Characterised and personalised predictive-prescriptive analytics using agent-based simulation

Abstract

Organisations rely on data analytics to process vast quantities of consumer data to inform strategy and policy decisions. Descriptive analytics is commonly used to provide insight into past consumer behaviour. However, greater value can be achieved by predicting future consumer behaviour through predictive and prescriptive analytics. Simulation plays a vital role in facilitating predictive-prescriptive analytics. Here we describe the practical use of an agent-based modelling and simulation approach applied to real-world scenarios, and describe its benefits over traditional econometric, statistical, and spreadsheet approaches. This paper describes two novel agent-based consumer modelling approaches and an associated case-study for each in water and energy forecasting respectively. The approaches described in this paper are termed characterised consumer modelling involving modelling individual consumers based on their characteristics/type; and personalised consumer modelling involving modelling specific identified consumers derived from data on an individual consumer to facilitate personalisation of strategies for that consumer. Validation of the models demonstrates a high level of accuracy and functionality, and suggests that an agent-based simulation approach can answer a range of complex consumer problems using minimal consumer demand data.