Parameters Of Support Vector Regression Research Articles

Minute-ahead stock price forecasting based on singular spectrum analysis and support vector regression

Time series modeling and forecasting is an essential and hard task in financial engineering and optimization. Various models have been proposed in the literature and tested on daily data. However, a limited attention has been given to intraday data. In this regard, the current work presents a model for intraday stock price prediction that uses singular spectrum analysis (SSA) and support vector regression (SVR) coupled with particle swarm optimization (PSO). In particular, the SSA decomposes stock price time series into a small number of independent components used as predictors. The SVR is applied to the task of forecasting and PSO is employed to optimize SVR parameters. The performance of our proposed model is compared to the performance of four models widely used in financial prediction: the wavelet transform (WT) coupled with feedforward neural network (FFNN), autoregressive moving average (ARMA) process, polynomial regression (PolyReg), and naïve model. Finally, the mean absolute error (MAE), mean absolute percentage error (MAPE), and the root mean of squared errors (RMSE) are used as main performance metrics. By applying all models to six intraday stock price time series, the forecasting results from simulations show that the presented SSA-PSO-SVR largely outperforms the conventional WT-FFNN, ARMA, polynomial regression, and naïve model in terms of MAE, MAPE and RMSE. Therefore, our proposed predictive system SSA-PSO-SVR shows evident potential for noisy financial time series analysis and forecasting.

Comparisons of transformer top oil temperature calculation models using support vector regression optimised by genetic algorithm

In this study, the comparisons of three available transformer top-oil temperature models for use in dynamic loading are discussed – the modified IEEE clause 7 model, Swift model, and Susa model. Considering the non-linear character of these commonly used models, support vector regression (SVR) is applied for non-linear map between input onsite data and model outputs. Genetic algorithm is adopted to optimise-related parameters of SVR. Three metrics are used for measuring the accuracy of transformer thermal models: mean-squared error, correlation coefficient square, and the maximum error. It is presented that when the cooler condition of transformer is natural oil, natural air, Susa model has the best performance while the modified IEEE 7 model has the best performance when the transformer is forced oil, forced air cooled.

Determination and prediction on “three zones” of coal spontaneous combustion in a gob of fully mechanized caving face

The precise division into “three zones” of coal spontaneous combustion in the gob plays a key role for coal fire fighting. This paper presents three-dimensional distribution maps and contour plots for the gases and temperature in the gob by the method of griddata interpolation according to the data (O2, CO, CO2, CH4, and temperature) acquired from in-situ test, and the variation of gases and temperature. It is proposed to comprehensively divide “three zones” by using O2 concentration of 5–18vol%, the appearance and disappearance of CO, and the heating rate K=0°C/m. The gas explosion conditions were considered to determine the danger zone of coal spontaneous combustion. The minimum mining speed was calculated to be 4.8m/day based on the division of the “three zones” in the gob in order to prevent spontaneous combustion phenomenon. Particle swarm optimization (PSO) was employed to optimize the parameters of support vector regression (SVR); the PSO-SVR model was established to predict the temperature of coal spontaneous combustion based on the gases’ concentration in the gob and distance from the measuring points to the working face. Prediction results and performance of PSO-SVR model were compared with standard SVR, back propagation neural network (BPNN), and multiple linear regression (MLR). The results indicated that PSO-SVR model had greater prediction accuracy and generalization ability, which can predict the temperature of coal spontaneous combustion in the gob.

A novel OFS–TLBO–SVR hybrid model for optimal budget allocation of government schemes to maximize GVA at factor cost

Support Vector Regression (SVR) has already been proved to be one of the most referred and used machine learning technique in various fields. In this study, we have addressed a predictive-cum-prescriptive analysis for finalizing fund allocations by the Government at center to the schemes under Central Plan and to the schemes under States and Union Territories Plan, with a goal to maximize Gross Value Added (GVA) at factor cost. Here, we have proposed a hybrid machine learning model comprising of OFS (Orthogonal Forward Selection), TLBO (Teaching Learning Based Optimization) and SVR for the prediction of GVA at factor cost. In this model, referred as OFS–TLBO–SVR hybrid model, SVR is at the core of prediction mechanism, OFS is for identifying the relevant features, and TLBO is to support in optimizing the free parameters of SVR and again TLBO is used for optimizing the governable attributes of data.

Determination of the excitation origin in the ventricles from the ECG using support vector regression

Abstract A common treatment of focal ventricular tachycardia is the catheter ablation of triggering sites. They have to be found manually by the physician during an intervention in a catheter lab. Thus, a method for determining the position of the focus automatically is desired. The inverse problem of electrocardiography addresses this problem by reconstructing the source of the ectopic beats using the surface ECG. This problem is ill-posed and therefore needs specific methods for solving it. We propose a machine learning approach for localisation of the ectopic foci in the heart to assist cardiologists with their therapy planning.We simulated 600 120-lead ECGs with different known excitation origins in the heart using a cellular automaton followed by a forward calculation. Features from the ECGs were used as input for a support vector regression (SVR). We assumed a functional relation between features from the ECG and the excitation origin. To benchmark SVR, we also used the well-known Tikhonov 0th order regularisation to reconstruct the transmembrane potentials in the heart and detect the location of the ectopic foci. Parameters for SVR and regularisation were chosen using a grid search minimising the error between estimated and true excitation origin. Compared to the Tikhonov regularisation method, SVR achieved a smaller deviation between estimated and real excitation origin evaluated with 6-fold cross validation. Future work could investigate on the behaviour on data from simulations with other torso and electrophysiological models, the influence of other methods for feature extraction and finally the evaluation with clinical data.

Parameter optimization of support vector regression based on sine cosine algorithm

Support vector regression is employed as a time series prediction model.A sine cosine algorithm based method is proposed for parameter tuning of SVR.The proposed SCA-SVR model is compared to other meta-heuristics algorithms.Benchmarks are selected to cover a range of possible practical situations.The SCA-SVR method has been demonstrated to be feasible efficiently and reliably. Time series prediction is an important part of data-driven based prognostics which are mainly based on the massive sensory data with less requirement of knowing inherent system failure mechanisms. Support Vector Regression (SVR) has achieved good performance in forecasting problems of small samples and high dimensions. However, the SVR parameters have a significant influence on forecasting performance of SVR. In our current work, a novel SCA-SVR model has been presented where sine cosine algorithm (SCA) is used to select the penalty and kernel parameters in SVR, so that the generalization performance on unknown data can be improved. To validate the proposed model, the results of the SCA-SVR algorithm were compared with those of grid search and some other meta-heuristics optimization algorithms on common used benchmark datasets. The experimental results proved that the proposed model is capable to find the optimal values of the SVR parameters and can yield promising results.

Black-box identification of a pilot-scale dryer model: A Support Vector Regression and an Imperialist Competitive Algorithm approach

This report describes system identification by means of the hybrid black-box method. The identification was carried out on the regulation object (a dryer model) representing the pilot-scale processes occurring during air conditioning and drying. The new approach proposed in the paper is the use of the imperialist competitive algorithm (ICA) as a tool for selecting the best parameters for support vector regression (SVR) and for selecting an optimal set of regressors. The advantage of this method is that the selection of an optimal set of regressors and the optimal parameters of SVR for this set is performed automatically, which reduces the time needed for identification. The results of the SVR with the ARX, ARMAX, OE, Box-Jenkins (BJ) and low-order transfer function (Tf) models were compared. The research was conducted for two fan speeds, equal to 40% and 60%. The Fit and MSE indicators for the SVR achieved a higher value with respect to those of the ARX, ARMAX, OE, BJ and Tf models. This method is sufficiently universal and can be applied to any plant as an efficient alternative method. This report is supported by National Science Centre grant UMO-2015/17/B/NZ7/02937

A Mixed-Kernel-Based Support Vector Regression Model for Automotive Body Design Optimization Under Uncertainty

Finite element (FE) models are commonly used for automotive body design. However, even with increasing speed of computers, the FE-based simulation models are still too time-consuming when the models are complex. To improve the computational efficiency, support vector regression (SVR) model, a potential approximate model, has been widely used as the surrogate of FE model for crashworthiness optimization design. Generally, in the traditional SVR, when dealing with nonlinear data, the single kernel function-based projection cannot fully cover data distribution characteristics. In order to eliminate the application limitations of single kernel SVR, a method for reliability-based design optimization (RBDO) based on mixed-kernel-based SVR (MKSVR) is proposed in this research. The mixed kernel is constructed based on the linear combination of radial basis kernel function and polynomial kernel function. Through the particle swarm optimization (PSO) algorithm, the parameters of the mixed kernel SVR are optimized. The proposed method is demonstrated through a representative analytical RBDO problem and a vehicle lightweight design problem. And the comparitive studies for SVR and MKSVR in application indicate that MKSVR surpasses SVR in model accuracy.

Swarm Optimized Grey SVR and ARIMA for Modeling of Larceny-Theft Rate with Economic Indicators

As real world data, larceny-theft rates are most likely to have both linear and nonlinear components. A single model such as the linear or nonlinear model may not be sufficient to model the larceny-theft rate. Thus, a hybridization of the linear and nonlinear models is proposed for modeling the larceny-theft rate. The proposed model combines Support Vector Regression (SVR) and Autoregressive Integrated Moving Average (ARIMA) models. Particle swarm optimization is used to optimize the parameters of SVR and ARIMA models. The proposed model is equipped with features selection that combines grey relational analysis and SVR to choose the significant economic indicators for the larceny-theft rate. The experimental results show that the proposed model has better accuracy than the linear, nonlinear, and existing hybrid models in modeling the larceny-theft rate of United States.

A method to precisely control the diesel substitution rate of diesel-natural gas dual fuel engine

Precise control of the diesel substitution rate (DSR) plays a vital role in ensuring the safety and low emissions of diesel-natural gas dual fuel engines. However, an open-loop control strategy is generally employed to control the DSR, which poses a significant challenge to the precise control of the DSR. Therefore, an effective and reliable method to precisely control the DSR is urgently needed for the diesel-natural gas dual fuel engine. In this study, a support vector regression (SVR)-based model is developed to estimate the DSR, and then, a closed-loop control algorithm with the feedback of the estimated DSR is proposed to control the DSR in real time. Five easily available engine variables are employed as the inputs of the estimation model: (1) injection quantity per cycle of diesel (IQPCd); (2) injection quantity per cycle of natural gas (IQPCg); (3) engine speed; (4) inlet air flow; and (5) excess air coefficient. Cross validation is introduced to determine the kernel function and some SVR parameters. Experiments reveal that the proposed estimation model produces a desirable estimation, with a root mean square error (RMSE) of 0.46% and a maximum absolute error (MAE) of 1.89%. Furthermore, in the actual operation, the actual DSR is effectively controlled in the vicinity of the given DSR via the proposed DSR closed-loop control algorithm, with a deviation of less than 1.32%.

Modeling Uncertainty in Tidal Current Forecast Using Prediction Interval-Based SVR

This paper proposes a prediction interval-based model for modeling the uncertainties of tidal current prediction. The proposed model constructs the optimal prediction intervals (PIs) based on support vector regression (SVR) and a nonparametric method called a lower upper bound estimation (LUBE) method. In order to increase the modeling stability of SVRs that are used in the LUBE method, the idea of combined prediction intervals is employed. As the optimization tool, a flower pollination algorithm along with a two-phase modification method is presented to optimize the SVR parameters. The proposed model employs fuzzy membership functions to provide appropriate balance between the PI coverage probability (PICP) and PI normalized average width (PINAW), independently. The performance of the proposed model is examined on the practical tidal current data collected from the Bay of Fundy, NS, Canada.

Support Vector Regression Based GARCH Model with Application to Forecasting Volatility of Financial Returns

In recent years, support vector regression (SVR), a novel neural network (NN) technique, has been successfully used for financial forecasting. This paper deals with the application of SVR in volatility forecasting. Based on a recurrent SVR, a GARCH method is proposed and is compared with a moving average (MA), a recurrent NN and a parametric GACH in terms of their ability to forecast financial markets volatility. The real data in this study uses British Pound-US Dollar (GBP) daily exchange rates from July 2, 2003 to June 30, 2005 and New York Stock Exchange (NYSE) daily composite index from July 3, 2003 to June 30, 2005. The experiment shows that, under both varying and fixed forecasting schemes, the SVR-based GARCH outperforms the MA, the recurrent NN and the parametric GARCH based on the criteria of mean absolute error (MAE) and directional accuracy (DA). No structured way being available to choose the free parameters of SVR, the sensitivity of performance is also examined to the free parameters.

Seismic response of soil-structure interaction using the support vector regression

In this paper, a different technique to predict the effects of soil-structure interaction (SSI) on seismic response of building systems is investigated. The technique use a machine learning algorithm called Support Vector Regression (SVR) with technical and analytical results as input features. Normally, the effects of SSI on seismic response of existing building systems can be identified by different types of large data sets. Therefore, predicting and estimating the seismic response of building is a difficult task. It is possible to approximate a real valued function of the seismic response and make accurate investing choices regarding the design of building system and reduce the risk involved, by giving the right experimental and/or numerical data to a machine learning regression, such as SVR. The seismic response of both single-degree-of-freedom system and six-storey RC frame which can be represent of a broad range of existing structures, is estimated using proposed SVR model, while allowing flexibility of the soil-foundation system and SSI effects. The seismic response of both single-degree-of-freedom system and six-storey RC frame which can be represent of a broad range of existing structures, is estimated using proposed SVR model, while allowing flexibility of the soil-foundation system and SSI effects. The results show that the performance of the technique can be predicted by reducing the number of real data input features. Further, performance enhancement was achieved by optimizing the RBF kernel and SVR parameters through grid search.

Efficient hyper parameter selection for support vector regression using orthogonal array

Support vector regression (SVR) is a nonlinear prediction method using kernel functions and has been widely applied to real-world problems. Although the accuracy of an effectively tuned SVR is high, its performance strongly depends on hyper parameters. Therefore, the determination of the parameters is important when applying SVR to real-world problems. Although the optimum parameters are usually determined by an exhaustive grid search, using this method is not realistic when the sample size is considerably large. To decrease the computational time required to determine the optimum parameters, we employ orthogonal array and propose two efficient methods for SVR parameter tuning based on variable selection in Taguchi method. The proposed methods can reduce the computational time to approximately one-twelfth of that taken by a grid search method. We also validated the actual computational time and accuracy of the proposed methods by applying it to five real datasets in UCI repository.

A Hybrid Accurate Model for Tidal Current Prediction

This paper proposes an accurate hybrid method based on support vector regression (SVR) and autoregressive integrated moving average (ARIMA) to predict the tidal current speed and direction. In the proposed hybrid model, the ARIMA model captures the linear component of the tidal current, and the remained residual components are modeled by SVR. In order to capture the maximum linear components, the appropriate order of the ARIMA model is determined by the Akaike information criterion. Autocorrelation and partial autocorrelation functions are used to verify the stationary or nonstationary characteristics of tidal data. In order to adjust the optimal values of SVR parameters, a new optimization method based on the crow search algorithm is developed to search the problem space globally. In addition, a three-phase modification method is proposed to increase the diversity of the algorithm. The proposed hybrid model is highly accurate and outperforms the artificial neural network, ARIMA, genetic algorithm, and SVR. This model is applied on the practical data collected from the Bay of Fundy, NS, Canada, in 2008.

Efficient hyper parameter selection for support vector regression using orthogonal array

Support vector regression (SVR) is a nonlinear prediction method using kernel functions and has been widely applied to real-world problems. Although the accuracy of an effectively tuned SVR is high, its performance strongly depends on hyper parameters. Therefore, the determination of the parameters is important when applying SVR to real-world problems. Although the optimum parameters are usually determined by an exhaustive grid search, using this method is not realistic when the sample size is considerably large. To decrease the computational time required to determine the optimum parameters, we employ orthogonal array and propose two efficient methods for SVR parameter tuning based on variable selection in Taguchi method. The proposed methods can reduce the computational time to approximately one-twelfth of that taken by a grid search method. We also validated the actual computational time and accuracy of the proposed methods by applying it to five real datasets in UCI repository.

A dynamic SVR–ARMA model with improved fruit fly algorithm for the nonlinear fiber stretching process

The fiber stretching process plays the key role in the process of fiber production and its effects is measured by the stretching ratio. The stretching ratio is determined by the relative speed of the winding roller. The stretching ratio has impact on the performance of the final fiber filament and production directly. Focused on the importance of the stretching ratio, the support vector regression (SVR) predictive model, called nonlinear auto-regressive exogenous model, for the fiber stretching rate based on existing industry data is proposed. Furthermore, the fruit fly optimization algorithm inspired by immune mechanism and cooperation functional (IFOA) is presented, and then is used to optimize the parameters in SVR. Furthermore, taking into account the high cost and accurate precision of the fiber stretching process, a time series autoregressive moving average (ARMA) model is introduced to reduce the prediction error of the IFOA–SVR model. Simulations results demonstrate that the proposed IFOA–SVR method can increase the prediction accuracy than the traditional FOA and the SVR method, and the ARMA model is essential to modify the prediction error of the IFOA–SVR model.

A Novel Hybrid Short Term Load Forecasting Model Considering the Error of Numerical Weather Prediction

In order to reduce the effect of numerical weather prediction (NWP) error on short term load forecasting (STLF) and improve the forecasting accuracy, a new hybrid model based on support vector regression (SVR) optimized by an artificial bee colony (ABC) algorithm (ABC-SVR) and seasonal autoregressive integrated moving average (SARIMA) model is proposed. According to the different day types and effect of the NWP error on forecasting prediction, working days and weekends load forecasting models are selected and constructed, respectively. The ABC-SVR method is used to forecast weekends load with large fluctuation, in which the best parameters of SVR are determined by the ABC algorithm. The working days load forecasting model is constructed based on SARIMA modified by ABC-SVR (AS-SARIMA). In the AS-SARIMA model, the ability of SARIMA to respond to exogenous variables is improved and the effect of NWP error on prediction accuracy is reduced more than with ABC-SVR. Contrast experiments are constructed based on International Organization for Standardization (ISO) New England load data. The experimental results show that prediction accuracy of the proposed method is less affected by NWP error and has higher forecasting accuracy than contrasting approaches.

A novel hybrid model for tourist volume forecasting incorporating search engine data

ABSTRACTThe precise prediction of tourism demand has long presented a challenge for both tourism professionals and academics. Tourist volume forecasting is a nonlinear problem, support vector regression (SVR) can approximate a nonlinear system with enough precision, but parameters tuning has always been an obstacle to developing SVR with good generalization potential. Furthermore, previous research mainly used historical observations of tourism demand as the inputs of SVR. This study introduces an approach that hybridizes SVR with the Bat algorithm (BA), namely BA-SVR, to forecast tourist volume by incorporating search engine data. In this model, BA is used to adjust the SVR parameters. To validate our proposed approach, tourist volume data for China’s Hainan province from August 2008 to October 2015 were used in conjunction with corresponding search engine data as numerical examples. The 12-month simulation forecasts indicate that the BA-SVR is an effective method that can outperform its traditional counterparts.

A Novel Hybrid Method of Parameters Tuning in Support Vector Regression for Reliability Prediction: Particle Swarm Optimization Combined With Analytical Selection

Support vector regression (SVR) is a widely used technique for reliability prediction. The key issue for high prediction accuracy is the selection of SVR parameters, which is essentially an optimization problem. As one of the most effective evolutionary optimization methods, particle swarm optimization (PSO) has been successfully applied to tune SVR parameters and is shown to perform well. However, the inherent drawbacks of PSO, including slow convergence and local optima, have hindered its further application in practical reliability prediction problems. To overcome these drawbacks, many improvement strategies are being developed on the mechanisms of PSO, whereas there is little research exploring a priori information about historical data to improve the PSO performance in the SVR parameter selection task. In this paper, a novel method controlling the inertial weight of PSO is proposed to accelerate its convergence and guide the evolution out of local optima, by utilizing the analytical selection (AS) method based on a priori knowledge about SVR parameters. Experimental results show that the proposed ASPSO method is almost as accurate as the traditional PSO and outperforms it in convergence speed and ability in tuning SVR parameters. Therefore, the proposed ASPSO-SVR shows promising results for practical reliability prediction tasks.