Neighborhood component analysis and reliefF based survival recognition methods for Hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is one of the most commonly seen malignancy type of liver cancer worldwide. In this paper, a model is presented to classify HCC. The proposed model consists of missing feature completion, feature reduction and feature classification phases. In the missing feature completion phase, a supervised algorithm is presented and this algorithm uses average values. To reduce features, neighborhood component analysis (NCA) and reliefF are utilized as weight generator and features are reduced by using a weight based feature reduction algorithm. By using these weight generators, two novel methods are presented and these are called as NCA based method and reliefF based method. In the classification phase, all of the classifiers of the MATLAB classification learner toolbox are used to obtain a comprehensively analysis. These classifiers are conventional machine learning methods. To evaluate the proposed method, a public dataset was used. Accuracy, recall, precision and F1-score parameters were utilized as performance metrics. The best accuracies for NCA and reliefF based methods were calculated as 92.12% and 83.03% respectively. The proposed methods were also compared to the other state-of-art methods. The comparatively results illustrated that NCA based method has the best success rate among all of them. These results clearly indicated the effectiveness of the proposed methods.