Design And Implementation Of User Interest Discovery System Based On Interaction

With the rapid development of mobile Internet technology,the number of users of mobile applications has also increased in an exponential state.Mobile Internet users are increasingly pursuing personalization in the process of using products,and many companies are committed to research on targeted delivery.The essence of personalization theory is to screen and integrate information resources based on the user's interests and preferences,and provide users with services that meet their interests and preferences.To realize the user's personalized service,the core is interest discovery.Mobile users generate a lot of data in the process of using the application.Most of these data have the value of deep mining.The traditional user interest analysis is particularly thin,and it can no longer support the increasingly hidden interest discovery.Web-based interest mining has been relatively mature,but the research on interest mining of mobile Internet users has not drawn attention.Under this background,this paper analyzes the current research methods of interest discovery and the related data analysis platforms at home and abroad.The status quo of mobile products proposes the design and implementation of an interactive interest discovery system based on mobile users.This article first establishes the research method of the system through the general method of analogous web-based data mining,and then optimizes the key technologies and methods for solving problems.In the realization of interest discovery,the system divides the user's interaction behavior into five categories:The user selects,forwards,reads,and reviews the information content using the improved TF-IDF algorithm to obtain the interest keyword,and uses the statistical analysis method to obtain the user's interest keyword for the focused type of interaction,and then separately addresses the two sets of keys.Words establish a relation graph,use TextRank algorithm to obtain the user's interest set,and finally use the harmonic factors to fuse the interest sets obtained by the two methods,and ultimately obtain the user's interest.Finally,this study uses a real mobile feed flow product as an experimental object to verify the accuracy of the system interest discovery algorithm,thus proving that the research of this system has a certain application value.In addition to interest discovery,the system also implements some basic statistical analysis modules such as user analysis,traffic statistics,channel analysis,retention analysis,and conversion analysis,and strives to provide users of the system with a data service platform with interest discovery as the core and full-featured functions.On the basis of selecting better key technologies and methods of system realization,the article introduce system's holistic designation and implementation:employ the structure of front-and-back end separation,whose interfaces interact in RESTful style;use Nginx for forward and reverse proxy and load balance;in consideration of system's huge concurrency in gathering user interaction data,and in order to mitigate data peak effectively and avoid the congestion of data collection flow which cause data lose,we use producer-and-consumer block array base on Zookeeper to achieve data cache;adopt Mongo to avoid data lose caused by database downtime;Mongo use ETL to load the data transformed from non-structure to structure into database.