Wide area monitoring protection and control system (WAMPAC), monitor and control the grid dynamics in real time. Availability of PMU data in WAMPAC opened the door for data centric modeling. This paper proposes a novel data centric model for blackout risk analysis. Analysis is based on the Kullback-Leibler divergence (KLD) or relative entropy between two data samples. The key contribution of this paper is probabilistic analysis of transmission line data to capture the power flow vulnerability in the cascade failure and early prediction of probable blackout based on the relative entropy between normal and the perturbed power flow data. For blackout prediction the reference KLD threshold is calculated from the past blackout events and used as a precursor for blackout early warning signal. Early prediction of blackout risk may prevent the power grid against massive blackouts in future. The proposed methodology is validated on the IEEE 30 bus prototype power system model.