Relevance Vector Machine Research Articles

Application of Big Data Unbalanced Classification Algorithm in Credit Risk Analysis of Insurance Companies

The 2008 global financial crisis triggered by subprime mortgage crisis in the United States and the ongoing European debt crisis have urged governments and academics to pay high attention to financial industry risk supervision. The financial industry has actively implemented comprehensive risk management. As an important component of the financial industry, the insurance industry implements comprehensive risk management to control the risks of insurance companies. Propose an integrated learning model based on imbalanced dataset resampling and apply it to UCI dataset (University of California Irvine). First, resampling technology is used to preprocess the unbalanced dataset to obtain a relatively balanced training set. Then, use the classic backpropagation neural network, classic k-nearest neighbor, and classic Naive Bayes three algorithms as the base classifier and use the Bagging strategy to get the ensemble learning model. In order to verify its effectiveness, F‐measure and G‐mean methods are used to measure the performance of the classifier. The subject mainly focuses on the classification of relevance vector machine (RVM) in two types of large‐scale datasets, imbalanced and balanced, and proposes solutions for these two types of data. This explains the effectiveness of the disequilibrium classification algorithm used in the risk analysis of insurance companies.

Combination Complex-Valued Bayesian Compressive Sensing Method for Sparsity Constrained Deconvolution Beamforming

Several deconvolution methods have been proposed to reduce the mainlobe width and sidelobe intensity of conventional beamforming results without increasing the array aperture; however, most of them cannot perform well in the face of coherent targets. Therefore, it is discerning to deconvolve the complex-valued beamforming result rather than the beamforming intensity. However, conventional deconvolution methods focus on the beam intensity, and only a few studies have investigated complex-valued deconvolution beamforming (CDB). Considering the sparse property of the targets in practice, the CDB can be converted to a complex-valued inverse compressive sensing problem and put into the Bayesian framework. To solve it efficiently, a complex-valued Relevance-Vector-Machine (RVM) tool is used to build the complex-valued Bayesian compressive sensing (C-BCS) method that has excellent performance in terms of recovery accuracy and operating speed. However, the C-BCS method is not stable for the CDB because the point spread function is a rank deficiency matrix. To overcome this problem, we propose the combination C-BCS method that combines the block-sparse and C-BCS methods to improve the performance of solving the CDB. The simulation results proved that the proposed method performed better than intensity-based deconvolution methods for the beamforming results generated by the coherent targets. A beamforming image of a pool scene captured by a narrowband forward-looking sonar system was adopted to test various deconvolution methods, and the proposed method exhibited superiority in sidelobe suppression.

Seismic fragility analysis of bridges by relevance vector machine based demand prediction model

Seismic fragility analysis of bridges by relevance vector machine based demand prediction model

Novel classification method of mechanical defects for GIS equipment based on mode function analysis and improved relevance vector machine

Novel classification method of mechanical defects for GIS equipment based on mode function analysis and improved relevance vector machine

Reduced-Space Relevance Vector Machine for Adaptive Electrical Capacitance Volume Tomography

We introduce an efficient synthetic electrode selection strategy for use in Adaptive Electrical Capacitance Volume Tomography (AECVT). The proposed strategy is based on the Adaptive Relevance Vector Machine (ARVM) method and allows to successively obtain synthetic electrode configurations that yield the most decrease in the image reconstruction uncertainty for the spatial distribution of the permittivity in the region of interest. The problem is first formulated as an instance of the Quadratic Unconstrained Binary Optimization (QUBO). By noting that the QUBO formulation is an NP-hard problem and thus prohibitive in practice, we then introduce the Reduced ARVM method, corresponding to the application of the ARVM method to a reduced search space. By using the Reduced ARVM method, good image reconstruction and low uncertainty levels can be achieved in AECVT with considerably fewer measurements. To corroborate our analysis, we present simulation results for three representative AECVT scenarios.

Multi-fault diagnosis for battery pack based on adaptive correlation sequence and sparse classification model

Aiming for an efficient fault diagnosis scheme for battery pack, this paper develops a complete framework for the diagnosis of faults in battery packs. First, an interclass sensor topology is introduced to cover multi-fault abnormalities, and an adaptive correlation coefficient between adjacent sensors is used to encompass system information. Then, the discrete wavelet packet transform (DWPT) is utilized to process the correlation sequences. Thereby, a variety of characteristic indicators are attained to and the main components are extracted by Principal Component Analysis (PCA) as fault features. Afterwards, two sparse classification models are developed, based on the multiclass relevance vector machine, to distinguish fault type and evaluate fault degree respectively. A fault injection platform is established to physically trigger the faults of external short, internal short, thermal abuse and loose connection on a series-connected four-cell pack. Finally, experimental verifications suggest that the proposed method gives accurate and reliable judgements on different fault types, and evaluates the fault degree accurately.

Adaptive relevance vector machine combined with Markov-chain-based importance sampling for reliability analysis

Many engineering systems involve complex implicit performance functions, and evaluating the failure probability of these systems usually requires time-consuming finite element simulations. In this research, a new reliability method is proposed by combining relevance vector machine and Markov-chain-based importance sampling (RVM-MIS), which improves computational efficiency by decreasing the number of expensive model simulations. Relevance vector machine (RVM) is a machine learning method based on the concept of probabilistic Bayesian learning framework. It is worth noting that RVM provides predicted value of the sample and corresponding variance. Due to this important feature, various active learning functions can be applied to improve the accuracy of RVM to approximate real performance functions. In addition, Markov-chain-based importance sampling (MIS) is utilized to generate important samples covering areas that significantly contribute to failure probability. The important samples are then predicted by a well-constructed RVM to obtain failure probability, rather than being evaluated using real performance functions, so the computation time is drastically decreased. RVM-MIS reduces the number of calls to real performance function while ensuring the accuracy of results. Four academic examples and a bearing statics problem with an implicit performance function are performed to verify the accuracy and efficiency of the proposed method.

Bayesian prediction of tunnel convergence combining empirical model and relevance vector machine

Convergence monitoring of the tunnel is a direct and reliable way to reveal its status. Accurate prediction of convergence is essential to prevent safety hazards, such as rock collapse, project delay, and even geological disasters. Convergence prediction is usually carried out based on numerical methods (NMs) and empirical models (EMs). The accurate model parameters of NMs are difficult to estimate from limited field geological tests, which severely undermines its prediction accuracy. Although the EMs involve the advantages of computational simplicity and efficiency, their prediction capacity is relatively limited. This paper presents a probabilistic model that combines the EM, Bayesian estimation, and relevance vector machine (RVM) to predict convergence. Firstly, various EMs integrated with Bayesian estimation are established and prediction results are compared to select the EM with higher prediction accuracy. Prediction residuals of the EM are then modeled by the RVM to further improve the accuracy. A high-speed railway tunnel is utilized to demonstrate the effectiveness of the presented approach. The results show that the root-mean-squared error values of the combined probabilistic model are reduced by 92.6% and 95.8% compared with the EM for two data sets. Moreover, the comparison results show that the presented model exhibits higher prediction accuracy than backpropagation neural network and Gaussian process regression.

Robust equation discovery considering model discrepancy: A sparse Bayesian and Gaussian process approach

Computational models are widely used to describe engineering systems and to predict their behaviour. However, in many applications, these computational models do not capture the complete physics of the real system, leading to model discrepancy. Model discrepancy causes bias in the inferred model parameters when it is not properly accounted for. This paper proposes a novel approach that seeks to capture the functional form of the model discrepancy and reduce bias in the estimated model parameters through and equation discovery procedure. A sparse Bayesian model is proposed, where sparsity is introduced through a hierarchical prior structure, providing a mechanism for removing erroneous candidate model terms from a series of potential equations as a part of an equation discovery procedure. At the same time, a Gaussian process model is used to account for model discrepancy. These two modelling assumptions are combined in a Bayesian formulation, allowing the system parameters and model discrepancy to be inferred in a probabilistic manner with associated uncertainties quantified based on their posterior distributions. The resulting method is capable of simultaneously providing physical insights into the system behaviour, by selecting the appropriate candidate model components and their respective system parameters, whilst compensating for model discrepancy that may occur due to an incomplete set of candidate terms. In order to efficiently solve the statistical model, an expectation maximisation algorithm is proposed for performing inference, and illustrative examples are presented to validate the proposed method. It is shown that compared to using a conventional sparse Bayesian approach for performing equation discovery, such as the Relevance Vector Machine, the proposed method provides better equation selection and parameter estimation, with less bias in the parameter estimates.

A fault diagnostic method for oil-immersed transformer based on multiple probabilistic output algorithms and improved DS evidence theory

Power transformer is one of the most vital equipment in the electric power systems, which serves as the connection of power networks at different voltage levels. However, the power transformers in service are subjected to various stresses like thermal stress and electrical stress, which may lead to several faults such as arching, partial discharge and overheating. In order to detect these incipient faults and ensure the normal operation of power transformers, the diagnostic methods based on Dissolved Gas Analysis (DGA) have been developed. However, the DGA methods suffer from low diagnostic accuracy. On the other hand, with the development of Artificial Intelligence (AI), a lot of intelligent algorithms have been applied in this area. However, most traditional AI algorithms are hard classification, which may cause the wrong diagnosis of faults. To overcome those disadvantages, a method based on the improved DS evidence theory and multiple probabilistic output algorithms is proposed in this paper to improve the performance of fault diagnosis of oil-immersed transformers. Firstly, the performance of the soft and hard classification algorithms is compared and analyzed. After that, three diagnostic models based on the Multiclass Relevance Vector Machine (MRVM), Multiclass Support Vector Machine (MSVM) and Back Propagation Neural Network (BPNN) are established. In order to improve the accuracy of these models, Particle Swarm Optimization (PSO) is used to optimize their hyper-parameters. Finally, the combination of these three probabilistic output models is achieved based on the improved Dempster-Shafer (DS) evidence theory, which can help to obtain a more accurate fault diagnosis result by fusing multiple algorithm results. The case study results show that the proposed method achieves a probabilistic output for the fault diagnosis of oil-immersed transformers and overcomes the deficiency of traditional DGA methods which are difficult to get accurate diagnosis results and can’t summarize the fault development rule inductively.

Hybrid ensemble soft computing approach for predicting penetration rate of tunnel boring machine in a rock environment

This study implements a hybrid ensemble machine learning method for forecasting the rate of penetration (ROP) of tunnel boring machine (TBM), which is becoming a prerequisite for reliable cost assessment and project scheduling in tunnelling and underground projects in a rock environment. For this purpose, a sum of 185 datasets was collected from the literature and used to predict the ROP of TBM. Initially, the main dataset was utilised to construct and validate four conventional soft computing (CSC) models, i.e. minimax probability machine regression, relevance vector machine, extreme learning machine, and functional network. Consequently, the estimated outputs of CSC models were united and trained using an artificial neural network (ANN) to construct a hybrid ensemble model (HENSM). The outcomes of the proposed HENSM are superior to other CSC models employed in this study. Based on the experimental results (training RMSE = 0.0283 and testing RMSE = 0.0418), the newly proposed HENSM is potential to assist engineers in predicting ROP of TBM in the design phase of tunnelling and underground projects. • Amalgamation of four conventional soft computing models and ANN. • Construction of hybrid ensemble model using MPMR, RVM, ELM and FN. • Implementation of accuracy matrix for visual interpretation of results. • Determination of best performing model using uncertainty analysis.

Prediction of Geotechnical Properties of Soil using Artificial Intelligence Framework

The present research work is carried out to predict the geotechnical properties (consistency limits, OMC, and MDD) of soil using AI technologies, namely regression analysis (RA), support vector machine (SVM), Gaussian process regression (GPR), artificial neural networks (ANNs), and relevance vector machine (RVM). The models of machine learning (SVM, GPR), hybrid learning (RVM), and deep learning (ANNs) are constructed in MATLAB R2020a with different configurations. The models of RA are built using the Data Analysis Tool of Microsoft Excel 2019. The input parameters of AI models are gravel, sand, silt, and clay content. The correlation coefficient is calculated for pair of soil datasets. The correlation shows that sand, silt, and clay content play a vital role in predicting soil's liquid limit and plasticity index. The performance of constructed AI models is compared to determine the optimum performance models. The limited datasets of soil are used in this study. Therefore, artificial neural networks and relevance vector machines could not perform well. Based on the performance of AI models, the Gaussian process regression outperformed the RA, SVM, ANNs, and RVM AI technologies. Hence, the GPR AI approach can predict the geotechnical properties of soil by gravel, sand, silt, and clay content. The Monte-Carlo global sensitivity analysis is also performed, and it is observed that the prediction of geotechnical properties of soil is affected by sand and clay content

Daily PM2.5 and PM10 forecasting using linear and nonlinear modeling framework based on robust local mean decomposition and moving window ensemble strategy

Highly accurate forecasting of particulate matter concentration (PMC) is essential and effective for establishing a reliable air pollution early warning system and has both theoretical and practical significance. To meet this demand, a novel multi-scale hybrid learning framework based on robust local mean decomposition (RLMD) and moving window (MW) ensemble strategy is developed for PM2.5 and PM10 forecasting. In this architecture, the RLMD is adopted to adaptively decompose the PMC time series (PMCTS) into several production functions and one residue with different frequencies. These subseries are simpler than the original PMCTS, but they still work alongside mode aliasing. Thus, following the well-established “linear and nonlinear” modeling philosophy, a novel hybrid learning framework, composed of the autoregressive integrated moving average (ARIMA) and combined kernel function relevance vector machine (RVMcom), is proposed to capture both the linear and nonlinear patterns in the subseries. To obtain better final outputs, based on the definition of the ensemble improvement degree, the MW ensemble method is used to merge the forecasting results of all subseries. A comprehensive experiment is conducted using PM2.5 and PM10 datasets from four municipalities in China to investigate the forecasting performance of our proposed framework, and the results demonstrate that our proposed RLMD-ARIMA–RVMcom-MW (R-A&Rcom-M) model is superior to other considered methods in terms of forecasting accuracy and generalization ability. This means that the developed forecasting architecture has a great application value in the field of PMCTS prediction.

Beamspace channel estimation in wideband lens antenna array-based mmWave mMIMO-OFDM systems under beam squint

Sparse Bayesian learning (SBL) and particularly relevant vector machines (RVMs) have drawn much attention to improving the performance of existing machine learning models. The methodology depends on a parameterized prior that enforces models with weight sparsity, where only a few are non-zeros. Wideband mmWave massive multiple-input multiple-output (mMIMO) systems with lens antenna array (LAA), expect to play a key role in future fifth-generation (5G) wireless systems. To provide the beamforming gain required to overcome path loss, we consider a lens antenna array (LAA)-based beamspace mMIMO system. However, the spatial-wideband influence causes the beam squint effect to emerge, making the beamspace channel path components exhibit a unique frequency-dependent sparse structure, and thus nullifies the frequency domain common support assumption. In this paper, we first propose a channel estimation (CE) algorithm, namely a reduced-antenna selection progressive support-detection (RAS-PSD), for the wideband mmWave mMIMO-OFDM systems with LAA, which considers the beam squint effect. Secondly, by exploring Bayesian learning (BL), a Gaussian Process hyperparameter optimization-based CE (GP-HOCE) algorithm is proposed for the considered system, where both its own hyperparameter and the hyperparameters of its adjacent neighbors governs the sparsity of each coefficient. The simulation results show that the wideband beamspace channel coefficients can be estimated more efficiently than those of the existing state-of-the-art algorithms in terms of normalized mean square error (NMSE) of CE for wideband mmWave mMIMO-OFDM systems.

Tropospheric Refractivity Profile Estimation by GNSS Measurement at China Big-Triangle Points

Atmospheric radio refractivity has an obvious influence on the signal transmission path and communication group delay effect. The uncertainty of water vapor distribution is the main reason for the large error of tropospheric refractive index modeling. According to the distribution and characteristics of water vapor pressure, temperature, and pressure, which are the basic components of the refractive index, a method for retrieving atmospheric refractivity profile based on GNSS (Global Navigation Satellite System) and meteorological sensor measurement is introduced and investigated in this study. The variation of the correlation between zenith wet delay and water vapor pressure is investigated and analyzed in detail. The partial pressure profiles of water vapor are retrieved with relevance vector machine method based on tropospheric zenith wet delay calculated by single ground-based GPS (Global Positioning System) receiver. The atmospheric temperature and pressure is calculated with the least square method, which is used to fit the coefficients of the polynomial model based on a large number of historical meteorological radiosonde data of local stations. By combining the water vapor pressure profile retrieving from single ground-based GPS and temperature and pressure profile from reference model, the refractivity profile can be obtained, which is compared to radiosonde measurements. The comparison results show that results of the proposed method are consistent with the results of radiosonde. By using over ten years’ (through 2008 to 2017) historical radiosonde meteorological data of different months at China Big-Triangle Points, i.e., Qingdao, Sanya, Kashi, and Jiamusi radiosonde stations, tropospheric radio refractivity profiles are retrieved and modeled. The comparison results present that the accuracies of refractivity profile of the proposed method at Qingdao, Sanya, Kashi, and Jiamusi are about 5.48, 5.63, 3.58, and 3.78 N-unit, respectively, and the annual average relative RMSE of refractivity at these stations are about 1.66, 1.53, 1.49, and 1.23%, respectively.

PERFORMANCE EVALUATION OF DATA-DRIVEN PROGNOSTIC BASED ON RVM-SBI TECHNIQUE

The Prognostic and Health Management (PHM) becomes a research topic in its own right and tends to be more and more visible within the scientific community such as in Nasa Society, which has provided datasets for experiments. The purpose of this paper is to evaluate the performance of a data-driven prognostic technique used for predicting Remaining Useful Life (RUL). The methodological support of the proposed approach integrates all data-driven prognostic sequential steps merged in offline and online part. To design the predictive degradation model on the offline part, the Relevance Vector Machine (RVM) algorithm was applied. On the online part, prediction of the RUL is based on the Similarity-Based Interpolation (SBI) algorithm. The different steps of the methodology are described and their implementation undertaken through a case study involving the degradation dataset of turbofan engines from the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS). Finally, results are compared with other techniques applied on the same dataset.

Computation of Permeability of Soil using Artificial Intelligence Approaches

The Gaussian Process Regression (GPR), Decision Tree (DT), Relevance Vector Machine (RVM), and Artificial Neural Network (ANN) AI approaches are constructed in MATLAB R2020a with different hyperparameters namely, kernel function, leaf size, backpropagation algorithms, number of neurons and hidden layers to compute the permeability of soil. The present study is carried out using 158 datasets of soil. The soil dataset consists of fine content (FC), sand content (SC), liquid limit (LL), specific gravity (SG), plasticity index (PI), maximum dry density (MDD) and optimum moisture content (OMC), permeability (K). Excluding the permeability of soil, rest of properties of soil is used as input parameters of the AI models. The best architectural and optimum performance models are identified by comparing the performance of the models. Based on the performance of the AI models, the NISEK_K_GPR, 10LF_K_DT, Poly_K_RVM, and GDANN_K_10H5 models have been identified as the best architectural AI models. The comparison of performance of the best architectural models, it is observed that the NISEK_K_GPR model outperformed the other best architectural AI models. In this study, it is also observed that GPR model is outperformed ANN models because of small dataset. The performance of NISEK_K_GPR model is compared with models available in literature and it is concluded that the GPR model has better performance and least prediction error than models available in literature study.

Detecting Symptom Errors in Neural Machine Translation of Patient Health Information on Depressive Disorders: Developing Interpretable Bayesian Machine Learning Classifiers.

Background: Due to its convenience, wide availability, low usage cost, neural machine translation (NMT) has increasing applications in diverse clinical settings and web-based self-diagnosis of diseases. Given the developing nature of NMT tools, this can pose safety risks to multicultural communities with limited bilingual skills, low education, and low health literacy. Research is needed to scrutinise the reliability, credibility, usability of automatically translated patient health information.Objective: We aimed to develop high-performing Bayesian machine learning classifiers to assist clinical professionals and healthcare workers in assessing the quality and usability of NMT on depressive disorders. The tool did not require any prior knowledge from frontline health and medical professionals of the target language used by patients.Methods: We used Relevance Vector Machine (RVM) to increase generalisability and clinical interpretability of classifiers. It is a typical sparse Bayesian classifier less prone to overfitting with small training datasets. We optimised RVM by leveraging automatic recursive feature elimination and expert feature refinement from the perspective of health linguistics. We evaluated the diagnostic utility of the Bayesian classifier under different probability cut-offs in terms of sensitivity, specificity, positive and negative likelihood ratios against clinical thresholds for diagnostic tests. Finally, we illustrated interpretation of RVM tool in clinic using Bayes' nomogram.Results: After automatic and expert-based feature optimisation, the best-performing RVM classifier (RVM_DUFS12) gained the highest AUC (0.8872) among 52 competing models with distinct optimised, normalised features sets. It also had statistically higher sensitivity and specificity compared to other models. We evaluated the diagnostic utility of the best-performing model using Bayes' nomogram: it had a positive likelihood ratio (LR+) of 4.62 (95% C.I.: 2.53, 8.43), and the associated posterior probability (odds) was 83% (5.0) (95% C.I.: 73%, 90%), meaning that approximately 10 in 12 English texts with positive test are likely to contain information that would cause clinically significant conceptual errors if translated by Google; it had a negative likelihood ratio (LR-) of 0.18 (95% C.I.: 0.10,0.35) and associated posterior probability (odds) was 16% (0.2) (95% C.I: 10%, 27%), meaning that about 10 in 12 English texts with negative test can be safely translated using Google.

Remaining useful life prediction with probability distribution for lithium-ion batteries based on edge and cloud collaborative computation

This paper proposes the architecture of the combination of the battery management system (BMS) and the cloud big data platform. Firstly, BMS measures and extracts the mean voltage falloff (MVF). A regression model of capacity and MVF based on historical data is established with generalized Box-Cox Transformation and least squares. The capacity and MVF are uploaded to the cloud big data platform, and then the mean and variance of the MVF is predicted based on the relevance vector machine, thereby realizing the 2σ range prediction of the lithium battery's state of health and the probability density function prediction of the remaining useful life. This paper makes two contributions to the data-driven prediction method. First, the edge-cloud collaborative computing architecture combining BMS and cloud is proposed, which effectively utilizes the advantages of BMS data quality and cloud computing power. Second, through the combination of relevance vector machine with particle swarm optimization and horizontal parameter transfer, the number of samples required for model learning is reduced to 30% and has better accuracy and robustness. Through the verification of NASA data, the results show that the average error is less than 2.18%.

Hybrid Model Based on Maxwell-Slip Model and Relevance Vector Machine

This article proposes a hybrid model based on Maxwell-slip (MS) model and relevance vector machine (RVM) for piezoelectric actuators (PEAs). The MS model maps the input voltage into the internal state space and provides an estimation of the output displacement. The output function of the MS model provides a linear relationship between the deformation vector and the output displacement. Replacing the regression function of the RVM by the output function of the MS model enables the RVM to update parameters of the MS model. The RVM is chosen since it is also able to provide the distribution of the estimated output displacement. Integration of the MS model with the RVM also reduces the dimension of the RVM from the number of support data points to the number of MS elements, which will reduce the computational time. Results demonstrate that the hybrid model is capable of describing rate-dependent and asymmetric hysteresis, is robust to excitations, enhances the accuracy of the MS model, and requires much shorter computational time than the standalone RVM. The modeling error under random excitation is 0.54% and for asymmetric hysteresis in a reluctance actuator is 0.31%. Compensation for hysteresis is carried out by connecting the inverse model in serial with a PEA and achieves tracking error of less than 0.81%.