Extension Of Support Vector Machine Research Articles

Sequence classification via large margin hidden Markov models

We address the sequence classification problem using a probabilistic model based on hidden Markov models (HMMs). In contrast to commonly-used likelihood-based learning methods such as the joint/conditional maximum likelihood estimator, we introduce a discriminative learning algorithm that focuses on class margin maximization. Our approach has two main advantages: (i) As an extension of support vector machines (SVMs) to sequential, non-Euclidean data, the approach inherits benefits of margin-based classifiers, such as the provable generalization error bounds. (ii) Unlike many algorithms based on non-parametric estimation of similarity measures that enforce weak constraints on the data domain, our approach utilizes the HMM's latent Markov structure to regularize the model in the high-dimensional sequence space. We demonstrate significant improvements in classification performance of the proposed method in an extensive set of evaluations on time-series sequence data that frequently appear in data mining and computer vision domains.

An extended support vector machine forecasting framework for customer churn in e-commerce

In order to accurately forecast and prevent customer churn in e-commerce, a customer churn forecasting framework is established through four steps. First, customer behavior data is collected and converted into data warehouse by extract transform load (ETL). Second, the subject of data warehouse is established and some samples are extracted as train objects. Third, alternative predication algorithms are chosen to train selected samples. Finally, selected predication algorithm with extension is used to forecast other customers. For the imbalance and nonlinear of customer churn, an extended support vector machine (ESVM) is proposed by introducing parameters to tell the impact of churner, non-churner and nonlinear. Artificial neural network (ANN), decision tree, SVM and ESVM are considered as alternative predication algorithms to forecast customer churn with the innovative framework. Result shows that ESVM performs best among them in the aspect of accuracy, hit rate, coverage rate, lift coefficient and treatment time. This novel ESVM can process large scale and imbalanced data effectively based on the framework.

Integration of Soft and Hard Classifications Using Extended Support Vector Machines

In this letter, the supervised classification algorithm support vector machines is extended to map both pure pixels and mixed pixels using hyperspectral data. The margins between the hyperplanes formed by the pixels on the class boundaries are recognized as mixed region, and the space beyond this region is related to pure pixels. In this way, each endmember is modeled by a set of training samples instead of a single (representative) spectrum to accommodate the variations within the relative pure pixels due to system noise. Unmixing outputs generate an integrated soft- and hard-classification map. The better performance comparing with conventional spectral unmixing method was demonstrated using hyperspectral data sets.

Underwater bottom still mine classification using robust time–frequency feature and relevance vector machine

A novel approach to the problem of detecting and classifying underwater bottom mine objects in littoral environments from acoustic backscattered signals is considered. We begin by defining robust short-time Fourier transform to convert the received echo into a time–frequency (TF) plane. Identify interest local region in spectrogram, then features in TF plane with robustness to reverberation and noise disturbances are built. Finally, echo features are sent to a relevance vector machine (RVM) classifier that represents a Bayesian extension of support vector machine (SVM). To evaluate the performace of the classifier based on this approach, the classification experiment of two typical types of mines lying on the bottom have been performed with a broad bandwidth active sonar. Each of the targets was lying on the lake bottom at a depth of 20 m. The case study exploits the robustness of a feature extraction scheme, and furthermore, RVM yields a much sparser solution and improves the classification accuracy than SVM in an impulse noise environment.

High impedance fault detection in distribution feeders using extended kalman filter and support vector machine

AbstractThe paper presents an intelligent technique for high impedance fault (HIF) detection using combined extended kalman filter (EKF) and support vector machine (SVM). The proposed approach uses magnitude and phase change of fundamental, 3rd, 5th, 7th, 11th and 13th harmonic component as feature inputs to the SVM. The Gaussian kernel based SVM is trained with input sets each consists of ‘12’ features with corresponding target vector ‘1’ for HIF detection and ‘−1’ for non‐HIF condition. The magnitude and phase change are estimated using EKF. The proposed approach is trained with 300 data sets and tested for 200 data sets including wide variations in operating conditions and provides excellent results in noisy environment. Thus, the proposed method is found to be fast, accurate, and robust for HIF detection in distribution feeders. Copyright © 2009 John Wiley & Sons, Ltd.

Mutagenic probability estimation of chemical compounds by a novel molecular electrophilicity vector and support vector machine

Mutagenicity is among the toxicological end points that pose the highest concern. The accelerated pace of drug discovery has heightened the need for efficient prediction methods. Currently, most available tools fall short of the desired degree of accuracy, and can only provide a binary classification. It is of significance to develop a discriminative and informative model for the mutagenicity prediction. Here we developed a mutagenic probability prediction model addressing the problem, based on datasets covering a large chemical space. A novel molecular electrophilicity vector (MEV) is first devised to represent the structure profile of chemical compounds. An extended support vector machine (SVM) method is then used to derive the posterior probabilistic estimation of mutagenicity from the MEVs of the training set. The results show that our model gives a better performance than TOPKAT (http://www.accelrys.com) and other previously published methods. In addition, a confidence level related to the prediction can be provided, which may help people make more flexible decisions on chemical ordering or synthesis. The binary program (ZGTOX_1.1) based on our model and samples of input datasets on Windows PC are available at http://dddc.ac.cn/adme upon request from the authors.

SAR Imagery Classification using Multi-class Support Vector Machines

In this paper, we present the application to SAR imagery classification of a novel pattern recognition technique named Multi-class Support Vector Machines (M-SVMs). M-SVMs are a n-ary extension of Support Vector Machines (SVM), introduced by Vapnik within the framework of the Statistical Learning Theory. In this article we use the M-SVMs in order to classify an ERS-1 SAR multi-frequency survey of Torre de Hercules coast, Spain (December 13, 1992). The main objective of this work is to evaluate the classification performances of M-SVMs in comparison with the most frequently employed Neural Networks and Fuzzy classifiers. M-SVMs provided interesting results with respect to Neural Networks and Fuzzy classifiers, having a reliability factor around to 94%.