Support vector machine

"Rainfall-runoff simulation and prediction in watersheds is one of the most important tasks in water resources management. In this research, an adaptive data analysis methodology, ensemble empirical mode decomposition (EEMD), is presented for decomposing annual rainfall series in a rainfall-runoff model based on a support vector machine (SVM). In addition, the particle swarm optimization (PSO) is used to determine free parameters of SVM. The study data from a large size catchment of the Yellow River in China are used to illustrate the performance of the proposed model. In order to measure the forecasting capability of the model, an ordinary least-squares (OLS) regression and a typical three-layer feed-forward artificial neural network (ANN) are employed as the benchmark model. The performance of the models was tested using the root mean squared error (RMSE), the average absolute relative error (AARE), the coefficient of correlation (R) and Nash–Sutcliffe efficiency (NSE). The PSO–SVM–EEMD model improved ANN model forecasting (65.99%) and OLS regression (64.40%), and reduced RMSE (67.7%) and AARE (65.38%) values. Improvements of the forecasting results regarding the R and NSE are 8.43%, 18.89% and 182.7%, 164.2%, respectively. Consequently, the presented methodology in this research can enhance significantly rainfall-runoff forecasting at the studied station.

Key words: annual rainfall-runoff, artificial neural networks, ensemble empirical mode decomposition, forecasting, particle swarm optimization, support vector machine"

"Due to the ability of malicious worms to spread quickly over the Internet, detection and reaction mechanisms are needed in order to stop the progress of the infection process. Existing worm thwarting techniques do not take into account the false negative and false positive probabilities when implementing the responses. In this paper, we study the impact of the efficiency of the worm detection process on the selected responses. To this purpose, we develop a cost-benefit model allowing the representation of the impact caused by the worm propagation as well as the outcome of the potential reaction. We show that our approach enables the selection of the convenient security solution, especially with respect to the reaction time, which is not considered by the traditional worm response techniques. The efficiency of the proposed approach is corroborated by practical experiments.

http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5893862&tag=1  "

Railway turnout systems are one of the most critical pieces of equipment in railway infrastructure. Early identification of failures in turnout systems is important to obtain increased availability and safety, and reduced operating and support costs. This paper aims to develop a method to identify ‘drive-rod out-of-adjustment’ failure mode, one of the most frequently observed failure modes. Support Vector Machine (SVM) with Gaussian kernel is used for diagnosis. In addition, the results of feature selection with statistical t-test and feature reduction with principal component analysis (PCA) are compared in the paper.

""Accurate time- and site-specific forecasts of streamflow and reservoir inflow are important in effective hydropower reservoir management and scheduling. Traditionally, autoregressive moving-average (ARMA) models have been used in modelling water resource time series as a standard representation of stochastic time series. Recently, artificial neural network (ANN) approaches have been proven to be efficient when applied to hydrological prediction. In this paper, the support vector machine (SVM) is presented as a promising method for hydrological prediction. Over-fitting and local optimal solution are unlikely to occur with SVM, which implements the structural risk minimization principle rather than the empirical risk minimization principle. In order to identify appropriate parameters of the SVM prediction model, a shuffled complex evolution algorithm is performed through exponential transformation. The SVM prediction model is tested using the long-term observations of discharges of monthly river flow discharges in the Manwan Hydropower Scheme. Through the comparison of its performance with those of the ARMA and ANN models, it is demonstrated that SVM is a very potential candidate for the prediction of long-term discharges.

Keywords: amoregressive moving-average (ARMA) models; long-term discharge prediction; neural networks; SCE-UA algorithm; support vector machine""

In this paper we study the problem of model selection for a linear programming-based support vector machine for regression. We propose generalized method that is based on a quasi-Newton method that uses a globalization strategy and an inexact computation of first order information. We explore the case of two-class, multi-class, and regression problems. Simulation results among standard datasets suggest that the algorithm achieves insignificant variability when measuring residual statistical properties.

"Condition classification is an important step in machinery fault detection, which is a problem of pattern recognition. Currently, there are a lot of techniques in this area and the purpose of this paper is to investigate two popular recognition techniques, namely hidden Markov model and support vector machine. At the begiIming, we briefly introduced the procedure offeature extraction and the theoretical background ofthis paper. The comparison experiment was conducted for gearbox fault detection and the analysis results from this work showed that support vector machine has better classification performance in this area.

Keywords: Pattern recognition; Feature extraction; Hidden markov model; Support vector machine"

Problem statement: The Pecan weevil was considered as the most dangerous pest of Pecan fruits. The aim of this research is to evaluate Support Vector Machine method (SVM) for identifying Pecan Weevil among other insects. Eventually, this recognition system will serve in a wireless imaging network for monitoring Pecan Weevils. Approach: SVM has been evaluated using two different kernel functions i.e., Polynomial Function and Radial Basis Function. Database of 205 Pecan Weevils and 75 other insects which typically exist in pecan habitat has been used. Three sets of input data for SVM have been generated by two standard region-based recognition methods. These sets are comprised of output obtained by Zernike Moments, Regional Properties and combination of these two methods. For each kernel function, the system had been trained by 25, 50 and 75% of data and remaining ratio in each case has been used for testing. Each experiment is repeated ten times and average results are considered for comparisons and analysis. Results: The optimum recognition rate had been found when system is trained by 75% of data. The results are approximately similar when the input data is obtained by Regional Properties and combination of Regional Properties and Zernike Moments methods. The optimum results are obtained when input data has been obtained by Zernike Moments alone for lower values of sigma ‘σ’. The proposed system is able to successfully recognize 99% of Pecan Weevil and 97% of the other insects using the radial basis function. The proposed system took approximately 31 sec for processing 75% of the data which include the time for training. The testing time is found to be 0.15 sec. Conclusion: Promising results can be obtained when input data is obtained by Zernike Moments and SVM is trained by RBF and 75% of data.

Problem statement: Red palm weevil (Rynchophorus Ferrugineous, Oliveir) is an insect which threatens the existence of palm trees. The proposed research is to develop a RPW identification system using Support Vector Machine method. The problem is to extract image features from an image and using SVM to find out the existence of RPW in an image. Approach: Images are snapped and image processing techniques of Regional Properties and Zernike Moments are used to extract different features of an image. The obtained features are fed into the SVM based system individually as well as in combination. The database used to train and test the system includes 326 RPW and 93 other insect images. The input data from database is selected randomly and fed into the system in three steps i.e., 25, 50 and 75% while remaining database is used for testing purpose. In SVM, polynomial kernel function and Radial Basis Function are used for training. Each experiment is repeated 10 times and the average results are used for analysis. Results: The optimal results are obtained by using Radial Basis Function in SVM at lower values of sigma ‘σ’ while Polynomial kernel function is not successful in returning adequate results. Further detailed analysis of results for ‘σ’ value of 10 and 15 revealed that proposed system works well with large training data and with inputs obtained by Regional Properties. The optimal value of ‘σ’ for proposed system is found to be 10 when training data ratio is 50%. The training time for proposed system depends on size of database and is found to be 0.025 sec per image while time consumed by proposed system for identification of RPW in an image is found to be 15 milli sec. The proposed system’s success in identification of RPW and other insect is found to be 97 and 93% respectively. Conclusion: It is concluded that SVM based system using Radial Basis Function having ‘σ’ value of 10 is optimal in identifying RPW from an image. The optimal input data for the proposed system needs to be obtained by Regional Properties only.