Survival Information System Using ReliefF Feature Selection and Backpropagation in Hepatocellular Carcinoma Disease

Abstract

This study was conducted to classify patients with hepatocellular carcinoma by dividing the dataset features. The purpose of this study was to select the features in medical records with the influential features on hepatocellular carcinoma. In this paper, the proposed model is the use of ReliefF at the feature selection stage and Backpropagation at the classification stage. At this feature selection stage, there are two steps, namely weight calculation and feature reduction process. In the feature weight calculation step, each feature is given a weight and the resulting features will be processed in the feature reduction process. The results of the feature weight calculation will produce a ranking from the highest value to the lowest value which will then be reduced by the feature ranking. The best features that have been produced will be used as input for the second stage, namely the classification stage using Backpropagation. The results showed that the 10 features used were the best features out of 39 features. The best accuracy is produced by the ReliefF+BPNN method of 85%. The comparison results show that the ReliefF feature selection method has the best success rate among all. The results of this study indicate that the proposed method is successful.