| With the continuous expansion of market scale and user groups of large Internet enterprises,the amount of user data is also increasing exponentially.Every large Internet company,after mastering such a number of users,makes the advertising industry become a key link in the company's profitability process.Nowadays,Internet technology has penetrated into all walks of life,how to use big data technology to maximize the expected effect and return on investment of the advertiser is one of the first problems to be solved by the Internet companies.Only the company with the best effect in the limited budget can grasp the customer resources and gain more profit.Therefore,every large Internet company expects to establish a big data management tool to manage the data of platform users and carry out accurate advertisingThe main problems faced by advertisers in the advertising industry today are as follows:(1)They need to be more in line with the actual business needs of advertisers and enterprises to put advertisements to target group users,that is to say,the user groups that advertisers expect to put are screened out from the mass users.(2)The management of the people who put advertisements is chaotic,and it is difficult to manage the historical people effectively.In the process of crowd computing,it is impossible to grasp the real-time status,which affects the subsequent delivery process.(3)In the process of advertising,we can't know the characteristics of the user group,so we can't make the corresponding response strategies and achieve better advertising effectivenessThis thesis first describes the research background and significance of the use of big data management platform for advertising in the advertising industry,and then explores the significance of crowd computing and user portrait for accurate advertising from the demand of the advertising industry for the big data management platform.Then it introduces the technology selection,technology architecture and network architecture used by the platform.Finally,the key functions of each module of the platform are discussed in detail.Combined with the architecture and detailed design of the platform,a big data management platform for advertising,is implemented.The system can be divided into four parts:user management,crowd management,crowd computing and portrait computing.The user management part can be divided into Shiro authentication process module,permission setting module and token acquisition module of authentication server.The crowd management module is divided into structured directory management,real-time state modification of crowd and portrait based on websocket,and theme crowd management.Population calculation includes keyword population calculation based on reading articles,APP population calculation based on usage records and population based on location services.According to the geographical location,keywords or app provided by the advertiser's customers,relevant users can be extracted from the TB level user data stored in the Hive table through spark,compress and store it for portrait computing.User portrait can be divided into calculation of user profile data,management of result label data.User portrait can be generated by comparing the user information of the crowd package with the user data of the label in RocksDB,including genderage,geographical distribution,mobile phone system,and show the user portraits to advertisers and marketing personnel,so that advertisers can analyze the characteristics and behaviors of users,adjust and customize the strategies in time.Then it puts the ads more accurately to the interested usersThe user data involved in this platform mainly comes from the internal data of the enterprise and the data of the third-party platform.At present,it has been officially launched and operated well,and has undertaken more than 90%of the company's advertising business.It also has a certain connection with the advertising platform of other departments,and has become an indispensable part of the advertising business. |
PDF Full Text Request