Principal Component Analysis Feature Extraction Research Articles

Hyperspectral Prediction Model of Nitrogen Content in Citrus Leaves Based on the CEEMDAN–SR Algorithm

Nitrogen content is one of the essential elements in citrus leaves (CL), and many studies have been conducted to determine the nutrient content in CL using hyperspectral technology. To address the key problem that the conventional spectral data-denoising algorithms directly discard high-frequency signals, resulting in missing effective signals, this study proposes a denoising preprocessing algorithm, complete ensemble empirical mode decomposition with adaptive noise joint sparse representation (CEEMDAN–SR), for CL hyperspectral data. For this purpose, 225 sets of fresh CL were collected at the Institute of Fruit Tree Research of the Guangdong Academy of Agricultural Sciences, to measure their elemental nitrogen content and the corresponding hyperspectral data. First, the spectral data were preprocessed using CEEMDAN–SR, Stein’s unbiased risk estimate and the linear expansion of thresholds (SURE–LET), sparse representation (SR), Savitzky–Golay (SG), and the first derivative (FD). Second, feature extraction was carried out using principal component analysis (PCA), uninformative variables elimination (UVE), and the competitive adaptive re-weighted sampling (CARS) algorithm. Finally, partial least squares regression (PLSR), support vector regression (SVR), random forest (RF), and Gaussian process regression (GPR) were used to construct a CL nitrogen prediction model. The results showed that most of the prediction models preprocessed using the CEEMDAN–SR algorithm had better accuracy and robustness. The prediction models based on CEEMDAN–SR preprocessing, PCA feature extraction, and GPR modeling had an R2 of 0.944, NRMSE of 0.057, and RPD of 4.219. The study showed that the CEEMDAN–SR algorithm can be effectively used to denoise CL hyperspectral data and reduce the loss of effective information. The prediction model using the CEEMDAN–SR+PCA+GPR algorithm could accurately obtain the nitrogen content of CL and provide a reference for the accurate fertilization of citrus trees.

Machine learning-based evaluation of eccentricity and acoustic impedance in oil well using VDL data

Logging procedure and interpretation are complicated mainly because of the large amount of different data inherited. One way to facilitate the problem is to use fewer tools but keep the same amount of information which can be feasible with the assistance of machine learning models. In this paper, Variable Density Logging data was used to evaluate the eccentricity and the acoustic impedance average of the medium behind the casing of oil wells. Different shallow and deep learning architectures were trained, and a Principal Components Analysis feature extraction approach was investigated. Actual data from acoustic logging tools acquired in different wells were adopted. A supervised learning approach was explored, and the K-Nearest Neighbors model obtained the best results when accounting for the cost-benefit and complexity of the models. State-of-the-art hyperparameter optimization techniques were used to tune the non-trainable parameters of the shallow learning algorithms evaluated. The proposed procedure was able to achieve mean balanced adjacency accuracy results greater than 80% for the free pipe pass and repeated pass of a similar well and results greater than 70% for the main pass of a distinguished oil well; which is a more challenging scenario.

Material Classification and Aging Time Prediction of Structural Metals Using Laser-Induced Breakdown Spectroscopy Combined with Probabilistic Neural Network.

In this paper, laser-induced breakdown spectroscopy (LIBS) combined with a probabilistic neural network (PNN) was applied to classify engineering structural metal samples (valve stem, welding material, and base metal). Additionally, utilizing data from the plasma emission spectrum generated by laser ablation of samples with different aging times, an aging time prediction model based on a firefly optimized probabilistic neural network (FA-PNN) was established, which can effectively evaluate the service performance of structural materials. The problem of insufficient features obtained by principal component analysis (PCA) for predicting the aging time of materials is addressed by the proposal of a time-frequency feature extraction method based on short-time Fourier transform (STFT). The classification accuracy (ACC) of time-frequency features and principal component features was compared under PNN. The results indicate that, in comparison to the PCA feature extraction approach, the time-frequency feature extraction method based on STFT demonstrates higher accuracy in predicting the time of aging materials. Then, the relationship between classification accuracy (ACC) and settings of PNN was discussed. The ACC of the PNN model for both the material classification test set and the aging time test set achieved 100% with Firefly (FA) optimization algorithms. This result was also compared with the ACC of ANN, KNN, PLS-DA, and SIMCA for the aging time test set (95%, 87.5%, 85%, and 62.5%, respectively). The experimental results demonstrated that the classification model using LIBS combined with FA-PNN could realize better classification accuracy.

A novel hybrid ensemble learning for anomaly detection in industrial sensor networks and SCADA systems for smart city infrastructures

Critical Infrastructures (CIs) use Supervisory Control and Data Acquisition (SCADA) systems for monitoring and remote control. Sensor networks are being integrated into all areas of the infrastructures of smart cities. The sensor network data stream contains information that can be utilized to model and control the activity of these infrastructures. However, SCADA systems are constantly exposed to a variety of diverse intrusions, making detection with traditional intrusion detection systems (IDS) extremely difficult. Due to their unique specifications, conventional security solutions, like antivirus and firewall software, are unsuitable for properly securing SCADA systems. In addition, anomaly detection in industrial sensor networks (ISNs) should occur in real time. Therefore, effectively identifying cyberattacks in major SCADA systems is unquestionably essential for enhancing their resilience, ensuring safe operations, and avoiding expensive maintenance. We developed a novel hybrid ensemble model approach to address these issues. This paper's primary objective is to detect hostile intrusions that have already circumvented firewalls and typical IDS. In this paper, we propose a hybrid Ensemble Learning Model (ELM) for intrusion detection in SCADA systems with ISNs utilizing a tangible data gathered from a gas pipeline system given by Mississippi State University (MSU), the water system, and the high-dimensional University of New South Wales-NB 2015 (UNSW-NB15) data that reflects a typical attack in the Internet of Things (IoT) environment. The unity normalization method was adopted for data preprocessing, and the Principal Component Analysis (PCA) was utilized for feature extraction of the high-dimensional datasets. Grey Wolf Optimizer (GWO) was used for optimizing the bagging, stacking, Adaboost, and an ensemble of classifiers Naive Bayes and Support Vector Machine with a majority voting technique. Then, we utilized the proposed approach founded on the bijective soft-set approach for efficient ELM selection. The experiment was conducted in two phases: Initially, without PCA + GWO for feature extraction and selection on the ELM, and subsequently, with PCA + GWO for feature extraction and selection on the ELM. PCA + GWO on the ensemble of classifiers NB + SVM provided an accuracy of 99%, precision of 100%, recall of 100%, and detection rate of 99.90%, outpacing the ensemble of classifiers without PCA feature extraction and GWO optimization approaches.

2D face recognition using PCA and triplet similarity embedding

The aim of this study is to propose a new robust face recognition algorithm by combining principal component analysis (PCA), Triplet Similarity Embedding based technique and Projection as a similarity metric at the different stages of the recognition processes. The main idea is to use PCA for feature extraction and dimensionality reduction, then train the triplet similarity embedding to accommodate changes in the facial poses, and finally use orthogonal projection as a similarity metric for classification. We use the open source ORL dataset to conduct the experiments to find the recognition rates of the proposed algorithm and compare them to the performance of one of the very well-known machine learning algorithms k-Nearest Neighbor classifier. Our experimental results show that the proposed model outperforms the kNN. Moreover, when the training set is smaller than the test set, the performance contribution of triplet similarity embedding during the learning phase becomes more visible compared to without it

Mapping of Shorea kunstleri King in mixed dipterocarp forest using remote sensing technology

The conventional method used in forest inventory is time-consuming, especially in field surveying and requires an expert’s knowledge and observation to do species identification. This standard method can be facilitated by advances in remote sensing technology. An airborne hyperspectral image mosaic was utilised in the research to evaluate the ability of this high spectral fidelity dataset that was incorporated into a feature extraction Support Vector Machine (SVM) classifier to successfully detect and map a tree species of interest (Shorea kunstleri King) within the mixed dipterocarp forest stand of the Lambir Hills National Park. In this study, two types of feature extraction Principal Component Analysis (PCA) and Spectral Derivative (SD) were applied to classifiers SVM. To demonstrate the effect, classification of the hyperspectral image with no feature extraction was done. The accuracies achieved from different feature extraction were compared statistically. Based on the results, Shorea kunstleri King was able to be discriminated against within 53 species of Shorea. SVM classifier is known for its ability to generalise well even with limited training samples and is commonly used in image classification. The Kernel parameter in SVM classifier alongside feature extraction significantly affects the classification accuracy. Therefore, feature extraction does affect the classification accuracy, but it depends on the morphology of the study area. It is concluded that the SVM method with SD feature extraction could detect tree crowns of Shorea kunstleri with a Kappa coefficient of 0.8126 which showed a great agreement with the observed samples and overall accuracy of 89.89% that shows the accuracy of the final map.

Machine learning-driven seismic failure mode identification of reinforced concrete shear walls based on PCA feature extraction

As the major lateral load-resisting components in building structures, reinforced concrete (RC) shear waolls exhibit a powerful capability of load-carrying as well as complicated mechanism under cyclic forces. Scenarios are there in regional risk assessment and post-earthquake renovation calling for stringent prerequisite of rapid failure mode identification. Great endeavor has been dedicated to utilize machine learning (ML) techniques to perform this task, while very few of them has paid adequate emphasis on exploratory data analysis or feature extraction during implementation. Hence, this study explored out a novel strategy by allocating principal component analysis (PCA) for experimental data mining and feature extraction, which was proved a powerful capability of enhancing the prediction accuracy compared with training with raw data. Based on the assembled dataset consisting of 181 RC shear walls by this study, nine classical and state-of-the-art ML algorithms were investigated, among which gradient boosting classifier achieved a relevantly high accuracy of 0.98 after performed with hyperparameter optimization. Besides, clustering patterns of experimental parameters in PCA analysis and feature importance in fitted model explainability were elaborated as well, which presented valuable information and consistent underlying mechanism. As the first attempt in RC shear walls scope to utilized PCA for feature extraction, this study achieved to rapidly predict the failure modes in an extremely satisfying accuracy, which presented obvious superiority compared with prevalent operation, indicating a powerful capability and vast feasibility in similar task implementation.

Optimize network intrusion detection system based on PCA feature extraction and three naïve bayes classifiers

Ransomware has become one of the major threats to private organizations, state and local organizations in the last three years, especially since the WannaCry malware. Ransomware is a child in the malware family that hijacks users’ data and related resources using security measures such as cryptographic and then demands bitcoin in compensation for the data that has been encrypted. Network Intrusion Detection Systems (NIDS) are now needed to protect data from hackers and keep information systems and networks from getting hurt as much as possible. It examines and predicts user behavior, which is then classified as an assault or expected behavior. This study will present a suggestion to improve hybrid (anomaly-misuse) NIDS by experimenting with three different Nave Bayes (NB) classifiers: Gaussian Naive Bayes (GNB), Multinomial Naive Bayes (MNB), and Complement Naive Bayes (CNB). To improve accuracy and reduce the computational time in NIDS, Principal Component Analysis (PCA) will be exploited. The UNSW-NB15 dataset was used to evaluate the suggested system’s performance, and the outcomes of the experiments are as follows: when using all features of UNSW-NB15 Datasets the Accuracy of GNB, MNB and CNB is 97%, 92%, and %94, when using PCA feature extraction with 10 PC the Accuracy of GNB, MNB and CNB is 100%, 88% and %100, when using PCA feature extraction with 20 PC the Accuracy of GNB, MNB and CNB is 100%, 92% and %100. When using PCA feature extraction with 30 PC, GNB, MNB, and CNB is 100%, 99%, and %100. According to the findings, 30 PCs improved the accuracy of the three algorithms and gave the best results, meaning that the increase in the number of PC leads to a rise in the system’s accuracy.

Flexible capped principal component analysis with applications in image recognition

Robust principal component analysis (PCA) for feature extraction, which takes various robust norms as the distance metric, has been shown to be valid in data reconstruction and recognition tasks. However, some existing approaches do not achieve satisfactory performance and robustness owing to the imperfection of norms used for measuring the distances in models. To solve this problem, we propose a new formulation of the robust PCA feature extraction method, called flexible capped PCA (FCPCA), in which we use capped L2,p-norm distance metric to minimize reconstruction errors of data. In addition to having rotational invariance, capped L2,p-norm distance metric can better weaken the role of outliers, even if the outliers’ effect is magnified when p is a large value. This characteristic makes our model more powerful than existing approaches. Also, the proposed model is more flexible than many other robust models, which can be derived by specifying the value of p. However, solving the resulting objective problem is a challenge. One of the key contributions of this study is the design of a new iterative algorithm for the derivation of an optimal solution, in which each iteration computes an eigenvalue problem associated with a weighted covariance matrix. Experimental results on several public datasets prove the algorithm’s effectiveness.

Research on physicochemical properties, microscopic characterization and detection of different freezing-damaged corn seeds

Seed freezing damage is an agricultural disaster. To explore how frostbite affects the growth and development of corn seeds, the germination conditions, and the biological indicators including the activities of related enzymes (SOD, POD, CAT, and AMS) of different frozen corn seeds (normal, -10°C,10h, and -20°C,10h) were measured. The texture of seed coat and the cell structure of seed embryo were observed by scanning electron microscope and transmission electron microscope respectively. The texture and cell structural changes reflect the influence of frostbite on corn seeds. To propose a quick, accurate and non-destructive method to identify the freezing-damaged corn seeds, near-infrared spectroscopy was used the identify the different frozen corn seeds. Different pretreatments, feature extraction methods and modeling methods were applied, result showed that in the case of standard normal variation pretreatment combined with principal component analysis feature extraction method and K-nearest neighbor model, 99.4% and 100% classification results of the training set and testing set were obtained respectively.

Performance Comparison of Feature Extraction and Machine Learning Classification Algorithms for Face Recognition

Face recognition is a highly active research topic in pattern recognition and computer vision, with numerous practical applications. Face recognition can provide the most natural interaction experience similar to the way humans can recognize others. This paper presents a performance comparison of various machine learning approaches and feature extraction algorithms. The feature extraction algorithm used is Principal Component Analysis (PCA), Latent Dirichlet Allocation (LDA), and a combination of PCA-LDA. The method used is to take a dataset sample and then evaluate and compare machine learning algorithms to analyze accuracy in recognizing faces. We also use feature extraction techniques on facial image capture to speed up data processing. The classification algorithms measured are k-nearest neighbor, naive Bayes, support vector machine, random forest, and gradient boosting. The results showed that the random forest classification algorithm was the most accurate face recognition method. On the other hand, the PCA-LDA combined feature extraction algorithm has lower false-negative and false-positive rates than PCA and LDA. In addition, the PCA feature extraction algorithm has the fastest performance in the process of recognizing faces

Design of Deep Learning Algorithm for IoT Application by Image based Recognition

The Internet of Things (IoT) is an ecosystem comprised of multiple devices and connections, a large number of users, and a massive amount of data. Deep learning is especially suited for these scenarios due to its appropriateness for "big data" difficulties and future concerns. Nonetheless, guaranteeing security and privacy has emerged as a critical challenge for IoT administration. In many recent cases, deep learning algorithms have proven to be increasingly efficient in performing security assessments for IoT devices without resorting to handcrafted rules. This research work integrates principal component analysis (PCA) for feature extraction with superior performance. Besides, the primary objective of this research work is to gather a comprehensive survey data on the types of IoT deployments, along with security and privacy challenges with good recognition rate. The deep learning method is performed through PCA feature extraction for improving the accuracy of the process. Our other primary goal in this study paper is to achieve a high recognition rate for IoT based image recognition. The CNN approach was trained and evaluated on the IoT image dataset for performance evaluation using multiple methodologies. The initial step would be to investigate the application of deep learning for IoT image acquisition. Additionally, when it comes to IoT image registering, the usefulness of the deep learning method has been evaluated for increasing the appropriateness of image recognition with good testing accuracy. The research discoveries on the application of deep learning in the Internet of Things (IoT) system are summarized in an image-based identification method that introduces a variety of appropriate criteria.

Identification of paralytic shellfish poison producing algae based on three-dimensional fluorescence spectra and quaternion principal component analysis

In view of the problem of the paralytic shellfish poison producing algae on-line measurement and identification, a new feature extraction method of paralytic shellfish poison producing algae measurement and identification based on quaternion principal component analysis (QPCA) is investigated. The three-dimensional (3D) fluorescence spectra of three common species of paralytic shellfish poison producing algae and eight species common of non paralytic shellfish poison producing algae are analyzed. The quaternion parallel representation model of algae three-dimensional fluorescence spectrum data is established, then the features of quaternion principal component is extracted to use as the input of k-nearest neighbor (KNN) classifier, and the identification of paralytic shellfish poison producing algae is realized by the three-dimensional fluorescence spectra coupled with quaternion principal component analysis. The results show that under the quaternion parallel representation model, the recognition accuracy rate of multiplication feature, modulus feature and summation feature is 90%, 95% and 100% respectively. Compared with that of the principal component analysis feature extraction method, the recognition accuracy rate in pure samples by summation feature of quaternion principal component is improved by 10%. This study provides an experimental basis for the accurate monitoring technology of three-dimensional fluorescence spectrum of paralytic shellfish poison producing algae.

Analysis of contrast limited adaptive histogram equalization (CLAHE) parameters on finger knuckle print identification

Recently, safe and reliable identification systems are strongly needed. One of them is Biometric Identification System. A Biometric characterization of Finger Knuckle Print ( FKP) that a person has is unique.A method to improve image enhancement is needed in order to produce a better and clearer image. Contrast Limited Adaptive Histogram Equalization (CLAHE) is an image enhancement method that can provide clip limit and region size. This research proves that changes in the region size and clip limit parameters can affect the accuracy of the finger knuckle print biometric identification by extracting Speeded-Up Robust Features (SURF) and PrincipalComponentAnalysis (PCA) features. The highest accuracy using SURF extraction is obtained at region size 5x5 and clip limit 0.14 which is 97%, while the highest accuracy with PCA feature extraction is obtained when the region size is 2x2 and clip limit is 0.12, 0.13, 0.14, 0.15, 0.18, 0.2, 0.4, and 0.8 which is 91%.

Reducing deep learning network structure through variable reduction methods in crop modeling

Crop models are widely used to predict plant growth, water input requirements, and yield. However, existing models are very complex and require hundreds of variables to perform accurately. Due to these shortcomings, large-scale applications of crop models are limited. In order to address these limitations, reliable crop models were developed using a deep neural network (DNN) – a new approach for predicting crop yields. In addition, the number of required input variables was reduced using three common variable selection techniques: namely Bayesian variable selection, Spearman's rank correlation, and Principal Component Analysis Feature Extraction. The reduced-variable DNN models were capable of estimating future crop yields for 10,000,000 different weather and irrigation scenarios while maintaining comparable accuracy levels to the original model that used all input variables. To establish clear superiority of the methodology, the results were also compared with a very recent feature selection algorithm called min-redundancy max-relevance (mRMR). The results of this study showed that the Bayesian variable selection was the best method for achieving the aforementioned goals. Specifically, the final Bayesian-based DNN model with a structure of 10 neurons in 5 layers performed very similarly (78.6% accuracy) to the original DNN crop model with 400 neurons in 10 layers, even though the size of the neural network was reduced by 80-fold. This effort can help promote sustainable agricultural intensifications through the large-scale application of crop models.

Feature Selection Method to Improve the Accuracy of Diabetes Mellitus Detection Instrument

The need for aroma recognition devices or often known as enose (electronic nose), is increasing. In the health field, enose can detect early diabetes mellitus (DM) type 2 from the aroma of urine. Enose is an aroma recognition tool that uses a pattern recognition algorithm to recognize the urine aroma of diabetics based on input signals from an array of gas sensors. The need for portable enose devices is increasing due to the increasing need for real-time needs. Enose devices have an enormous impact on the choice of the gas sensor Array in the enose. This article discusses the effect of the number of sensor arrays used on the recognition results. Enose uses a maximum of 4 sensors, with a maximum feature matrix. After that, the feature matrix enters the PCA (Principal Component Analysis) feature extraction and clustering using the FCM (Fuzzy C Means) method. The number of sensors indicates the number of features. Enose using method for feature selection, it’s a variation from 4 sensors, where experiment 1 uses 4 sensors, experiment 2 uses a variation of 3 sensors and experiment 3 uses a variation of 2 sensors. Especially for sensors 3 and 4 using feature extraction method, PCA (Principal Component Analysis), to reduce features to only 2 best features. As for the variation of 2 sensors use primer feature matrix. After feature selection, the number of features is 2 out of 11 variations. Next, do the grouping using the FCM (Fuzzy C Means) method. The results show that using two sensors has a high accuracy rate of 92.5%.

Hyperspectral image classification using support vector machines

In this paper, a novel approach for hyperspectral image classification technique is presented using principal component analysis (PCA), bidimensional empirical mode decomposition (BEMD) and support vector machines (SVM). In this process, using PCA feature extraction technique on Hyperspectral Dataset, the first principal component is extracted. This component is supplied as input to BEMD algorithm, which divides the component into four parts, the first three parts represents intrensic mode functions (IMF) and last part shows the residue. These BIMFs and residue image is further taken as input to the SVM for classification. The results of experiments on two popular datasets of hyperspectral remote sensing scenes represent that the proposed-model offers a competitive analyticalperformance in comparison to some established methods.

Pulsed terahertz spectroscopy combined with hybrid machine learning approaches for structural health monitoring of multilayer thermal barrier coatings.

Structural health monitoring of multilayer thermal barrier coatings (TBCs) is very vital to ensure the structural integrity and service performance of the hot-section components of the aero-engine. In this paper, we theoretically and numerically demonstrated that the terahertz time domain spectrum and the terahertz reflectance spectrum could be adopted to estimate the structure parameters, based on the finite difference time domain (FDTD) algorithm, 64 samples which were imported with three kinds of 64 sets structure parameters had been calculated to obtain the time domain and terahertz reflectance signals. To mimic the actual test signals, the original FDTD simulation signals were processed by adding the Gaussian white noise and wavelet noise reduction. To reduce the data dimension and improve the calculation efficiency during modeling, the principal component analysis (PCA) algorithm was adopted to reduce the dimensions of time-domain data and reflectance data. Finally, these data after multiple signal processing and PCA feature extraction were used to train the extreme learning machine (ELM), combining the genetic algorithm (GA) could optimize the PCA-ELM model and further improve the prediction performance of the hybrid model. Our proposed novel and efficient terahertz nondestructive technology combined with the hybrid machine learning approaches provides great potential applications on the multilayer TBCs structural integrity evaluation.

Facial Recognition Implementation using K–NN and PCA Feature Extraction in Attendance System

Attendance is the fact that someone is present at an event or goes regularly to an institution, or attendance at an event is the number of people present at that time. The Saifiatul Amaliyah school itself is one of the many schools in Indonesia where the attendance of students or attendance is still done manually. This can cause problems, namely allowing fraud when filling in attendance and errors in data recapitulation. Therefore, in this study a computerized face attendance was created, which was formed using the K-Nearest Neighbor (K-NN) method and combined with the extraction of the Principal Component Analysis (PCA) feature where the attendance process can be done with a person's face. The face attendance system using the K-NN and PCA methods has an accuracy of 82%.

Principal Component Analysis in Projection and Image Domains—Another Form of Spectral Imaging in Photon-Counting CT

We explore the feasibility of principal component analysis (PCA) as a form of spectral imaging in photon-counting CT. Using the data acquired by a prototype system and simulated by computer, we investigate the feasibility of spectral imaging in photon-counting CT via PCA for feature extraction and study the impacts made by data standardization and de-noising on its performance. The PCA in the projection domain maintains the data consistence that is essential for tomographic image reconstruction and performs virtually the same as that in the image domain. The first three primary components account for more than 99.99% covariance of the data. Within anticipation, the contrast-to-noise ratio (CNR) between the target and background in the first principal component image can be larger than that in the image generated from the data acquired in each energy bin. More importantly, the CNR in the first principal component image may be larger than that in the image formed by the summed data acquired in all energy bins (i.e., the conventional polychromatic CT image). In addition, de-noising can not only reduce the noise in images but also improve the effectiveness/efficiency of PCA in feature extraction. The PCA in either projection or image domain provides another form of spectral imaging in photon-counting CT that fits the essential requirements on spectral imaging in true color. The verification of PCA's feasibility in CT as a form spectral imaging and observation of its potential superiority in CNR over conventional polychromatic CT are meaningful in theory and practice.