Recommandation Teachnology Research Based On User Portrait

Micro network blog(Micro Blog)is a network platform for short message sharing through attention mechanism.Because of its characteristics of complex user network,short text information,redundant information and high timeliness of information,the traditional method of recommendation can not adapt well to the network of micro-blog environment.In order to alleviate the "information overload" effect to a certain extent and improve the user experience of Internet micro-blog,this paper studies the content structure characteristics of the text in the network blog,and combines the text classification algorithm and the user portrait model construction technology to improve the traditional user label based recommendation.A user portrait is constructed and designed based on time sequences growth and dynamic updating model label.After in-depth study,a recommendation algorithm based on dynamic user portrait is proposed and used as the core algorithm of the recommendation system.Finally,a dynamic model is designed and implemented.The system shows the feasibility of the algorithm by the off-line experiment,and proves that the proposed algorithm is better than the traditional recommendation method in the accuracy of recommendation.This article has mainly done the following three aspects:1.In obtaining the original text information for building the system,this paper deeply studies the web crawler technology based on WebMagic,describes the principle,work flow and deployment method of the network crawler,and constructs a web crawler real example using JAVA language on the Windows platform to mine the text required for the recommendation system.2.In the research process of constructing the user portrait text label model,this paper deeply studied the text categorization based on SVM and describes the text classification process,algorithm principles,advantages and disadvantages of the algorithm in detail.Secondly,this paper deeply studied the related technology of user portrait modeling,describes the concept of user portrait and the user portrait are constructed.On this basis,this paper proposes a user portrait based on time sequence and a recommendation algorithm based on the dynamic user portrait.The algorithm analyzes user's text labels through time periods,updates user labels or label weights.3.In this paper,the recommendation system and related technologies arestudied,the attributes of the recommended system and the evaluation experiments are introduced in detail,and the principle and process of the current mainstream recommendation algorithms are described in detail.Based on the recommendation algorithm,this paper builds a basic recommendation system based on the recommendation algorithm.The system builds the user's basic portrait model by processing the user text information and dynamically updates the user portrait with the time sequences,thus improving the accuracy of the recommendation.