期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2015
卷号：6
期号：3
页码：2354-2360
出版社：TechScience Publications
摘要：Data dimensionality refers to the number of variables that are measured on each observation. Recent trends in technology and applications result in the generation of huge volume of high dimensional data. The analysis of these data is inevitable for various research and production activities. Data analysis focuses on understanding, manipulating and interpreting large scale data. Relevant information is hidden in this huge volume dataset which needs to be extracted for analysis. Several methods have been developed in the field of data mining for automated data processing. Owing to the huge dimension of data these methods fail to meet the requirement efficiently. Dimensionality reduction offers an optimal solution to this problem by reducing the data dimension. It transforms data into a meaningful and reduced dimension space with minimal information loss. This reduces the computational cost involved in data analysis and founds effective in data compression, visualization and big data analysis. Dimension reduction is applicable in many real world domains such as regression analysis, cluster analysis, computer vision, image processing, text categorization and so on. There are various classes of dimension reduction techniques such as supervised, unsupervised, linear, nonlinear etc. The paper presents a concise review of some relevant linear and nonlinear dimensionality reduction techniques.
