Fire risk in the context of social development and government control: Evidence from 10 years of multivariate statistics in China

A study of fire risk from fire statistics can provide a global-oriented view for regional fire risk mitigation. Using the fire statistics from 2010 to 2019 in China, a fire risk matrix-based framework was constructed to reflect the regional fire risk level comprehensively, which combined the frequency with the consequence of fires. Under this framework, the regional disparity was observed, and most regions in China are at a medium level of fire risk overall. Furthermore, the regional socio-economic and governmental data were integrated to identify the positive and negative factors that may influence fire risk. The correlation between socio-economic development factors, risk management factors and fire risk were explored, and four statistically significant indicators were identified: population size (−0.20, p < 0.001), per capita GDP (0.35, p < 0.001), income level (0.26, p < 0.001), and employed persons in state-owned agencies and organizations (−0.15, p < 0.01). Based on these four correlation indicators, three typical machine learning methods, namely K-Nearest Neighbors (KNN), Support Vector Machine (SVM) and Random Forest (RF), the explanatory models for regional fire risk were constructed and the results were validated based on statistical data. The experimental results show that the fire risk can be explained to a certain extent based on the four indicators with correlation, with classification accuracies of 47.31 % (KNN), 53.76 % (SVM) and 54.84 % (RF) on the test set.