Simplified & Novel Predictive Model using Feature Engineering over MIMIC-III Dataset

Abstract

Predictive mode of analysis is one of the preferred techniques in formulating treatment plans for resisting the consequences of critical disease in the healthcare sector. The existing literature review finds machine learning to be dominantly used for this purpose over standard medical dataset with no effective or full-proof predictive models to date. Therefore, the proposed system contributes toward developing a computational model which offers a simplified yet effective series of operations for feature engineering followed by applying a machine learning approach. This methodology leverages the medical dataset concerning its informative features, significantly reducing the computational burden on the learning platform. The experimental analysis is carried out on a standard MIMIC-III dataset to find that the proposed scheme has better accuracy and computational efficiency than existing learning schemes. The outcome of study shows proposed scheme to achieve approximately 95% of accuracy and reduced algorithm processing time of 0.255s when compared to existing frequently adopted learning schemes.