Primary Liver Cancer Early Screening Based on Gradient Boosting Decision Tree and Support Vector Machine

Primary liver cancer has no obvious clinical symptoms and no effective screening method in the early stage, which leads that its first diagnosis is generally late. Early detection is considered to be the main measure to improve this situation. An early screening method utilizing clinical laboratory dataset was proposed. Firstly, gradient boosting decision tree (GBDT) was used for feature selection, and then two classification methods, support vector machine (SVM) and GBDT, were used for training and testing. The results show that the Kappa index reaches the almost perfect level and the accuracy is over 90%. This method can assist doctors to screen primary liver cancer early.