Personalized Packages Recommendation Algorithm Based On Feature Augmentation And Cascade Tactics

As the Internet is incorporated into the social economy and daily life constantly, people are becoming accustomed to a variety of online information interactions(such as shopping, social networking, navigation, etc.). So that huge amounts of context data are produced. The data includes many feature information, such as consumption time, location, shopping records, and so on. Then the information implies consumption intention of customers. At the same time, in order to meet the needs of different users, companies often need to design different packages combined with different consumption situation and business elements for enhancing user viscosity. Therefore, it is important to recommend commodities for user personally and intelligently through analyzing massive user consumption data and discovering useful knowledge based on the "consumer-oriented" as the core idea.According to the ideas of the personalized recommendation based on hybrid strategy, this paper proposes a new recommendation algorithm based on Feature Augmentation and Cascade Collaborative Filtering(FACCF). The algorithm can improve accuracy of packages recommendation. And the algorithm combines the time-domain characteristics, spatial characteristics, consumption propensity and package characteristics of consumption data. Firstly, the algorithm will construct CTAP probabilistic topic model for topic computing, which mixed with time, area and behavior information. Secondly, the algorithm obtains the result of CTAP(new features and package topics) with package topics and features. Then combine user's consumption tendency informatio n to do cascade optimization, finally obtain the personalized recommendation list. Experimental results show that FACCF recommendation algorithm is more accurate than traditional methods. At last, use Hadoop distributed platform to run the FACCF algorithm for improving the time efficiency, and apply it into recommend system of C hina Telecom for personalized package recommendation.