Data-driven satisficing measure and ranking

Risk assessment is the process where we identify hazards, analyze or evaluate the risk associated with that hazard, and determine appropriate ways to eliminate or control the hazard. Risk assessment techniques have been widely applied in many area including quantitative financial engineering (Krokhmal et al. 2002), health and environment study (Zhang and Wang 2012; Van Asselt et al. 2013), transportation science (Liu et al. 2017), etc. Paltrinieri et al. (2014) point out that traditional risk assessment methods are often limited by static, onetime processes performed during the design phase of industrial processes. As such they often use older data or generic data on potential hazards and failure rates of equipment and processes and cannot be easily updated in order to take into account new information, giving a more complete view of the related risks. This failure to account for new information can lead to unrecognized hazards, or misunderstandings about the real probability of their occurrence under current management and safety precautions. With the rapid development of computational intelligence and corresponding decision support system, as well as the launch of "Big data" era, nowadays, new risk assessment technique should allow decision maker to update the assessment results by observing new information or data and realize quick response to dynamic environment. In this paper, we develop a satisficing measure based model to assess, compare and ranking random outcomes, and propose both online and offline data-driven computational frameworks.