In silicon testing, a Shmoo plot is commonly used to give us an insight into the silicon manufacturing development health. A Shmoo plot is a graphical display of the response of a component or system varying over a range of conditions and inputs, which is a good representative to silicon manufacturing development and silicon heath. Currently, a large amount of Shmoo plots have been generated by for test content robustness. Shmoo plots and other silicon characterization data has high value, however, analysis of them is a time-consuming work. This paper establishes a machine learning based model to improve and automate the procedure in silicon data analysis for HVM test content development. Our experiment shows that the supervised learning model has good accuracy on VMIN estimation across various kinds of Shmoo issues (crack/sprinkle/ceiling). The accuracy attained is greatly improved over previous tools. The framework can be easily integrated into any automated tester software and would save time to market during first silicon characterization. Additionally, the methodology discussed in this work can be extended to the HVM test flow for silicon behavior.