Reducing the dimensionality of data for analysis using the principal component method

Despite the fact that modern data mining systems have high computing power, the amount of data to analyze is constantly increasing and can become a critical factor. Thus, the task of reducing the dimensionality of the source data for analysis without reducing the quality of the analysis itself becomes relevant. One of the methods that allows you to reduce the dimensionality of the data is the principal component method. The paper considers the application of this method in data analysis in sensor network nodes. The advantage of the method is that there are no preliminary hypotheses about the condition of the object under study. The implementation of the method is linear and cyclic, which determines its good algorithmization by computer technology. As the initial data set, a set of wireless sensor network operation data is used, which consists of one thousand nodes. For each node, a selection of measurements on the main parameters of the quality of service is presented. The initial data is being preprocessed. A covariance matrix is constructed for which the eigenvalues and eigenvectors are found. The result of the method is the main components obtained by converting the eigenvectors. These components are used for data analysis. The result of this work is a reduction in the dimension of the data.