Research On Technologies Of Heterogeneous Data Processing And Analysis For Public Security Business

With the development of policing informatization and digitization,a large amount of data resources have been accumulated in the public security policing system.In the era of big data,the traditional manual analysis and experience-based policing model has been unable to adapt to the current data volume level.The establishment of intelligent policing based on intelligence data and artificial intelligence and other technologies has become an important practice in the new era.The construction of smart policing requires efficient data mining techniques to leverage the value of data resources.At present,the process of building smart policing faces the problems of low efficiency methods of network intelligence data processing,difficulties in abnormal detection,and delays in early warning mechanisms.To a large extent,it limits the development of smart policing and can not effectively exert the value of massive data resources.This paper conducts in-depth research and analysis of issues that arised during the construction of smart policing,and proposes the effective solutions.Firstly,a detailed analysis of the problems existing in the network intelligence data processing method during the construction of smart policing has been conducted.After that,in order to improve the efficiency of network intelligence data extraction in the public security department,an automatic template generation algorithm is proposed.The extraction templates are generated automatically based on the structure of the web page data and the repeated pattern information.The content extraction is completed at a lower time consumption and the accuracy rate exceeds 92%.Secondly,aiming at the difficulty of anomaly detection in the complex transaction relationship network during the construction of smart policing,this paper adopts deep learning to build the detection model,and proposes a two-way heterogeneous convolutional neural network model which combines transaction data between transaction nodes and local topology relation data in a transaction network.The model can process transaction network data in parallel which provides a scientific basis for the detection of criminal cases.The accuracy of the experimental test results reach 95% and the detection recall rate exceed 97%.Then,focused on the problem of the lagging in the crime early warning mechanism during the process of construction of smart policing,a universal early warning model integrated the personnel data in the policing system is proposed.The model is based on the analysis of policing data and then achieves early warning of high-risk personnel which provides a scientific reference for the deployment of police resources.The average accuracy under unbalanced data sets exceeds 90%.Finally,we design a smart policing prototype system based on the above researches which provides a practical and effective processing and analysis method for the policing system data.The system test results show that the system can achieve the expected design goals.