A Novel DIT User Preference Prediction Model Combing Social Network Information

Under the social network environment,the information overload is an urgent problem to be solved.In this background,that personalized recommendation technology has been widely used.Traditional personalized recommendations based on social networks,mainly include user based and item based collaborative filtering recommendation,demographic based recommendations.Although great progress has been made,on this basis,has been provided a rich and varied personalized service for users,but there are still a lot of limitations.Based on the social network document attribute,social network text information and social network interactive behavior,this paper proposes a method of DIT user preference prediction based on social network information.This method takes into account the document attributes in the social network,the text information in the social network and the interactive behavior in the social network,by constructing user portraits,obtaining the user's points of interest and speculating on the user's affective tendencies,obtained the user's preference predication on certain types of products.The main contents of this paper include the following aspects:(1)Construct a DIT user preference prediction model that incorporates social network information.The model takes into account the document attributes,text information and interactive behavior in the social network,and calculates the user's preference prediction for a certain type of products.The model is composed of three sub-models weighted by organic means,compared with a single dimension preference prediction model,the prediction result is more accurate than that.(2)Construct a user preference prediction model based on social network document attributes.Taking into account the user's gender,age,location,number of friends and other factors,the product preference prediction based on social network document attributes is calculated.(3)Construct a user preference prediction model based on social network text information.Considering the emotional tendency and theme features of the text message,the product preference prediction based on the social network text information is calculated.(4)Construct a user preference prediction model based on social network interactive behavior.Taking into account the user's likes and dislikes,forwarding and friendships on the social network,the product preference prediction based on social network interaction is calculated.(5)In order to evaluate the model's prediction better,the existing datasets from July 1,2015 to June 30,2017 are respectively divided into weeks,months,quarters and years,and the adjacent cycle datasets are used as a training set and a test set respectively to calculate the prediction accuracy in each cycle.