The use of information systems artefacts and expertise in a study of prediction in the strategic management of house fires
House fires have posed a threat to life and property in every society for millennia, such that laws, organisations and work systems have been established to protect communities. Brave fire fighters need to attend instances of house fires within seven minutes in New South Wales, Australia in order to minimise damage (Smith et al, 2016). Statistics published annually by the Australian Productivity Commission (2021) show little change in the fatalities, injuries or costs associated with house fires for the last 10 years. Research shows house fires kill more people annually in Australia than any other natural disaster (Edwards 2017, Gissing 2019). The question was thus posed - what if we could predict the likelihood of when and where house fires occur in the future? Common features in fire data were identified enabling two models to be developed, using different features to predict the likelihood of house fires. One model uses fuzzy logic, whilst the second was developed using Machine Learning techniques (Shmueli & Koppious, 2011, Padmanabhan et al, 2022). Using Information Systems expertise, the two models were integrated into an artifact called the M2inder Computer Application incorporating the principles of Design Science (Henver et al 2004, Dreschler, 2015), and tested against real world data for the Australian Winter of 2021. The results from the machine learning model returned an r2 value of 0.774202 and when compared against real world data had an accuracy rate of 55%. 69% of survey respondents deemed this to be acceptable in this research domain. Examined through an Activity Theory lens (Vygotsky 1978, Leontiev 1981, Engeström’s 1987) the complexity of decision-making in-house fire prevention, preparation and response is aligned with Situational Awareness Theory (Endsley, 1995), creating a generalisable problem-solving research tool - Situationally Aligned Activity Theory.
