EFFECTIVE DIMENSION REDUCTION OF HIGH DIMENSIONAL DATA SETS BASED ON PRINCIPAl COMPONENTS ANALYSIS AND WAVELETS

Dimension reduction is a critical and essential data preprocessing technique for high dimensional data, where the data contains hundreds of attributes (features). It can be used to improve both the efficiency and the effectiveness of mining algorithms. One of the most important techniques for solving the dimension reduction problem is the statistical method, principal components analysis (PCA). PCA method applies global linear transformation that transforms the original dataset into a reduced space that capture most of the information in the original dataset. Many algorithms that exploit PC4 method to solve the dimension reduction problem have been proposed. At the same time, recently there has significant development in the use of wavelets for solving the dimension reduction problem, but has largely focused on data that is univariate (i.e. one feature or signal). In this paper, we propose an algorithm that combines PCA method with wavelets method to solve the dimension reduction problem. Dimension reduction is based on finding a linear transformation that transforms the high dimensional dataset to a low dimensional dataset. Our experiments on a test dataset show that the proposed algorithm is more effective and more condense to the information than PCA algorithm. However, the computational complexity of the proposed algorithm is close to a recently fast PCA proposed algorithm.