A fruitful approach various dimension reduction techniques

Dimensionality reduction is a fruitful approach to diminish the number of arbitrary variables under consideration. This research study is carried out with an intent to study and analyze various dimensionality reduction techniques. In this study two main techniques principal component analysis and linear discriminant analysis have been studied in detail and analyzed. Principal component analysis is a statistical based technique that reduces the data size and preserves most of the variance in the form of new variables called principal components. Linear Discriminant Analysis is a generalization of Fisher's linear discriminant. It searches for linear combination of features that best separates the classes. Both the techniques are applied separately on the datasets MATLAB is used to evaluate and analyze both the techniques on all these datasets separately. PCA does not take any class information, it just converts the variables into principal components on the basis of eigenvalues and the components with highest eigenvalues contain highest variance of the datasets and thus the components with lowest eigenvalues can be ignored.