Medical diagnosis is an important task that needs to be executed accurately and efficiently. Medical domain complexities are represented by multidimensional heterogeneous datasets. Computer aided diagnosis must deal with processing and analyzing high dimensional data. Optimization of features in datasets reduces time and memory complexity of learning algorithms. It is necessary to have a tool that gives relationship between features and eliminate redundant ones. Feature selection or feature extraction reduce dimensions and essentially influence the performance of classifier. Many techniques have been used to determine essential features of medical data. We investigate two feature extraction techniques, Principal component analysis (PCA) and common Factor Analysis (FA) techniques for classification of heart disease. These techniques expose the structure, while maintaining the integrity of the data, thus improving diagnosis performance.

Dimensionality reduction PCA FA neural networks AUC heart disease