High-risk Factor Mining with Gray Relation Analysis for Pulmonary Infection after Renal Transplantation

Pulmonary infection is often one of the serious complications after renal transplantation. In serious cases, it could even endanger the survival of patients. Therefore, infection risk prediction has always been a highly-concerned research topic before and after transplantation operation. Aiming to explore an efficient risk prediction of pulmonary infection, this paper proposes a data-driven scheme to mine a group of optimal risk factor combination. Firstly, it pre-processes and cleans clinical patient data. After that, it reduces sample dimension and computes the comprehensive influence degrees of patient cases by Primary Component Analysis (PCA). Then, a Gray Relation Analysis (GRA) approach is adopted to generate the correlation degree to pulmonary infection happening or not. Finally, a fully-connected neural network (FCNN) with multi-layer structure is designed to determine optimal risk factors, via searching the highest infection prediction accuracy in given patient samples. The experiments with performance comparisons verify that our risk factor mining method is effective, obviously outperforming the other ones having been compared on given clinic patent cases.