Software using random forest for risk prediction of heart valve surgery patients

Abstract

CASPeR - Cardiac surgery prediction tool for risk stratification of heart valve surgeries is presented. The base builds a machine learning pipeline for training a random forest classifier which predicts the mortality after a certain amount of days after the surgery was performed. The classifier also offers a list of potential risk factors through its in build feature selection. With a survival analysis the groups "high-risk" and "low-risk" are compared with each other to check for statistical difference. The tool uses "Shiny" a R package which offers a web frame work to develop data analysis visualizations for the User Interface. CASpeR is delivered as a Microsoft Windows standalone desktop application, that comes with a .exe installer and a detailed manual.