Dimensionality Reduction Based Diabetes Detection Using Feature Selection and Machine Learning Architectures

Abstract

One of the illnesses that is sweeping the globe like an epidemic is diabetes. Every generation, including children, teenagers, young adults, and elderly, is seen to be affected by it. The study raised difficulties regarding the necessity to establish a connection between the primary causes of diabetes development. This paper compares pre-processing accuracies of various dimensionality reduction models. Here the proposed technique use two datasets, one is normal diabetic dataset and another is heart disease dataset. Both dataset has been pre-processed using dimensionality reduction (DR). In proposed work, DR process is divided into two stages: unsupervised DR and supervised DR. Prior to processing, improved DR unsupervised principle component analysis was performed. The two datasets were then combined.