抽象的な

Comparative Analysis of Dimensionality Reduction Techniques

S.Vijayarani, S. Maria Sylviaa

Datasets are most important for performing all the type of data mining tasks. Every dataset has many numbers of attributes and instances. Dimensionality reduction (DR) is one of the preprocessing steps which is used to reduce the dimensions (attributes or features) without losing the data. There are two divisions of reduction they are feature extraction and feature reduction. Feature extraction is the process of decomposition of attributes of the original data (i.e.) merging the attributes of the data Feature selection is the process of selecting the subset of attributes by eliminating features with little or no predictive information. Feature extraction techniques are more adequate than the feature selection. Reduction is done to the larger dataset to decrease the curse of dimensionality. The main objective of this paper is to provide a systematic comparative analysis on feature reduction algorithms such as PCA, LDA and FA to medical dataset (Thyroid, Oesophagal).The performance factor considered are number of attributes reduced and time is observed.