ABSTRACT : |
In data mining dimension reduction is widely used in Medicine, Bioinformatics etc. Selection of features from huge dataset is tedious. Usually high dimension data contains noise , irrelevant information and small amount of relevant information. Reduction of dimensionality is very important to extract the important features, which is useful for predicting the results. This proposal layouts the high dimension data reduction using three ways.(i)Feature Selection (ii)Linear Dimensionality Reduction (iii)Non-Linear Dimensionality Reduction. In this work Feature Selection based on mutual information for feature filtering to select the relevant features with minimal redundancy. Linear Dimension Reduction is used in high dimension dataset for extracting the latent variables. The Non-linear dimension reduction is used to reduce the dimension for visalizing. Results are presented to show the efficiency of this work.
Keywords: Microarray Dataset, Dimension Reduction, Feature Selection |
|