Plenary lecture 2: advanced classification and regression algorithms
Dana Simian · 2009
The aim of this paper is to present and analyzed new classification and regression algorithms and applications of these algorithms for different kind of data. One of the practical applications of our algorithms is in the ecological characterization of different areas. Generally, the task of classification is to find a rule, which based on external observations assigns an object to one of several classes. A classification task supposes the existence of training and testing data given in the form of data instances. Each instance in the training set contains one target value, named class label and several attributes named features. One of the approaches used for solving the problem of binary or multiclass classification is represented by SVM algorithm. SVM models are obtained by convex optimization and are able to learn and generalize in high dimensional input spaces. The goal of SVM is to produce a model which predicts target value of data instances in the testing set which are given only the attributes. A very powerful idea, which can be used not only in SVM algorithms, is the kernel method. Using an appropriate kernel, the data are projected in a space with higher dimension in which they are separable by a hyperplane. Usually simple kernels are used but the real problems require more complex kernels. The kernel substitution can be used to define many other types of learning machines distinct from SVMs. We introduce and analyze multiple kernels based on simple kernels. Our intention was to study the possibility of obtaining nonlinear multiple kernels using simple classifiers and to analyze their performance comparing with other multiple kernels. Therefore we choose, first, for experiments the most common data sets, used in a great number of papers and taken from libsvm website. Second, we used different types of data obtained in the monitoring process of many ecological areas from Romania. In order to take advantage of possible correlations between the outputs to improve the quality of the predictions, we also consider the Support Vector Regression method (SVR). SVR is based on the theory of Support Vector Machines and belongs to the category of reproducing kernel methods. The kernel is viewed as the covariance of a second order Gaussian process. SVR builds a model, f, of the output of a system that depends on a set of factors. One of the problems we confront with, in the ecological characterization of different areas is the quantification of the outputs of our system. In order to apply SVR method, we make a regression analysis for studying the relationships among the outputs of our system.