Search

Machine learning methods have been successfully applied to the phenotype classification of many diseases based on static gene expression measurements. More recently microarray data have been collected over time, making available datasets composed by time series of expression gene profiles. In this paper we propose a new method for time series classification, based on a temporal extension of L1-norm support vector machines, that uses dynamic time warping distance for measuring time series similarity. This results in a mixed-integer optimization model which is solved by a sequential approximation algorithm. Computational tests performed on two benchmark datasets indicate the effectiveness of the proposed method compared to other techniques, and the general usefulness of the approaches based on dynamic time warping for labeling time series gene expression data. [ABSTRACT FROM AUTHOR]

The Human Immunodeficiency Virus (HIV) encodes an enzyme, called HIV protease, which is responsible for the generation of infectious viral particles by cleaving the virus polypeptides. Many efforts have been devoted to perform accurate predictions on the HIV-protease cleavability of peptides, in order to design efficient inhibitor drugs. Over the last decade, linear and nonlinear supervised learning methods have been extensively used to discriminate between protease-cleavable and non cleavable peptides. In this paper we consider four different proteins encoding schemes and we apply a discrete variant of linear support vector machines to predict their HIV protease-cleavable status. Empirical results indicate the effectiveness of the proposed method, that is able to classify with the highest accuracy the cleavable and non cleavable peptides contained in two publicly available benchmark datasets. Moreover, the optimal classification rules generated are characterized by a strong generalization capability, as shown by their accuracy in predicting the HIV protease cleavable status of peptides in out-of-sample datasets. [ABSTRACT FROM AUTHOR]

A vast majority of classification problems are characterized by an intrinsic vagueness in the knowledge of the class label associated to each example. In this paper we propose a classifier based on fuzzy discrete support vector machines, that takes as input a binary classification problem together with a membership value for each example, and derives an optimal separation rule by solving a mixed-integer optimization problem. We consider different methods for computing the membership function: some are based on a metric defined in the attribute space; some derive the membership function from a scoring generated by a probabilistic classifier; others make use of frequency voting by an ensemble classifier. To evaluate the overall accuracy of the fuzzy discrete SVM, and to investigate the effect of the alternative membership functions, computational tests have been performed on benchmark datasets. They show that fuzzy discrete SVM is an accurate classification method capable to generate robust rules and to smooth out the effect of outliers. [ABSTRACT FROM AUTHOR]

Discrete support vector machines are models for classification recently introduced in the context of statistical learning theory. Their distinctive feature is the formulation of mixed integer programming problems aimed at deriving optimal separating hyperplanes with minimum empirical error and maximum generalization capability. A new family of discrete SVM is proposed in this paper, for which the hyperplane establishes a variable softening of the margin to improve the separation among distinct classes. Theoretical bounds are derived to finely tune the parameters of the optimization problem. Computational tests on benchmark datasets in the biolife science application domain indicate the effectiveness of the proposed approach, that appears dominating against traditional SVM in terms of accuracy and percentage of support vectors. [ABSTRACT FROM AUTHOR]

We present a rule induction method based on decision trees for classification and prediction problems. Our approach to tree construction relies on a discrete variant of support vector machines, in which the error is expressed by the number of misclassified instances, in place of the misclassification distance considered by traditional SVMs, and an additional term is included to reduce the complexity of the generated rule. This leads to the formulation of a mixed integer programming problem, whose approximate solution is obtained via a sequential LP-based algorithm. The decision tree is then built by means of a multivariate split derived at each node from the approximate solution of the discrete SVM. Computational tests on well-known benchmark datasets indicate that our classifier achieves a superior trade-off between accuracy and complexity of the induced rules, outperforming other competing approaches. [ABSTRACT FROM AUTHOR]