Fire Risk Assessment Based on SMOTE and Bayesian Network

Abstract

Establishing fire historical data and index system is the key to solving the fire risk assessment problem. Based on this, this paper proposes a new fire risk assessment method. Firstly, aiming at the imbalance problem of fire history data, SMOTE algorithm is used to expand the small class of sample data and build an improved fire sample data set. Secondly, the improved fire data and expert knowledge are used to learn the structure and parameters of the Bayesian network. Finally, the risk assessment value is determined through the probabilistic inference of the Bayesian network. Applying this method to building fire risk assessment can effectively solve the problem of unbalanced fire data and improve the level of fire risk assessment.