Research On Urban Road Police Deployment And Patrol Methods Based On High-Incidence Traffic Accident Early Warning

Under the background of Digital China,borrowing technology to supplement police force has gradually become an important means to solve the shortage of police resources.Promoting the transformation of traffic service management from a traditional mode mainly based on experience judgment to a data-driven and judgmentled modern mode is an important approach to achieving technological empowerment.This article is based on traffic accident data from Yinzhou District,Ningbo City.Following the approach of "identification-warning-patrol," it studied the city’s road police deployment and patrol method based on high-incidence traffic accident early warning:Firstly,starting from the collection of accident data,in-depth analysis of the spatiotemporal distribution characteristics of traffic accidents in the jurisdictional area was carried out through methods such as time series analysis and spatial statistics.It was found that there is a strong periodicity and spatial clustering of traffic accidents in the Yinzhou District,which provided a basis for subsequent predictions.The adaptive k-means clustering algorithm was used to analyze the spatial features of traffic accidents,and 30 high-risk black spots were identified based on historical accident data collected from April 1,2020 to September 9,2021.Secondly,in order to better judge the changing trend of black spot traffic accidents,a Bi LSTM neural network model was established to predict the daily average number of traffic accidents in black spot based on accident time series and features such as weather,traffic flow,and accident environment.The results showed that the model’s prediction accuracy reached 93.1%,with a mean square error of 0.092,which had better accuracy and robustness compared to similar methods.Based on the accident prediction results,a high-risk traffic accident warning mechanism was established for each black spot.Finally,based on the improvement of the efficiency of the daily patrol work of traffic police,a black spot area police deployment and patrol model was established based on indicators such as the early warning level,coverage rate,arrival rate,and patrol intensity of each black spot.A two-stage police deployment and patrol optimization algorithm was proposed to solve the model: in the first stage,the differential evolution particle swarm optimization algorithm was used to solve the initial police deployment position;in the second stage,the improved Q-learning algorithm was used to plan the police patrol path.Under different work scenarios,police deployment and patrol plans were provided for three different police coverage requirements of 70%,80%,and 90%.The results showed that the minimum number of police motorcycles required under the three coverage requirements were 5,7,and 8,respectively;the average patrol length was 10.56 km,7.20 km,and 6.75 km,respectively.The proposed two-stage algorithm improved the solution speed by 51.5% and 69.6%compared to the traditional algorithm,and had better solution results,providing theoretical basis and practical reference for the intelligent and digital transformation of police force service management.