Feature Selection for 30-Day Heart Failure Read-mission Prediction Using Clinical Drug Data

Lu S, Ye Y, Tsui FC, Liu X, Hwa R. Feature Selection for 30-Day Heart Failure Readmission Prediction Using Clinical Drug Data. NIPS workshop on Machine Learning for Clinical Data Analysis and Healthcare; 2013 Dec 10.

Large-scale clinical drug data, which can form high dimensional feature space, could potentially hinder the efficiency and performance for machine learning algorithms. Drug ontologies, which provide relationships and similarities between medication concepts, can be used to reduce redundancy by clustering drug data. In this paper, we study the problem of using clinical drug data to predict the probability of 30-Day heart failure readmission. We propose a feature selection method using correlation-based feature measures with tree-structure drug ontology. Experimental results show that our method can reduce the feature space dramatically and improve the performance of heart failure readmission prediction using only drug data.

Publication Year: 
2013
Faculty Author: 
Publication Credits: 
Lu S, Ye Y, Tsui FC, Liu X, Hwa R.
^