Research On Community Security Risk Prediction Model Based On Deep Learning

Community is a social community composed of several social groups or people living in a certain area.So the community policing is stable or not is a mirror of the local reaction to the social order is stable,community security directly affects(reflection)the whole social security level,so can the timely and effective assessment community policing risk degrees,risk of community policing in risk assessment,improve community policing stability has important significance.At present,there is a serious shortage of grassroots police force,and many public security work is in a passive state.In order to better maintain the stability of community security,accurate prediction of community security risks is carried out to provide decision-making support for the optimization of police allocation and improve the rapid response ability of public security organs to crack down on illegal crimes.Based on the current community security situation,this thesis analyzes and extracts the key factors affecting the community security risk,and excavates the relationship between each factor,so as to accurately predict the community security cases,and finally make an assessment and early warning of the community security risk degree.The main contents of this thesis are as follows:(1)based on the relevant factor analysis theory,select the characteristic factors affecting the community security risk.By analyzing and preprocessing the collected data,this thesis USES correlation analysis to preliminarily screen out the time series characteristic factors that affect community security risk,and eliminates the characteristic factors that have little or no change in the short time,thus laying a foundation for the construction of community security case prediction model.(2)based on the multi-channel convolutional neural network,the prediction model of community security cases was proposed.The prediction model improves the traditional convolutional neural network and changes from a single channel network to a multi-channel network to adapt to the demand of multiple characteristic factors input at the same time.The improved network enlarges the sensing field of the convolutional neural network and improves the accuracy of the prediction model.The experimental results show that the prediction accuracy of multichannel convolutional neural network can reach 87.7%.(3)according to the difference of prediction performance of different models in different tasks,a comprehensive prediction model based on neural network was proposed.Two single model of comprehensive prediction model are respectively CART regression tree model with multi-channel convolution neural network prediction model,the output value of each single model as the input values of the model,according to different task learning performance is good or bad to assign weights,eventually return based on the predictive results of the input,namely for the community public security case prediction results.The experimental results show that the prediction accuracy of the comprehensive prediction model can reach 92.5%.In this thesis,two single models,CART regression tree model and multi-channel convolutional neural network model,are respectively constructed for community security case prediction.The prediction accuracy of CART regression tree model and multi-channel convolutional neural network model are above 80%,which have better prediction performance.However,compared with the comprehensive prediction model proposed in this thesis,the comprehensive prediction model has higher prediction accuracy and robustness.