Data Classification Based on the Features Reduction and Piecewise Linear Separation

Abstract

The article discusses information technology for classifying data using space reduction methods to the level of data visualization. The components of the steps of using information technology are discussed in detail. The basis is the reduction of the multidimensional feature space into a space that can be visualized. The next step is the construction of piecewise linear class separators by the user on the training set. Further, the borders are projected into the original space of multidimensional objects. Thus, the boundaries of classes are determined, which is in the proposed technology a trained classification system. The advantage is the construction of a flexible non-linear classification system using piecewise linear separators. The proposed method is based on the use of visualization data. A distinctive feature of information technology is that the construction of class delimiters is done by user. Based on visual data analysis, user determines the location of class boundaries and spaces. Thus, technology provides user with convenient tools for data analyzing and classifying.