Search

Variable Selection and Oversampling in the Use of Smooth Support Vector Machines for Predicting the Default Risk of Companies

Many real-world visual classification tasks require one to recognize test data in a particular domain of interest, while the training data can only be collected from a different domain. This can be viewed as the problem of unsupervised domain adaptation, in which the domain difference and the lack of cross-domain label/correspondence information make the recognition task very difficult. In this paper, we propose to exploit the cross-domain data correspondence using both observed data similarity and labels transferred from the source domain. This allows us to perform distribution matching for cross-domain data with recognition guarantees. Our experiments on three different cross-domain visual classification tasks would confirm the effectiveness of our method, which is shown to perform favorably against state-of-the-art unsupervised domain adaptation approaches.

In real-world video surveillance applications, one often needs to recognize face images from a very long distance. Such recognition tasks are very challenging, since such images are typically with very low resolution (VLR). However, if one simply downsamples high-resolution (HR) training images for recognizing the VLR test inputs, or if one directly upsamples the VLR inputs for matching the HR training data, the resulting recognition performance would not be satisfactory. In this paper, we propose a joint face hallucination and recognition approach based on sparse representation. Given a VLR input image, our method is able to synthesize its person-specific HR version with recognition guarantees. In our experiments, we consider two different face image datasets. Empirical results will support the use of our approach for both VLR face recognition. In addition, compared to state-of-the-art super-resolution (SR) methods, we will also show that our method results in improved quality for the recovered HR face images.

Considering the diverse application scenarios involving wireless sensor networks (WSNs), accurate continuous monitoring requires a solution to the essential task of estimating unmeasured locations in the monitored space. In this paper, we utilize Epsilon-Smooth Support Vector Regression (Epsilon-SSVR) to report monitoring information of environment, furthermore we combine spatial and temporal correlation to strengthen monitoring accuracy. However if our sensors are too sparsely deployed, the resulting coverage holes problem will adversely impact the monitoring result. Therefore, we utilize Uniform Design and different local interpolation methods to assist Epsilon-SSVR to mitigate the coverage holes problem. In our experiment, we compare our method with different methods applied to different sensors deployments. Epsilon-SSVR has better accuracy and computation speed than others. Besides continuous monitoring, we also propose a distributed anomaly detection mechanism to report anomaly information, in order to provide a reliable and real time anomaly monitoring system.

Time series classification has attracted increasing attention in machine learning and data mining. In the analysis of time series data, how to represent data is a critical step for the performance. Generally, we can regard each time stamp as a feature dimension for time series data instance. However, this naive representation might be not suitable for data analysis due to the over-fitting of data. To address this problem, we proposed a temporal bag-of-words representation for time series classification. A codebook is generated by the representative subsequences from the time series data. Consequently, we encode a time series data instance by the codebook, which describes different local patterns of time series data. In our experiments, we demonstrate that our proposed method can achieve better results by comparing with competitive methods.

We present a novel transfer learning approach to cross-camera action recognition. Inspired by canonical correlation analysis (CCA), we first extract the spatio-temporal visual words from videos captured at different views, and derive a correlation subspace as a joint representation for different bag-of-words models at different views. Different from prior CCA-based approaches which simply train standard classifiers such as SVM in the resulting subspace, we explore the domain transfer ability of CCA in the correlation subspace, in which each dimension has a different capability in correlating source and target data. In our work, we propose a novel SVM with a correlation regularizer which incorporates such ability into the design of the SVM. Experiments on the IXMAS dataset verify the effectiveness of our method, which is shown to outperform state-of-the-art transfer learning approaches without taking such domain transfer ability into consideration.

This paper presents a novel sparse representation for robust face recognition. We advance both group sparsity and data locality and formulate a unified optimization framework, which produces a locality and group sensitive sparse representation (LGSR) for improved recognition. Empirical results confirm that our LGSR not only outperforms state-of-the-art sparse coding based image classification methods, our approach is robust to variations such as lighting, pose, and facial details (glasses or not), which are typically seen in real-world face recognition problems.

We aim at providing a comprehensive introduction to Support Vector Machines and their applications in computational finance. Based on the advances of the statistical learning theory, one of the first SVM algorithms was proposed in mid 1990s. Since then, they have drawn a lot of research interests both in theoretical and application domains and have became the state-of-the-art techniques in solving classification and regression problems. The reason for the success is not only because of their sound theoretical foundation but also their good generalization performance in many real applications. In this chapter, we address the theoretical, algorithmic and computational issues and try our best to make the article self-contained. Moreover, in the end of this chapter, a case study on default prediction is also presented. We discuss the issues when SVM algorithms are applied to bankruptcy prognosis such as how to deal with the unbalanced dataset, how to tune the parameters to have a better performance and how to deal with large scale dataset.

We propose a novel multiple kernel learning (MKL) algorithm with a group lasso regularizer, called group lasso regularized MKL (GL-MKL), for heterogeneous feature selection. We extend the existing MKL algorithm and impose a mixed l 1 and l 2 norm constraint (known as group lasso) as the regularizer. Our GL-MKL determines the optimal base kernels, including the associated weights and kernel parameters, and results in a compact set of features for comparable or improved recognition performance. The use of our GL-MKL avoids the problem of choosing the proper technique to normalize the feature attributes collected from heterogeneous domains (and thus with different properties and distribution ranges). Our approach does not need to exhaustively search for the entire feature space when performing feature selection like prior sequential-based feature selection methods did, and we do not require any prior knowledge on the optimal size of the feature subset either. Comparisons with existing MKL or sequential-based feature selection methods on a variety of datasets confirm the effectiveness of our method in selecting a compact feature subset for comparable or improved classification performance.

Standard linear discriminant analysis (LDA) is known to be computationally expensive due to the need to perform eigen-analysis. Based on the recent success of least-squares LDA (LSLDA), we propose a novel rank-one update method for LSLDA, which not only alleviates the computation and memory requirements, and is also able to solve the adaptive learning task of concept drift. In other words, our proposed LSLDA can efficiently capture the information from recently received data with gradual or abrupt changes in distribution. Moreover, our LSLDA can be extended to recognize data with newly-added class labels during the learning process, and thus exhibits excellent scalability. Experimental results on both synthetic and real datasets confirm the effectiveness of our propose method.

Outlier detection is an important issue in datamining and has been studied in different research areas. It can be used for detecting the small amount of deviated data. In this article, we use “Leave One Out” procedure to check each individual point the “with or without” effect on the variation of principal directions. Based on this idea, an over-sampling principal component analysis outlier detection method is proposed for emphasizing the influence of an abnormal instance (or an outlier). Except for identifying the suspicious outliers, we also design an on-line anomaly detection to detect the new arriving anomaly. In addition, we also study the quick updating of the principal directions for the effective computation and satisfying the on-line detecting demand. Numerical experiments show that our proposed method is effective in computation time and anomaly detection.

In the era of Basel II a powerful tool for bankruptcy prognosis is vital for banks. The tool must be precise but also easily adaptable to the bank's objections regarding the relation of false acceptances (Type I error) and false rejections (Type II error). We explore the suitability of Smooth Support Vector Machines (SSVM), and investigate how important factors such as selection of appropriate accounting ratios (predictors), length of training period and structure of the training sample influence the precision of prediction. Furthermore we show that oversampling can be employed to gear the tradeoff between error types. Finally, we illustrate graphically how different variants of SSVM can be used jointly to support the decision task of loan officers.

Anomaly detection aims to identify rare events that deviate remarkably from existing data. To satisfy real-world appli- cations, various anomaly detection technologies have been proposed. Due to the resource constraints, such as limited energy, computation ability and memory storage, most of them cannot be directly used in wireless sensor networks (WSNs). In this work, we proposed a hierarchical anomaly detection framework to overcome the challenges of anomaly detection in WSNs. We aim to detect anomalies by the accurate model and the approximated model learned at the re- mote server and sink nodes, respectively. Besides the framework, we also proposed an approximated local outlier factor algorithm, which can be learned at the sink nodes. The proposed algorithm is more efficient in computation and storage by comparing with the standard one. Experimental results verify the feasibility of our proposed method in terms of both accuracy and efficiency.