Fire Loss Assessment Model Based on Internet Search Engine Query Data

There is an interaction between public fire safety awareness and fire direct economic loss, while how to quantify public fire safety awareness is still a challenge. Driven by the rapid development of Internet networks and information technology, Internet search engine query data provides an opportunity for quantifying the level of public fire safety awareness. Based on the fire-related Baidu Search Volume (BSV) data, this paper investigates the relationship between public fire safety awareness and fire direct economic loss with an integration machine learning method. Firstly, the fire-related keywords selection framework is constructed based on the analysis of the fire safety demand generated at different responding stages of fire accidents. Then, the most important keywords associated with fire direct economic losses are identified by correlation analysis. Finally, four types of assessing models, including multiple regression model, principal component analysis model, vector auto-regression model, and artificial neural network model are developed to simulate the ability of fire-related keywords to assess fire direct economic losses. The results show that the fire-related keywords BSV data have a strong correlation with the fire direct economic loss. Meanwhile, BSV data presents a good applicability in evaluating the fire direct economic loss. The artificial neural network model presents the best assessment performance in this work. This research could contribute to quantifying the public awareness of fire safety, indirectly present the public fire safety behavior, and also provide a new solution for how to assess urban fire risk.