Research On Mobile Personalized Service System

The aim of the personalized service is how to transfer the right information to the right person. With the development of information science and technology, information has infiltrated into every aspect of people's life. Meanwhile, with the popularity of mobile phones and the development of mobile data service, personalized information services to mobile phone users are inevitable demands. According to the actual demands, the research concern of this paper is personalized service for the mobile application systems.After analysis of the current research status and introduction of application system----Mobile Devices Off-line Reading System, the architecture of personalized information recommendation service based on Mobile Devices Off-line Reading System is provided. The research and analysis works are as followings:1, Users' Behavior Sequence Analysis: The K near neighbors classification method is adjusted so as to trace, log, model and analyze the series of users' actions for news browsing. Then the personalized features are classified and saved, which could be used later in the system to provide the real-time individual information recommendation for each user's access to news. Simulation experiment result has certified the utility and feasibility of the individual recommendation algorithm. The accuracy of the recommendation can reach 70%.2, News Content Classification Analysis: via the text segmentation, feature extraction, and with the history items provided by the User Model, the text classification method based on Bayesian arithmetic is designed to divide the unread news into two classes----interested news and uninterested news. The simulation experiment shows that the accuracy of the recommendation by content classification analysis is higher than by behavior sequence analysis.3, Comprehensive Analysis Module: The advantages and disadvantages are compared between the two recommendation means. Finally, under the actual demand of the application system, the integration arithmetic is proposed to make the tow means work together and output more desirable result sets. At last, the interface for relation rules means is included to improve the extension ability of the recommendation system.