| In the mass information of the car user community,there are many marketing advertisements and navy information that are difficult to identify.The traditional method in the past is to monitor the operation students with naked eyes,and for those who have obvious intention to irrigate,and advertise for dealers and brands.Posts will be deleted,banned and blocked.The cost of this method is large and the profit is low.Therefore,it is necessary to develop a real-time monitoring system for posting articles to detect various article genres in the community,strengthen the identification and processing of abnormal groups such as dealers and naval forces,and minimize the negative impact on the community content ecology as much as possible.Use the CNN model to predict the text category to improve the work efficiency of operators.The real-time monitoring system for posting documents fully considers low latency and high availability,and implements a separate monitoring system,which can support operations personnel to process user postings in a timely manner,visually configure monitoring rules,and support convolutional neural network pairing.The newly-added user forum posts categorized operations and provides auxiliary suggestions to operators.The monitoring system mainly consists of three parts.The first is the most basic monitoring part,which realizes the counting of users and car circle dimensions,and supports the millisecondlevel split monitoring based on the database binlog,and is decoupled from the business code,even if the follow-up Adding new genres or expanding the monitoring function can be achieved with minimal code changes;the second is to configure the platform,which achieves high availability,even if some nodes are down,it can still provide services to the outside world,and supports visualized operations on web pages.Entering the company’s unified SSO can cope with frequent changes in monitoring rules by R&D personnel and operations personnel without changing the code;third,with the help of deep learning CNN model and Tensor Flow framework,classification model training with historical stock data is finally realized.The real-time recognition of the content of newly issued texts can distinguish the texts of navy,dealers and normal users,reducing the workload of manual recognition by operators.The real-time monitoring system for postings uses key technologies such as binlog technology and CNN text classification algorithm to achieve real-time monitoring of user postings in automotive forums.It can monitor user postings in different genres,and achieves low latency,high availability,and resolution of monitoring rules.Coupling and secure encrypted storage of user data. |
