Risk Analysis in Big Data

Society has become increasingly interconnected with networks linking people, the environment, information, and technology. This complexity is a challenge for risk analysis. Traditional risk analysis methods typically underestimate the probability and impact of risks (e.g., terrorist attacks, power failures, and natural disasters such as hurricanes) because normal data and independent observations are assumed. Cascading failures occur in complex systems, such as a hurricane that causes a power failure resulting in flooding. Big data offers enormous promise for improving risk analysis, but cannot replace the importance of appropriate assumptions, data quality, continued validation, and human understanding of risk analysis for effective risk mitigation. The blessing and curse of big data for risk analysis is illustrated by the example of Google Flu Trends, which was accurate at first but then became inaccurate