Association Rule Mining for Accidental Dwelling Fire Incidence Data Analysis in Greater Manchester, UK

ABSTRACT Association rule mining was used to identify patterns in accidental dwelling fire incidents attended by Greater Manchester Fire and Rescue Service over the period 2013/14 to 2023/24. The association rule mining process identified relationships between cooking fire incidence, distraction, living alone, deprivation, and fire injury. The association rule mining was undertaken using the Python programming language utilizing the apriori algorithm to find frequent combinations of accidental fire circumstances, and the association rules function to find meaningful associations in terms of which combinations of circumstances are most strongly associated with specific fire outcomes. The parameters of the apriori and association rules functions were adjusted in order to produce the most informative and useful set of association rules from the accidental dwelling fire incident dataset. The strength of the relationships was assessed using three key metrics: support, which quantifies the frequency of a rule's occurrence within the dataset; confidence, which determines the proportion of instances where the rule holds true; and lift, which indicates the extent to which the rule's occurrence exceeds what would be expected by chance. Distraction appeared to substantially increase the risk of cooking fires (60% more likely than chance), and distraction appeared to be a more common fire incidence factor in deprived areas. This extends previous research into accidental dwelling fire incidence in relation to deprivation and cooking fires by providing the probabilities associated with co‐occurrence of different fire risk factors.