The Impact And Prediction Of Mega-sporting Events On Crime Risk ——Taking The Beijing Summer Olympics As An Example

With the continuous improvement of Chinese comprehensive national strength and international influence,more and more mege-sporting events are held in China,which has brought great challenges to the security of mega-sporting events.mega-sporting events have a significant impact on the crime risk of cities,and the existing research mostly analyzes the overall impact of mega-sporting events on crime risk from the time dimension,and lacks refined analysis and accurate prediction from the time and space level.This study takes the Beijing 2008 Summer Olympics（including the Olympic and Paralympic Games）as an example to study the impact of mega-sporting events on the spatiotemporal risk of crime,and to carry out accurate predictions of the spatiotemporal risk of crime during mega-sporting events.First,based on multiple regression analysis,this study researches the impact of the Olympic Games on the crime risk in different types of areas such as the whole city of Beijing,the Olympic area and the non-Olympic area.We find that the Olympic Games had a significant inhibitory effect on various crime risks in Beijing,among which robbery-type crimes had a greater impact,followed by burglary,car theft,and bicycle theft crimes,and minimal impact on fraud.Secondly,according to the deployment of the security work of the Olympic Games,this study constructs a spatiotemporal correlation analysis model based on Poisson regression,and then analyzes the related factors of crime risk in the core area of the Olympic Games during different security stages.We found that the correlation between various types of crime risk and space facilities in the core area varied with security intensity: before the Olympic Games,burglary and fraud crimes were strongly related to medical education,living services,and government agencies and social group space facilities;during the Olympic Games,bicycle theft was significantly correlated with catering,shopping,leisure and entertainment,as well as government agencies and social organizations;and after the Olympic Games,the correlation between auto theft and life service factors is more significant.Finally,in view of the demand for crime risk prediction during mega-sporting events,this study proposes a crime time series data augmentation method based on time series decomposition,and uses this method to construct a Catboost model combined with data augmentation（DACatboost）.This study conducts a prediction effect test on the crime spatiotemporal risk of the2008 Beijing Summer Olympics,the 2015 Toronto Pan American Games,and the 2019 Wuhan Military Games.The results show that compared with traditional methods such as LSTM,Random Forest,and XGBoost,the proposed DA-Catboost model has good prediction performance and generalization ability,and can accurately predict and analyze the crime risk of mega-sporting events at home and abroad.This study has a certain reference significance for social security and stability maintenance during mega-sporting events,and can provide decision support and reference for the security work of public security departments.