| Abstract: |
Simulation plays a key role in the workforce analysis and planning capability of the Australian Defence Force (ADF). This is done using Athena, an advanced workforce simulation tool suite, specifically developed to provide the ADF with a capability that can accurately model and simulate the complexity of the Defence workforce, with its hierarchical nature and highly interconnected structure. Athena Lite, a Discrete Event Simulation engine that simulates the progression of personnel through their careers, including postings, promotions and loss of personnel through attrition, while respecting workforce-specific prerequisites and conditions, is a part of this capability. It is a web-based, scalable decision support tool currently in use by ADF workforce planners.
Athena Lite gives ADF users the ability to forecast their workforce, run ‘what-if’ scenarios, and find optimal recruitment and promotion policies to meet demand. However, in Defence workforces there is a high level of uncertainty – unexpected attrition, recruitment, and workforce unavailability, as well as new platforms and capabilities, are just some examples of this. With this high level of uncertainty, forecasting individual scenarios, or manually running multiple ‘what-if’ scenarios becomes time-consuming and places a high demand on the analyst to determine the ‘cause-and-effect’ of input changes on the large number of outputs.
To meet this challenge, IRIS was created, a web-based data analytics toolbox designed to interface directly with Athena Lite. IRIS is based on a previous prototype outlined in (Mortimer, et al., 2021). It has two main goals: to sample a large input space and run Athena Lite many times with this input space, and to use data analytics to explore the created dataset. It automatically explores the health of a workforce, provides insights into the sensitivity of various parameters and outputs, and examines the nth order interdependencies and effects of inputs on outputs, as well as outputs on other outputs.
To do this, users build their model in Athena Lite and enter input parameter ranges in the IRIS interface, before running the simulator up to 2000 times, in each run sampling the input parameters from between the user-input ranges. Three visualisations are then created to allow users to explore the dataset. In an interactive time-series plot, the user is able to explore the performance of units or ranks and find points that were particularly vulnerable. A correlation heatmap, created using Spearman’s correlation coefficient, displays the dependencies between simulation inputs and outputs, and an interactive Bayesian network gives users the ability to explore the effect of the dependencies. The user can explore the interactive network through mouse hovers and clicks, and make queries to see the direct effect of the relationships between parameters.
IRIS is currently being used by ADF workforce planners. However, IRIS development continues, to increase the available analysis techniques and visualisations, with inclusion of techniques such as decision trees and principal component analysis. A planned future feature of IRIS involves further automation of the analysis, using natural language generation to create reports that provide plain English explanations summarising and highlighting key results.
Mortimer, K., Nguyen, V., Caelli, T. & Hill, B., 2021. Simulation and Data Analytics for a Defence Workforce Transition. Sydney, NSW, MSSANZ, pp. 92-98. |