT results allowed the classification of the different pasta samples only on the basis of their surface texture features (sensitivity>0.963; specificity>0.648).

Multivariate feature extraction from textural images of bread

In order to compute the classical texture measures there is often a need to perform extensive calculations on the images and do a preprocessing in a specialised manner. Some of these texture measures are constructed to estimate specific information. Other texture measures seem to be more global in nature. The techniques presented in this paper define algorithms applied on the raw image without extensive preprocessing. We want to show that mathematical transformations of images on a vectorised form will easily enable the use of multivariate techniques and possibly model several features hidden in the images at the same time. In this paper we will compare five different methods of extracting features from textural images in food by multivariate modelling of the sensory porosity of wheat baguettes. The sample images are recorded from factorial designed baking experiments on wheat baguettes. The multivariate feature extraction methods to be treated are the angle measure technique (AMT), the singular value decomposition (SVD), the autocorrelation and autocovariance functions (ACF) and the so-called size and distance distribution (SDD) method. The methods will be tested on equal basis and the modelling of sensory porosity from extracted features is done using principal component regression (PCR) and partial least square regression (PLS). The difference between the behaviour of the methods will be discussed. The results show that all the methods are suited to extract sensory porosity but the AMT method prove to be the best in this case.