A sparse algorithm for adaptive pruning least square support vector regression machine based on global representative point ranking

Least square support vector regression(LSSVR)is a method for function approximation,whose solutions are typically non-sparse,which limits its application especially in some occasions of fast prediction.In this paper,a sparse algorithm for adaptive pruning LSSVR algorithm based on global representative point ranking(GRPR-AP-LSSVR)is proposed.At first,the global representative point ranking(GRPR)algorithm is given,and relevant data analysis experiment is implemented which depicts the importance ranking of data points.Furthermore,the pruning strategy of removing two samples in the decremental learning procedure is designed to accelerate the training speed and ensure the sparsity.The removed data points are utilized to test the temporary learning model which ensures the regression accuracy.Finally,the proposed algorithm is verified on artificial datasets and UCI regression datasets,and experimental results indicate that,compared with several benchmark algorithms,the GRPR-AP-LSSVR algorithm has excellent sparsity and prediction speed without impairing the generalization performance.

Future changes in rainfall, temperature and reference evapotranspiration in the central India by least square support vector machine

Climate change affects the environment and natural resources immensely. Rainfall, temperature and evapotranspiration are major parameters of climate affecting changes in the environment. Evapotrans- piration plays a key role in crop production and water balance of a region, one of the major parameters affected by climate change. The reference evapotranspiration or ETo is a calculated parameter used in this research. In the present study, changes in the future rainfall, minimum and maximum temperature, and ETo have been shown by downscaling the HadCM3 （Hadley Centre Coupled Model version 3） model data. The selected study area is located in a part of the Narmada river basin area in Madhya Pradesh in central India. The downscaled outputs of projected rainfall, ETo and temperatures have been shown for the 21st century with the HADCM3 data of A2 scenario by the Least Square Support Vector Machine （LS-SVM） model. The efficiency of the LS-SVM model was measured by different statistical methods. The selected predictors show considerable correlation with the rainfall and temperature and the application of this model has been done in a basin area which is an agriculture based region and is sensitive to the change of rainfall and temperature. Results showed an increase in the future rainfall, temperatures and ETo. The temperature increase is projected in the high rise of minimum temperature in winter time and the highest increase in maximum temperature is projected in the pre-monsoon season or from March to May. Highest increase is projected in the 2080s in 2081-2091 and 2091-2099 in maximum temperature and 2091-2099 in minimum temperature in all the stations. Winter maximum temperature has been observed to have increased in the future. High rainfall is also observed with higher ETo in some decades. Two peaks of the increase are observed in ETo in the April-May and in the October. Variation in these parameters due to climate change might have an impact on the future water resource of the study area, which is mainly an agricultural based region, and will help in proper planning and management.

Nonlinear correction of photoelectric displacement sensor based on least square support vector machine

A model of correcting the nonlinear error of photoelectric displacement sensor was established based on the least square support vector machine.The parameters of the correcting nonlinear model,such as penalty factor and kernel parameter,were optimized by chaos genetic algorithm.And the nonlinear correction of photoelectric displacement sensor based on least square support vector machine was applied.The application results reveal that error of photoelectric displacement sensor is less than 1.5%,which is rather satisfactory for nonlinear correction of photoelectric displacement sensor.

Application of particle swarm optimization-based least square support vector machine in quantitative analysis of extraction solution of yangxinshi tablet using near infrared spectroscopy

A particle swarm optimization(PSO)-based least square support vector machine(LS-SVM)method was investigated for quantitative analysis of extraction solution of Y angxinshi tablet using near infrared(NIR)spectroscopy.The usable spectral region(5400-6200cm^(-1))was identified,then the first derivative spectra smoothed using a Savitzky-Golay filter were employed to establish calibration models.The PSO algorithm was applied to select the LS-SVM hyper-parameters(including the regularization and kernel parametens).The calibration models of total flavonoids,puerarin,salvianolic acid B and icarin were established using the optimumn hyper-parameters of LS SVM.The performance of LS SVM models were compared with partial least squares(PLS)regression,feed forward back propagation network(BPANN)and support vector machine(SVM).Experimental results showed that both the calibration results and prediction accuracy of the PSO-based LS SVM method were superior to PLS,BP-ANN and SVM.For PSO-based LS-SVM models,the determination cofficients(R2)for the calibration set were above 0.9881,and the RSEP values were controlled within 5.772%.For the validation set,the RMSEP values were close to RMSEC and less than 0.042,the RSEP values were under 8.778%,which were much lower than the PLS,BP-ANN and SVM models.The PSO-based LS SVM algorithm employed in this study exhibited excellent calibration performance and prediction accuracy,which has definite practice significance and application value.

Least Squares One-Class Support Tensor Machine

One-class classification problem has become a popular problem in many fields, with a wide range of applications in anomaly detection, fault diagnosis, and face recognition. We investigate the one-class classification problem for second-order tensor data. Traditional vector-based one-class classification methods such as one-class support vector machine (OCSVM) and least squares one-class support vector machine (LSOCSVM) have limitations when tensor is used as input data, so we propose a new tensor one-class classification method, LSOCSTM, which directly uses tensor as input data. On one hand, using tensor as input data not only enables to classify tensor data, but also for vector data, classifying it after high dimensionalizing it into tensor still improves the classification accuracy and overcomes the over-fitting problem. On the other hand, different from one-class support tensor machine (OCSTM), we use squared loss instead of the original loss function so that we solve a series of linear equations instead of quadratic programming problems. Therefore, we use the distance to the hyperplane as a metric for classification, and the proposed method is more accurate and faster compared to existing methods. The experimental results show the high efficiency of the proposed method compared with several state-of-the-art methods.

Image Denoising Using Local Adaptive Least Squares Support Vector Regression

Rather than attempting to separate signal from noise in the spatial domain, it is often advantageous to work in a transform domain. Building on previous work, a novel denoising method based on local adaptive least squares support vector regression is proposed. Investigation on real images contaminated by Gaussian noise has demonstrated that the proposed method can achieve an acceptable trade off between the noise removal and smoothing of the edges and details.

Least-Square Support Vector Machine and Wavelet Selection for Hearing Loss Identification

Hearing loss(HL)is a kind of common illness,which can significantly reduce the quality of life.For example,HL often results in mishearing,misunderstanding,and communication problems.Therefore,it is necessary to provide early diagnosis and timely treatment for HL.This study investigated the advantages and disadvantages of three classical machine learning methods:multilayer perceptron(MLP),support vector machine(SVM),and least-square support vector machine(LS-SVM)approach andmade a further optimization of the LS-SVM model via wavelet entropy.The investigation illustrated that themultilayer perceptron is a shallowneural network,while the least square support vector machine uses hinge loss function and least-square optimizationmethod.Besides,a wavelet selection method was proposed,and we found db4 can achieve the best results.The experiments showed that the LS-SVM method can identify the hearing loss disease with an overall accuracy of three classes as 84.89±1.77,which is superior to SVM andMLP.The results show that the least-square support vector machine is effective in hearing loss identification.

A predictive model of chemical flooding for enhanced oil recovery purposes:Application of least square support vector machine

Applying chemical flooding in petroleum reservoirs turns into interesting subject of the recent researches.Developing strategies of the aforementioned method are more robust and precise when they consider both economical point of views(net present value(NPV))and technical point of views(recovery factor(RF)).In the present study huge attempts are made to propose predictive model for specifying efficiency of chemical flooding in oil reservoirs.To gain this goal,the new type of support vector machine method which evolved by Suykens and Vandewalle was employed.Also,high precise chemical flooding data banks reported in previous works were employed to test and validate the proposed vector machine model.According to the mean square error(MSE),correlation coefficient and average absolute relative deviation,the suggested LSSVM model has acceptable reliability;integrity and robustness.Thus,the proposed intelligent based model can be considered as an alternative model to monitor the efficiency of chemical flooding in oil reservoir when the required experimental data are not available or accessible.

An Approximate Linear Solver in Least Square Support Vector Machine Using Randomized Singular Value Decomposition

In this paper, we investigate the linear solver in least square support vector machine（LSSVM） for large-scale data regression. The traditional methods using the direct solvers are costly. We know that the linear equations should be solved repeatedly for choosing appropriate parameters in LSSVM, so the key for speeding up LSSVM is to improve the method of solving the linear equations. We approximate large-scale kernel matrices and get the approximate solution of linear equations by using randomized singular value decomposition（randomized SVD）. Some data sets coming from University of California Irvine machine learning repository are used to perform the experiments. We find LSSVM based on randomized SVD is more accurate and less time-consuming in the case of large number of variables than the method based on Nystrom method or Lanczos process.

Tribological properties and wear prediction model of TiC particles reinforced Ni-base alloy composite coatings

TiC particles reinforced Ni-based alloy composite coatings were prepared on 7005 aluminum alloy by plasma spray. The effects of load, speed and temperature on the tribological behavior and mechanisms of the composite coatings under dry friction were researched. The wear prediction model of the composite coatings was established based on the least square support vector machine (LS-SVM). The results show that the composite coatings exhibit smaller friction coefficients and wear losses than the Ni-based alloy coatings under different friction conditions. The predicting time of the LS-SVM model is only 12.93%of that of the BP-ANN model, and the predicting accuracies on friction coefficients and wear losses of the former are increased by 58.74%and 41.87%compared with the latter. The LS-SVM model can effectively predict the tribological behavior of the TiCP/Ni-base alloy composite coatings under dry friction.

Intelligent control for large-scale variable speed variable pitch wind turbines

Large-scale wind turbine generator systems have strong nonlinear multivariable characteristics with many uncertain factors and disturbances. Automatic control is crucial for the efficiency and reliability of wind turbines. On the basis of simplified and proper model of variable speed variable pitch wind turbines, the effective wind speed is estimated using extended Kaiman filter. Intelligent control schemes proposed in the paper include two loops which operate in synchronism with each other. At below-rated wind speed, the inner loop adopts adaptive fuzzy control based on variable universe for generator torque regulation to realize maximum wind energy capture. At above-rated wind speed, a controller based on least square support vector machine is proposed to adjust pitch angle and keep rated output power. The simulation shows the effectiveness of the intelligent control.

Application of multi-outputs LSSVR by PSO to the aero-engine model

Considering the modeling errors of on-board self-tuning model in the fault diagnosis of aero-engine, a new mechanism for compensating the model outputs is proposed. A discrete series predictor based on multi-outputs least square support vector regression （LSSVR） is applied to the compensation of on-board self-tuning model of aero-engine, and particle swarm optimization （PSO） is used to the kernels selection of multi-outputs LSSVR. The method need not reconstruct the model of aero-engine because of the differences in the individuals of the same type engines and engine degradation after use. The concrete steps for the application of the method are given, and the simulation results show the effectiveness of the algorithm.

Regressive approach for predicting bearing capacity of bored piles from cone penetration test data

In this study, th e least sq u are su p p o rt v ecto r m achine （LSSVM） alg o rith m w as applied to predicting th ebearing capacity o f b ored piles e m b ed d ed in sand an d m ixed soils. Pile g eo m etry an d cone p e n e tra tio nte s t （CPT） resu lts w ere used as in p u t variables for pred ictio n o f pile bearin g capacity. The d ata u se d w erecollected from th e existing litera tu re an d consisted o f 50 case records. The application o f LSSVM w ascarried o u t by dividing th e d ata into th re e se ts： a train in g se t for learning th e pro b lem an d obtain in g arelationship b e tw e e n in p u t variables an d pile bearin g capacity, and testin g an d validation sets forevaluation o f th e predictive an d g en eralization ability o f th e o b tain ed relationship. The predictions o f pilebearing capacity by LSSVM w ere evaluated by com paring w ith ex p erim en tal d ata an d w ith th o se bytrad itio n al CPT-based m eth o d s and th e gene ex pression pro g ram m in g （GEP） m odel. It w as found th a t th eLSSVM perform s w ell w ith coefficient o f d eterm in atio n , m ean, an d sta n d ard dev iatio n equivalent to 0.99,1.03, an d 0.08, respectively, for th e testin g set, an d 1, 1.04, an d 0.11, respectively, for th e v alidation set. Thelow values o f th e calculated m ean squared e rro r an d m ean ab so lu te e rro r indicated th a t th e LSSVM w asaccurate in p redicting th e pile bearing capacity. The results o f com parison also show ed th a t th e p roposedalg o rith m p red icted th e pile bearin g capacity m ore accurately th a n th e trad itio n al m eth o d s including th eGEP m odel.

An optimal method for prediction and adjustment on gasholder level and self-provided power plant gas supply in steel works

An optimal method for prediction and adjustment on byproduct gasholder level and self-provided power plant gas supply was proposed.This work raises the HP-ENN-LSSVM model based on the Hodrick-Prescott filter,Elman neural network and least squares support vector machines.Then,according to the prediction,the optimal adjustment process came up by a novel reasoning method to sustain the gasholder within safety zone and the self-provided power plant boilers in economic operation,and prevent unfavorable byproduct gas emission and equipment trip as well.The experiments using the practical production data show that the proposed method achieves high accurate predictions and the optimal byproduct gas distribution,which provides a remarkable guidance for reasonable scheduling of byproduct gas.

Model Parameter Transfer for Gear Fault Diagnosis under Varying Working Conditions

Gear fault diagnosis technologies have received rapid development and been effectively implemented in many engineering applications.However,the various working conditions would degrade the diagnostic performance and make gear fault diagnosis(GFD)more and more challenging.In this paper,a novel model parameter transfer(NMPT)is proposed to boost the performance of GFD under varying working conditions.Based on the previous transfer strategy that controls empirical risk of source domain,this method further integrates the superiorities of multi-task learning with the idea of transfer learning(TL)to acquire transferable knowledge by minimizing the discrepancies of separating hyperplanes between one specific working condition(target domain)and another(source domain),and then transferring both commonality and specialty parameters over tasks to make use of source domain samples to assist target GFD task when sufficient labeled samples from target domain are unavailable.For NMPT implementation,insufficient target domain features and abundant source domain features with supervised information are fed into NMPT model to train a robust classifier for target GFD task.Related experiments prove that NMPT is expected to be a valuable technology to boost practical GFD performance under various working conditions.The proposed methods provides a transfer learning-based framework to handle the problem of insufficient training samples in target task caused by variable operation conditions.

Determination of rock depth using artificial intelligence techniques

This article adopts three artificial intelligence techniques, Gaussian Process Regression（GPR）, Least Square Support Vector Machine（LSSVM） and Extreme Learning Machine（ELM）, for prediction of rock depth（d） at any point in Chennai. GPR, ELM and LSSVM have been used as regression techniques.Latitude and longitude are also adopted as inputs of the GPR, ELM and LSSVM models. The performance of the ELM, GPR and LSSVM models has been compared. The developed ELM, GPR and LSSVM models produce spatial variability of rock depth and offer robust models for the prediction of rock depth.

A pre-warning system of abnormal energy consumption in lead smelting based on LSSVR-RP-CI

The pre-warning of abnormal energy consumption is important for energy conservation of industrial engineering. However, related studies on the lead smelting industries which usually have a huge energy consumption are rarely reported. Therefore, a pre-warning system was established in this study based on the intelligent prediction of energy consumption and the identification of abnormal energy consumption. A least square support vector regression (LSSVR) model optimized by the adaptive genetic algorithm was developed to predict the energy consumption in the process of lead smelting. A recurrence plots (RP) analysis and a confidence intervals (CI) analysis were conducted to quantitatively confirm the stationary degree of energy consumption and the normal range of energy consumption, respectively, to realize the identification of abnormal energy consumption. It is found the prediction accuracy of LSSVR model can exceed 90% based on the comparison between the actual and predicted data. The energy consumption is considered to be non-stationary if the correlation coefficient between the time series of periodicity and energy consumption is larger than that between the time series of periodicity and Lorenz. Additionally, the lower limit and upper limit of normal energy consumption are obtained.

Combined forecast method of HMM and LS-SVM about electronic equipment state based on MAGA

For the deficiency that the traditional single forecast methods could not forecast electronic equipment states, a combined forecast method based on the hidden Markov model（HMM） and least square support vector machine（LS-SVM） is presented. The multi-agent genetic algorithm（MAGA） is used to estimate parameters of HMM to overcome the problem that the Baum-Welch algorithm is easy to fall into local optimal solution. The state condition probability is introduced into the HMM modeling process to reduce the effect of uncertain factors. MAGA is used to estimate parameters of LS-SVM. Moreover, pruning algorithms are used to estimate parameters to get the sparse approximation of LS-SVM so as to increase the ranging performance. On the basis of these, the combined forecast model of electronic equipment states is established. The example results show the superiority of the combined forecast model in terms of forecast precision,calculation speed and stability.

Improved Model of Radial Vibration in Switched Reluctance Motor Including Magnetic Saturation

This paper proposes an improved method for the prediction of radial vibration in switched reluctance motor(SRM)considering magnetic saturation.In this paper,the basic modeling principle is briefly introduced,it is based on the derivation that the peak acceleration is dependent on the product of phase current and current gradient idi/dt.However,the derivation may cause errors due to saturation effect.Thus in this paper,the discrete sample data are firstly acquired based on DC pulse measurement method,by which electromagnetic,torque and peak acceleration characteristics can all be acquired.Then the entire peak acceleration characteristics are obtained by improved Least Square Support Vector Machine(LSSVM).Based on the obtained static peak acceleration characteristics,the time-varied radial vibration model is established based on superposition of natural oscillations of dominant vibration modes.Finally,a simulation model is built up using MATLAB/Simulink.The good agreement between simulation and experiment shows that the proposed method for modeling is feasible and accurate,even under saturation.In addition,since LSSVM does not need any prior knowledge,it is much easier for modeling compared with other existing literatures.

Soft measurement model on torque of alternating current electrical dynamometer including copper loss and iron loss

Alternating current electrical dynamometer is a common device to measure the torque of engines, such as the gasoline engine. In order to solve the problems such as high cost, high energy consumption and complicated measurement system which exists in the direct measurement on the torque of alternating current electrical dynamometer, copper loss and iron loss are taken as two key factors and a soft-sensing model on the torque of alternating current electrical dynamometer is established using the fuzzy least square support vector machine (FLS-SVM). Then, the FLS-SVM parameters such as penalty factor and kernel parameter are optimized by adaptive genetic algorithm, torque soft-sensing is investigated in the alternating current electrical dynamometer, as well as the energy feedback efficiency and energy consumption during the measurement phase of a gasoline engine loading continual test is obtained. The results show that the minimum soft-sensing error of torque is about 0.0018, and it fluctuates within a range from -0.3 to 0.3 N·m. FLS-SVM soft-sensing method can increase by 1.6% power generation feedback compared with direct measurement, and it can save 500 kJ fuel consumption in the gasoline engine loading continual test. Therefore, the estimation accuracy of the soft measurement model on the torque of alternating current electrical dynamometer including copper loss and iron loss is high and this indirect measurement method can be feasible to reduce production cost of the alternating current electrical dynamometer and energy consumption during the torque measurement phase of a gasoline engine, replacing the direct method of torque measurement.