Hepatocellular carcinoma (HCC) is one of the most common cancers worldwide, accounting for an estimated 600,000 deaths annually. It is estimated to have around 17,000 deaths by HCC in 2009 and expected to double over the next 10 to 20 years. Although recent advance of microarray technologies significantly improve our capability to screen tens of thousands of genes and identify valuable signatures, the use of this technology in clinics is hampered by complexity of prediction models applied to the data. Thus, in current proposal, we seek to develop and validate easy-to-use prediction model with the use of low-cost and easily accessible technology in clinics. The robustness of prediction model will be validated throughout series of experiments. Our method will generate continuous values (risk score) that are positively associated with high risk of HCC recurrence after treatment. Unlike many other prediction models that subdivide patients into two or three subgroups by applying pre-fixed cut-off criteria during prediction, our approach will employ more flexible cut-off during or after generating risk score since risk score of each patients is independent of other patients.