Research On A Personalization Content Push System Based On User Behaviors

With the applications of "Triple play" technology becoming more and more populartoday, the number and series of the set-top box increases rapidly. Through the connectionof set-top boxes and TV set, users can enjoy rapid growth of content resources. However,accessing to the content they are interested accurately is a challenge. The research ofpersonalized push system is conducive to solve this challenge.To get the users’watching behavior and interest information, the system pre-definedthe content and format of the data which will be reported by all the users. The systemimplicitly records the user watching behavior data in the entire process from the start toshutting down. Based on those data, user interest database will be established whichplays a great role in preprocessing of the data and reducing the noise data.Building user interest model and improving the push technology is the two corecontent of the personalized push system. According to a thorough study of currentmainstream user interest model-building technology, a improved interest model will beestablished which is called space vector interest model based on the time factor (TSVM).The model which has a three-tier tree structure can be able to calculate the user dynamicinterest and to take into account the long-term interests and short-term interest of the user,to simple user interest model updates at the same time. Finally, using a hybrid pushtechnology based on users’ interests change, can make up for high-dimensional sparseddata and over fitting which traditional push technology bring about.Making use of the user watching behavior data collected by Routon ElectronicGrand View Garden management platform, the comparing experiment verifies the spacevector user interest model based on the time factor in accuracy as well as the correlationcoefficient is superior to the traditional user interest model; besides the hybrid pushtechnology based on users’ interest change improves the MAE indicators. Thepersonalization content push system can present the interested topic to users bypersonalized program interface.