Dimensionality Reduction Method Based Feature Ranking Techniques for Diagnosis of Cardiac Disease Using AI

The computational time and cost of medical diseases diagnosis are so critical. One possible solution to reduce the dataset size of the medical data, ideally without losing predictive performance. High-dimensional data poses a severe challenge for data mining. This paper presents a systematic survey of the computational steps in automated disease diagnosis based on its features (attributes).These computational steps are: 1.) feature extraction to quantify the important features of the disease 2.) Ranking algorithms to reduce the dimensions (features) obtained after feature extraction 3.)Obtaining the final reduced number of features using feature Re-ranking algorithms and 4.) Classifying the selected features to identify the type of disease. Among various dimensionality reduction methods, PCA plays a vital role. However, the experiments also show that PCA is more sensitive to the choice of reduced dimensions. After reaching a peak, the various classifier accuracy increases performing feature selection using Ranking algorithms and Re-Ranking algorithms of dimensions for PCA.