Stadium fires can easily cause massive casualties and property damage. The early risk prediction of stadiums will be able to reduce the incidence of fires by making corresponding fire safety management and decision making in an early and targeted manner. In the field of building fires, some studies apply data mining techniques and machine learning algorithms to the collected risk hazard data for fire risk prediction. However, most of these studies use all attributes in the dataset, which may degrade the performance of predictive models due to data redundancy. Furthermore, machine learning algorithms are numerous and applied to fewer stadium fires, and it is crucial to explore models suitable for predicting stadium fire risk. The purpose of this study was to identify salient features to build a model for predicting stadium fire risk predictions. In this study, we designed an index attribute threshold interval to classify and quantify different fire risk data. We then used Gradient Boosting-Recursive Feature Elimination (GB-RFE) and Pearson correlation analysis to perform efficient feature selection on risk feature attributes to find the most informative salient feature subsets. Two cross-validation strategies were employed to address the dataset imbalance problem. Using the smart stadium fire risk data set provided by the Wuhan Emergency Rescue Detachment, the optimal prediction model was obtained based on the identified significant features and six machine learning methods of 12 combination forms, and full features were input as an experimental comparison study. Five performance evaluation metrics were used to evaluate and compare the combined models. Results show that the best performing model had an F1 score of 81.9% and an accuracy of 93.2%. Meanwhile, by introducing a precision-recall curve to explain the actual classification performance of each model, AdaBoost achieves the highest Auprc score (0.78), followed by SVM (0.77), which reveals more stable performance under such imbalanced data.

Machine learning methods can establish complex nonlinear relationships between input and response variables for stadium fire risk assessment. However, the output of machine learning models is considered very difficult due to their complex “black box” structure, which hinders their application in stadium fire risk assessment. The SHapley Additive exPlanations (SHAP) method makes a local approximation to the predictions of any regression or classification model so as to be faithful and interpretable, and assigns significant values (SHAP value) to each input variable for a given prediction. In this study, we designed an indicator attribute threshold interval to classify and quantify different fire risk category data, and then used a random forest model combined with SHAP strategy in order to establish a stadium fire risk assessment model. The main objective is to analyze the impact analysis of each risk characteristic on four different risk assessment models, so as to find the complex nonlinear relationship between risk characteristics and stadium fire risk. This helps managers to be able to make appropriate fire safety management and smart decisions before an incident occurs and in a targeted manner to reduce the incidence of fires. The experimental results show that the established interpretable random forest model provides 83% accuracy, 86% precision, and 85% recall for the stadium fire risk test dataset. The study also shows that the low level of data makes it difficult to identify the range of decision boundaries for Critical mode and Hazardous mode.