Research And Implementation Of OTT Terminal Identification Technology Based On Operator's Big Data And Deep Learning

With the rapid development of home internet,TV service has entered the multi-terminal and multi-channel large video consumption upgrading cycle with the Internet as the carrier.With the continuous innovation of technology,OTT intelligent terminal has become an important carrier and traffic entry of home Internet,which has great marketing value.Operators,as providers of basic broadband,have unique advantages of home user traffic.Through the terminal dimension,they deeply analyze and study the data of home Internet traffic,understand the behavior preferences of home users,grasp the behavior and traffic characteristics of users in home Internet,optimize and rationally allocate network resources and optimize users' online experience,etc.Provide support,so as to adjust their own video services and content.Accordingly,this paper proposes an OTT terminal identification method based on large data of operators.This method solves the problems of large workload,low efficiency and high error rate in traditional terminal identification methods.For this method,the main contributions and innovations of this paper are as follows:1.This paper uses Hadoop technology,DPI,regular matching technology and distributed crawler technology to identify terminals.DPI technology is used to deeply parse data packets;high-speed regular matching algorithm is used to obtain UA strings;user-defined function UDF in Hive is written to parse UA and get terminal model;and terminal matching program is written through Trie dictionary tree structure to match terminal model and terminal name.The experimental results show that the method can identify terminals more quickly and accurately,and the accuracy of terminal recognition can reach more than 90%,which has greatly improved compared with traditional methods.2.Using the technology of distributed crawler to obtain the detailed information of terminal models on e-commerce websites,and store the captured detailed information,so as to establish terminal matching library,and constantly update and maintain the information of terminal library,so as to improve the success rate of terminal matching.3.Terminal library information is always changing.Manual updating terminal library information is time-consuming,laborious and error-prone.This paper studies and analyses the performance of CNN algorithm in word segmentation recognition,and proposes a deep learning method for automatic updating terminal library.