Modeling Data-driven Liver Cancer Prediction with Medical Knowledge on Chinese Population

According to the report in 2018, liver cancer is the top ten of malignant tumors. Although various techniques were developed to assess the liver cancer risk, but there is no specific model targets to the characteristics of Chinese population. In this study, we collected a total of 49,076 Chinese patients with cancer risk using questionnaires, which are designed by the domain experts. We extracted 14 features of liver cancer from the raw 107 questions under the instruction, and compared six competitive methods for risk prediction. The experimental results indicated the Support Vector Machine (SVM) outperform other methods of medical knowledge. By leveraging the combination of the appropriate features and methods, it is found that our model is efficient in modeling liver cancer risk prediction on Chinese population.