Modelling and Analysis Method using BEST

Jointly presented by The Defence Science & Technology Group – A. Coutts, M.-T. Nguyen, J. Knoll, T. Heseltine, K. Owen, M. Van Der Merwe & D. Blumson & The University of Melbourne – I. Widjaja, R. Susanto & J. Jayaputera
Thursday, 23rd January 2025

Bayesian networks (BNs) have been used to analyse Defence capabilities and concepts for supporting strategic planning, evaluation, and force design. In collaboration with the University of Melbourne (UoM), Defence Science & Technology Group (DSTG) developed the Bayesian Elicitation Support Tool (BEST)* for creating model structures and capturing the required data to generate model parameters. BEST automates data collection by organising the data, facilitating data generation and distributing surveys to multiple subject matter experts (SMEs) via email to provide timely responses.

As BEST allows for distributed, asynchronous survey completion, it is a critical tool for eliciting data from different SMEs in large, complex and multi-domain Defence problems. BEST also provides the flexibility to allocate subsections of the model to match participants’ expertise. In particular, BEST employs various elicitation techniques to significantly reduce the size of data sets and minimise bias by aggregating the
opinions of multiple SMEs. Various built-in analyses (what-if, sensitivity, optimisation design/strategy exploration, cause-effect tracing, etc.) are available in our single web-based tool.

Within BEST we have recently been implementing a Multi-Criteria Decision Analysis (MCDA) method, namely the Analytic Hierarchy Process (AHP). By utilising both BN and MCDA methodologies, we present a hybrid decision-making model for evaluating option ranking and selection problems. We also combine the BN modelling approach with simulation and propose a stochastic model for quantitatively analysing the uncertainty associated with eliciting data from SMEs. The proposed approach provides BN models to deal with the uncertainty inherent in data collection and gives the flexibility to decision makers in option comparison and sensitivity analysis. The public version of BEST can be accessed through the University of Melbourne server