Research On Dynamic Evolutionary Recommendation Algorithm For Community Question Answering

Human progress is inseparable from the inheritance and development of civilization.In this process of inheritance and development,the way of mutual learning allows knowledge to be spread from generation to generation.The simplest way to communicate is question and answer.However the current recommendation technology is dominated by the similarity of historical interests among users,the users' interest may change at any time in the question and answer(QA)community,and the dynamic change of users' interest will bring inaccuracy to the relevant recommendations of the QA community.In order to solve these problems,this paper proposes a modeling method based on time series of users' interest vector and designed a new type of users' interest change model from the users' various interest distribution and interest in the growth curve,to simulate the dynamic evolution of recommendation.The major contributions are as follows:(1)Modeling of the users' interest growth.By matching the words in topic dictionary,and the word vector similarity measure based on Word2 Vec,this paper models the users' interest,and obtain labels of different types of users by users' interest clustering.On this basis,by analyzing the questions that users have answered in the time series,this paper builds the users' interest vector based on the time sequence and applies it to the recommendation of users' friends and the questions.Finally,the recommendation results are optimized to some of the existing recommended algorithms,which are embodied on the recommended priority,and the evolutionary process naturally obtains to some community discovery functions.(2)Prediction of the best answer for questions.Based on the extraction of interest features of time series and the topic modeling of the questions,the best answer of the question is predicted.The feature of the user and the question is extracted based on the previous time series feature extraction method,and the SVR(support vector regression)algorithm is used to predict the like number of answers.Answers with a higher like number would be recommended to question issuers.(3)The dynamic evolution of ranking optimization for recommending friends.This paper proposes an evolutionary ranking algorithm for friend recommendation based on adaptive PSO(particle swarm optimization)evolution,and implements ranking optimization of recommendation priority on the initial recommendation result of users' friends,at the same time,some community discovery functions are implemented,such as friend recommendation delete function,potential friend discovery,etc.And uses variant EM algorithm to optimizeimportant parameters of the evolutionary algorithm.(4)The dynamic evolution of ranking optimization for recommending question answerers.This paper proposes an evolutionary ranking algorithm based on adaptive first-order Markov random walks to implements the optimization of priority,which is applied to the initial recommended results of question answerers,at the same time,we can realize some community discovery functions,such as finding latent users who are interested in problems,etc..And adopts the variant EM algorithm to optimize the important parameters of the evolution algorithm.The experiment is based on ZhiHu real data sets and the experimental results show that this paper puts forward are effectiveness.