Personalized Recommendation Mechanism For Knowledge Service System Research And Construction

Information explosion is a typical feature of the big data era.Learners can easily find a variety of knowledge information from the Internet,but the expansion of information also makes it difficult for learners to find the information they want and fall into the problem of knowledge overload.How to effectively understand the user's knowledge service needs,provide users with personalized knowledge recommendation services,and provide effective technical support and guarantee for the development of knowledge services has become a problem worthy of research.At present,when we carry out knowledge push services,it is often because of the sparseness of the data that we cannot fully tap the potential information of users and projects,which causes our push accuracy to decrease and affects the quality of our knowledge push.In order to solve these problems effectively,the following three parts of the research work are carried out:(1)A content-based optimization algorithm for knowledge services(CBOKS)is proposed.By synthesizing multiple data sources to extract features of documents,documents can be represented as the required indicator vector as accurately as possible.The extracted feature words are not always equivalent,and the literature also has its own focus,according to the rules to give its corresponding weight.The characteristics of users' preferences are not immutable,and the knowledge points concerned will be different in a certain period of time.In order to reflect the dynamic characteristics,users' preferences can be dynamically migrated to meet the actual needs of users.When recommending features for a user,the content tends to be unitary.In order to avoid this problem,we use the method of multi-user joint comparison to evaluate the candidate literature to increase the novelty.(2)A repeated prediction mechanism based on Collaborative matrix factorization(RPCMF)is proposed.Consider the fact that different users have different scoring schemes.In order to eliminate the shortcomings of vector based similarity,we subtract the average score of each user from each user's score of related items.Avoid the influence of similar user preferences on numerical prediction.At the same time,it integrates the advantages of various algorithms,makes full use of the processing effect of collaborative filtering method for sparse matrix and the high accuracy of matrix decomposition,and reevaluates the values with different evaluation values through the comparison mechanism,so as to achieve the purpose of accurate prediction and accurate push.(3)Based on the above two mechanisms,a personalized recommendation system for knowledge service is designed and constructed.The system architecture,design purpose and applicable objects are described in detail.This system mainly adopts the design mode of browser/server,mainly including user history display,personalized recommendation,literature display and other related functions.