Missing data imputation using machine learning based methods to improve HCC survival prediction

Hepatocellular carcinoma (HCC) is the most common type of cancer in the liver. Early diagnosis is very important for predicting the course of HCC. In recent years, important research studies have been carried out to help physicians make a decision about survival prediction related to HCC disease with information from clinical data. However, most of these studies do not take into account differences or incomplete data among HCC patients. Recovering data is very important as missing data may adversely affect survival prediction of HCC patients. In this study, 5 different models (median, mode, mean, decision treebased regression and linear regression) were used for the first time in order to estimate the missing clinical characteristics of real complex HCC data. The results were evaluated in two different machine learning algorithms (Naive Bayes, Decision tree based classifier) for survival prediction. Accuracy, specificity, precision, sensitivity and F1 score were chosen as performance criteria. The highest success in HCC survival prediction model was obtained by using decision tree based regression and decision tree based classifier.