The Impact of Data Mining on Prediction of Renal Transplantation Survival and Identifying the Effective Factors on the Transplanted Kidney

Chronic kidney failure is a common disease in the world and kidney transplantation is the most effective treatment in patients with chronic kidney failure. The aim of this study was to predict the survival of transplanted kidney and identify its effective factors, and also to provide a model for higher prediction accuracy.
In this retrospective study, data from 423 cases of kidney transplant patients during 2006-2011 in Afzalipour Teaching Hospital in Kerman were obtained. The neural networks, decision tree and support vector machine were used to predict kidney transplantation survival and information fusion was used to combine the results of these classifiers and design a model with higher prediction accuracy. In addition, for identifying factors affecting the survival of transplanted kidney, genetic algorithm was used and for data analysis and implementation of algorithms, Clementine 12 and Weka 2.3 were used.
The accuracy of neural networks, decision tree, and support vector machine were 94%, 92%, and 92%, respectively, and the accuracy of information fusion was 95.74%. Also, recipient BMI and gender, donor age, compatibility of donor and recipient blood group, and history of kidney transplantation as the effective factors on renal transplantation survival were identified by genetic algorithm. The prediction accuracy of this model was 91.67%.
The results show that information fusion can increase the prediction accuracy. Also, the genetic algorithm as an effective method can be used for identifying the optimal features.