Intelligent Decision Support for Military Operations under Weather Uncertainty

Tan Jing Xian

Abstract

The weather, an important factor which affects almost every outdoor activity on Earth, is often undermined in military situational assessment, perhaps because there are other variables deemed even more crucial. Weather's high level of uncertainty is best modeled using the Bayesian network (BN) approach, which is renowned for its ability to illustrate situations where information is incomplete.

However, most BNs are static and therefore unsuitable to conduct decision-making in dynamic, real-world situations such as fast-paced military operations. Therefore, a dynamic Decision Support System (DSS) is proposed. A special feature of its framework is the organization of domain knowledge in bigger generic blocks where related variables and their interrelationships are grouped together and represented by a BN, known as a fragment. These fragments contain data that are already pre-processed and ready to be used, understood from relevant historical weather data.

In this thesis, Singapore's weather is illustrated in a BN, based on weather data and information obtained from National Environmental Agency (NEA) of Singapore. Two fictitious military examples are cited to demonstrate the use of this proposed DSS. When required, the DSS extracts relevant BN fragments and connect them via common variable nodes to form an Influence Diagram (ID). Together with the user's input, the ID is compiled to obtain the optimal decision with the highest utility value. Run-time to obtain results is greatly decreased when compared to a system which carries out computation from raw data from scratch only when required.

Besides its usage on military operations, the weather BN and proposed DSS framework can be applied to areas where weather elements are crucial in determining the success and outcome. They have the potential to expand and include more depth of highly complicated problems.