In 2014, hepatocellular carcinoma (HCC) cancer ranks second in the list of reasons for cancer-oriented deaths over the world. The incidences of hepatocellular cancer have approximately doubled, in the last two decades and the mortality rate due to HCC has also increased. There will be approximately 30,200 liver cancer deaths in 2018. The clinician provides treatments to HCC patients by using evidence-based medicine, which may not effectively resolve the problem of each patient. For assisting the decision making of the clinician, research works have been done in this field to extract information from these clinical data using computational methods. In this paper, a methodology has been proposed for the prediction of hepatocellular carcinoma patient data. A two-phase cluster based feature ranking procedure has been proposed and applied to the pre-processed data. Markov Blanket-based clustering method has been proposed, in which, the redundancy among the features is computed to rank the features. Total 6 different classifiers namely C4.5, ENSEMBLE, ANN, kNN, Naive Bayes, and SVM have been used for evaluation of the proposed methodology in terms of classification accuracy on HCC data by comparing it with some other most common feature selection methods (ReliefF, mRMR, MIM, and FCBF). The better the classification accuracy of the proposed methodology shows its effectiveness for prediction of HCC data.
Select your language of interest to view the total content in your interested language