Detection and recognition of crisis using Markov models and Case-based reasoning
This project pertains to the modeling, detection, and monitoring of crises in geopolitical dynamic environments. As risks are inherent to crises, we need tools to cope with the uncertainty factors involved in these situations. The objective of this project is to conduct research activities to support the understanding of crisis situations and to model their potential evolution. Our main goal is to explore how to find patterns from episodes of conflict that can be reused as templates by human operators in their analysis process. The technologies targeted for this project are Markov models and Case-Based reasoning (CBR). Markov models are efficient tools for modeling sequences of events as they capture the uncertainty inherent to a process. Various extensions, such as Hidden Markov models (HMM), are available for modeling complex stochastic processes with partial information. CBR is a framework, originating from artificial intelligence, which can be used to find similarities in situations and to help the interpretation of these situations. The CBR community has proposed various algorithms for acquiring cases from previous experiences and for exploiting them when facing new situations. We will study how each of these two technologies can contribute to the modeling of crisis situations. We will adapt some of these algorithms for complex situations described by sequential and uncertain events. And we aim to propose hybrid approaches for finding similar sequences of events from geopolitical conflict data sets currently available on the Web.